Add mdoc as input format (Evan Silberman). This change introduces a reader for mdoc, a roff-derived semantic markup language for manual pages. This reader has been developed almost exclusively against mandoc’s documentation and implementation of mdoc as a reference, and the real-world manual pages tested against are those from the OpenBSD base system. Of ~3500 manuals in mdoc format shipped with a fresh OpenBSD install, 17 cause the mdoc reader to exit with a parse error. Any further chasing of edge cases is deferred to future work.

New module: Text.Pandoc.Readers.Mdoc, exporting readMdoc [API change].

Issue warnings for duplicate YAML metadata keys (#10312).

Ensure that --sandbox affects --embed-resources. Previously it did not (contrary to what was implied by the manual), which means that an image with URL /etc/passwd would leak an encoded version of that file to HTML output with --self-contained or --embed-resources, even if --sandbox was used. Thanks to Samuel Mortenson for pointing out the issue.

Text.Pandoc.App.OutputSettings: add sandbox' function. This computes the sandboxed files from Opt and avoids code repetition.

Docx reader:

• Parse index references as empty spans with attributes (#10171). Attributes included are entry, and optionally bold, italic, yomi, see.
• Don’t create multiple paragraphs for title or subtitle (#10359). If there are multiple paragraphs with Title or Subtitle style, use only the first for metadata.
• Handle case where Zotero itemData has different id from the citationItem id. In this case we use the citationItemId in the bibliography as well, overriding the referenceId in the itemData (#10366).

LaTeX reader:

• Put parsed minipage in specially marked Div (#10266).

HTML reader:

• Parse footnotes defined by dpub-aria roles (#5294).

MediaWiki reader:

• Fix indented tables with caption (#10390).
• Fix parsing of col/rowspan (#6992).

Typst reader:

• Avoid generating empty paragraphs.
• Support underparen, overparen.
• Fix #quote attribution. If attribution is not present, don’t print the -- (#10320).
• Fix typo in unicode code point for em dash (see #10320).

Commonmark reader:

• implicit_figures should check for empty caption and not produce an implicit figure in this case (#10429).

RST reader:

• Use a new one-pass parsing strategy. Instead of having an initial pass where we collect reference definitions, we create links with target ##SUBST##something or ##REF##something or ##NOTE##something, and resolve these in a pass over the parsed AST. This allows us to handle link references that are not at the top level (#10281).
• Ignore newlines in URL in explicit link (#10279).
• Handle block level substitutions.
• Support :file: on raw directive (#8584).
• Implement option lists (#10318).
• Avoid putting metadata in Para (#7766). Create MetaInlines when possible, just as with markdown input. MetaBlocks is still used when there are multiple paragraphs or non-paragraph content. This change also affects field lists.
• Fix linked substitutions (#6588). E.g. |Python|_.
• Support inline anchors (#9196).
• Explicit links define references (#5081). For example, Go to `g`_ `g <www.example.com>`_. should produce two links to www.example.com.

EPUB writer:

• Use standardized filename for cover image instead of the original name (#10404). This avoids problems with e.g. filenames containing spaces.

Markdown writer:

• Issue INFO warning when not rendering table, e.g., when raw_html is disabled and the table can’t be fit into a supported markdown table format (#10407).
• Respect empty LineBlock lines in plain output (Evan Silberman). The plain writer behaved as a markdown variant with Ext_line_blocks turned off, and so empty lines in a line block would get eliminated.

LaTeX writer:

Ensure that beamer footnotes go on frame, not column (#5769).

HTML writer:

• Unwrap empty incremental divs (#10328, Albert Krewinkel). Divs are unwrapped if the only purpose of the div seems to be to control whether lists are presented incrementally on slides.

Typst writer:

• Make template sensitive to a page-numbering variable (#10370). This can be set to an empty string (or, in metadata, to false) for no page numbers.
• Make smart extension work (#10271). If smart is not enabled, a command in the default template will disable smartquote substitutions. When smart is enabled, render curly apostrophes as straight and escape straight apostrophes. When smart is disabled, render curly apostrophes as curly and don’t escape straight apostrophes. Similarly for quotes, em and en dashes. This should give more idiomatic typst output, with fewer unnecessary escapes.

ANSI writer:

• Respect empty LineBlock lines (Evan Silberman).

JATS writer:

• Correct spelling of suppress attribute (#10350, Andreas Deininger).

Typst template:

• Remove definitions.typst partial.
• Remove unnecessary definition of endnote.
• Incorporate the one remaining definition into default.typst.
• Use typst 0.12 code for two column layout (#10294, Luis Rivera).
• Note: the new templates presuppose typst 0.12; if you try to use an earlier version of typst, an error will be raised.

LaTeX/Beamer template:

• Split fonts.latex partial into two parts: fonts.latex and font-settings.latex.
• In beamer template, load beamer theme between fonts.latex and font-settings.latex. This allows a theme (such as metropolis) to set its own default font, while still allowing the user to override it. This fixes a regression in pandoc 3.5 (#10297).
• Note: Users who have custom templates based on pandoc 3.5 templates will need to add font-settings.latex() after fonts.latex() in the latex template. In a beamer template, the beamer theme-setting code needs to be moved between these two partials.

ConTeXt template: Ensure that font names don’t wrap (#10305).

epub.css: remove background-color (#10264, Suraj Patil). With this greyish background color, epubs look bad on a Kindle (#10263).

Text.Pandoc.ImageSize: add WebP support (Evan Silberman, #10397). Add Webp constructor on ImageType [API change].

Text.Pandoc.Readers.Roff and a new unexported module Text.Pandoc.Readers.Roff.Escape: parameterize Roff escaping (Evan Silberman) [API change]. This allows code to be reused between the mdoc and man readers, despite the differing Token types.

Text.Pandoc.PDF:

• PDF via LaTeX: always do max runs if toc is present (#10308). The old method (checking to see if toc hash had changed) is not completely reliable.
• Use .source extension, not .html, in toPdfViaTempFile (#10314).

Text.Pandoc.Logging: add YamlWarning constructor to LogMessage [API change] (#10312).

Text.Pandoc.Format: remove duplicate typst entry (#10388, Caleb Mclennan).

Fix a typo in the ua.yaml localization for ‘See’ (Jens).

Lua subsystem (Albert Krewinkel):

• Remove prefixes from Lua type names (#8574). Lua type names were inconsistent with regard to the use of prefixes; all prefixes are removed now, and Lua types now have the same name as the Haskell types. The use of app-specific prefixes is suggested by the Lua manual to avoid collisions. However, this shouldn’t be a problem with pandoc, as it cannot be used as a Lua package.

doc/libraries.md: Add newly developed Haskell packages. Sort list alphabetically (Albert Krewinkel).

doc/lua-filters.md: document pandoc.List:iter method (Albert Krewinkel). List objects have a new function iter that returns an iterator function that returns the next list item on each call.

MANUAL.txt:

• Clarify what the example of YAML EPUB metadata shows (#10405).
• Fix typo in template syntax (#10265, Pascal Wagler).
• Update manual with information on openxml template (#10273).
• Clarify that --variable can only assign string values (#10298).

Fix comments in TEI writer referring to DocBook (#10430, Evan Silberman).

Fix several typos in documentation (#10349, Andreas Deininger).

Allow Diff 1.0.

Add font-settings.latex partial to pandoc.cabal (#10379).

Bump upper bound for data-default.

Use latest typst, texmath, pandoc-lua-marshal, commonmark-pandoc, commonmark-extensions, skylighting, skylighting-format-blaze-html.

Add command-line options --list-of-figures/--lof and --list-of-tables/--lot (#10029, Akash Patel). Only docx, latex, and context are affected by these options currently. Setting the lof and lot variables will also work for the formats that are currently supported.

Defaults files: interpolation of environment variables now works for to and from fields (#8024). This is needed because these files can contain paths of custom readers/writers.

Docx reader:

• Reset lists after headers in same list numId (#10258). To accomplish this, we add a Heading constructor to BodyPart and include on it all the information list items have.

DocBook reader:

• Parse id, class, and tabstyle on tables (#10181, Erik Rask). Add parsing of id (xml:id), class, and tabstyle XML attributes for table and informaltable in the DocBook reader. The tabstyle value is put in the ‘custom-style’ attribute.

Dokuwiki reader:

• Be more forgiving about misaligned lists, like dokuwiki itself (#8863).
• Improve blockquote parsing in dokuwiki. Allow for quoted code blocks.
• Enable smart extension.
• Properly parse -- and --- as dashes.
• Fix block quote behavior (#6461). Blockquotes are not really block containers in DokuWiki; the lines are interpreted literally (so, e.g., you can’t start a list), and line breaks are added at the ends.

EPUB reader:

• Fix links to other files in the EPUB, making them internal links to a fragment derived from the filename (#10207). There was already code to handle links like #foo, but not to handle links like ch0001.html#foo.

LaTeX reader:

• Add em, ex, px, mu to list of units for dimension args (#10212).

ANSI writer:

• Fix subscripts (Evan Silberman).

DokuWiki writer:

• Don’t emit <HTML> tags (#7413). The use of these tags is now strongly discouraged for security reasons, and will be removed. We previously used them as a fallback for lists that could not be represented using DokuWiki syntax, e.g. ordered lists with fancy numbers or lists with multiple blocks in their items. We also used them for block quotes with multiple blocks as their contents. We now use the <WRAP> syntax (from the optional WRAP plugin) to handle lists with multiple blocks as their contents. A new method of handling block quotes with complex contents has the side benefit of also handling nested block quotes, which weren’t supported before. <HTML> and <html> tags are only for raw HTML blocks and inlines, and only if the raw_html extension is enabled. (It is now a valid extension for dokuwiki, though off by default.)

Docx writer:

• Support --list-of-figures and --list-of-tables (or lof and lot variables) (Akash Patel).

HTML writer:

• Don’t emit missing title/lang warnings if templates does not contain the pagetitle or lang variables respectively (#9370).

LaTeX writer:

• Better fix for lists in definition lists (#10241). In commit a26ec96d89ccf532f7bca7591c96ba30d8544e4a we added an empty \item[] to the beginning of a list that occurs first in a definition list, to avoid having one item on the line with the label. This gave bad results in some cases (#10241) and there is a more idiomatic solution anyway: using \hfill.
• Avoid error on refs div with empty citations (#10185). If there are no citations, don’t emit an empty CSLReferences environment.

RST writer:

• Change bullet list hang from 3 to 2. This accords with the style in the RST reference docs.
• Handle cases where indented context starts with block quote (#10236). In these cases we emit an empty comment to fix the point from which indentation is measured; otherwise the block quote is not parsed as a block quote. This affects list items and admonitions.
• Don’t enclose the list table in a .. table::; this leads to doubled captions (#10226).
• Fix alignment of list table items corresponding to cells (#10227).

JATS template:

• Support floats-group (Albert Krewinkel, see #10196). The content of the floats-group variable is now rendered in a <floats-group> element when using the publishing or archiving tag sets.

LaTeX and Beamer templates:

• Split old default.latex into two templates, default.latex and default.beamer, factoring common parts into partials: fonts.latex, common.latex, passoptions.latex, hypersetup.latex, after-header-includes.latex.
• Make default.beamer the default template for beamer.
• Add shorttitle, shortsubtitle, shortauthor, shortinstitute, shortdate variables to beamer template (#10248, Thomas Hodgson).
• Make --number-sections work with beamer (#12045, Thomas Hodgson).
• Support a list of images for titlegraphic in beamer template (#10246, Thomas Hodgson). Title graphic options will be applied to each title graphic. Images will be separated by \enspace.
• Beamer theme options (#10243)
• Add theme options to beamer template: colorthemeoptions, fontthemeoptions, innerthemeoptions, outerthemeoptions (#10243, Thomas Hodgson).
• Don’t load amsmath, amssym in beamer template. These are loaded by beamer automatically.

Text.Pandoc.SelfContained:

• Improve handling of links to remote CSS (#10261).

Text.Pandoc.Class:

• Allow extracting data: URIs even in PandocPure (--sandbox) (#10249).
• Export extractURIData [API change].

Text.Pandoc.PDF:

• Read .toc and .log files from output directory (#10186). When this is different from the input directory, this is where .toc and .log files are written.

Text.Pandoc.Shared:

• Modify addPandocAttributes for changes in commonmark-pandoc. The new commonmark-pandoc version automatically adds the attribute wrapper="1" on all Divs and Spans that are introduced just as containers for attributes that belong properly to their contents. So we don’t need to add the attribute here. This gives much better results in some cases. Previously the wrapper attribute was being added even for explicit Divs and Spans in djot, but it is not needed in these cases.

Text.Pandoc.Options:

• Add writerListOfFigures and writerListOfTables fields to WriterOptions (#8245, Akash Patel). [API change]

Text.Pandoc.App:

• Add optListOfFigures and optListOfTables to Opt (#8245) [API change].

Lua subsystem (Albert Krewinkel):

• Update List module (#9835). The module now comes with a method :at(index[, def]) that allows to access indices, accepts negative indices to count from the end, and will return the def value as a default if the list has no item at the given position. Furthermore, the list constructor pandoc.List now accepts iterators. E.g., pandoc.List(text:gmatch '%S+') returns the list of words in text.

• Support character styling via pandoc.layout. The Doc values produced and handled by the pandoc.layout module can now be styled using bold, italic, underlined, or strikeout. The style is ignored in normal rendering, but becomes visible when rendering to ANSI output. The pandoc.layout.render function now takes a third parameter that defines the output style, either plain or ansi.

• It is now possible to return a single filter from a filter file, e.g.

  -- Switch single- and double quotes
  return {
    Quoted = function (q)
      elem.quotetype = elem.quotetype == 'SingleQuote'
        and 'DoubleQuote' or 'SingleQuote'
      return elem
    end
  }

  The filter must not contain numerical indexes, or it might be treated as a list of filters.

• Add list_of_figures and list_of_tables to writer options (Akash Patel).

Use latest releases of commonmark, commonmark-pandoc, texmath, djot.

Stop depending on package SHA (Albert Krewinkel). Use crypton instead.

linux/make_artifacts.sh: add riscv64 support (Olivier Benz).

Fix invalid XML in test/docx/normalize.docx (#10242).

doc/lua-filters.md: list functions in pandoc.utils alphabetically (Albert Krewinkel).

MANUAL.txt:

• Clarify use of beamerarticle variable (#10250).
• Add clarification to address user issues like #6704 (Yehuda Katz).

New output format: ansi (for formatted console output) (Evan Silberman). Most Pandoc elements are supported and printed in a reasonable way, if not always ideally. This version does no detection of terminal capabilities, nor does it fall back to different output styles for less-capable terminals.

Add command line options --table-caption-position and --figure-caption-position. These allow the user to specify whether to put captions above or below tables and figures, respectively. The following output formats are supported: HTML (and related such as EPUB), LaTeX (and Beamer), Docx, ODT/OpenDocument, Typst.

Change default --pdf-engine via HTML to WeasyPrint (#10142). wkhtmltopdf is deprecated. weasyprint is the easiest-to-install, maintained alternative. For better results, one might prefer pagedjs-cli.

Org reader:

• Fix parsing of src blocks with an -i flag (#10071, Albert Krewinkel). Tabs are now preserved in the contents of src blocks if the the block has the -i flag.

RTF reader:

• Handle images inside shp contexts (#10145).

RST reader:

• Improve simple table support (#10093). Multiline rows occur only when the first cell is empty; we were previously treating lines with any empty cell as row continuations. In addition, we no longer wrap multiline cells in Para if they can be represented as Plain. This is consistent with docutils behavior.

LaTeX reader:

• Math environments don’t have bracketed options (#10160).
• Parse nested tabular environments (#4746).

Typst reader:

• Change how “block” elements are handled. Previously they were always parsed as divs. But actually they can occur in some “inline” contexts. Now we first try to parse them as inlines, and only as blocks if that fails. A surrounding Div or Span element is added only if there is an identifier.

HTML reader:

• Only parse main element’s contents (if present) (#10140). If main has an id or class, we include a div with that id or class; otherwise just the contents.
• Read TeX annotation in MathML content if present (#9971).
• Better handle KaTeX-generated math (#9971). KaTeX emits the mathml followed by a span with an HTML fallback. Previously pandoc was converting both. We now ignore the HTML fallback span, marked with class katex-html.

Docx reader:

• Add “SuppressAuthor” and “AuthorOnly” to citationMode when +citations is used (thomjur).

New module: Text.Pandoc.Writers.ANSI [API change] (Evan Silberman).

Docx writer:

• Support custom-style attribute for docx table (Sebbones).
• Support --number-offsets.
• Make table/figure rendering sensitive to caption position settings.

OpenDocument writer:

• Make table/figure rendering sensitive to caption position settings.

Typst writer/template:

• Implement figure caption positions by triggering a show rule in the default template, which determines caption positions for figures and tables globally.
• Don’t include trailing semicolon after @ style citations with suffixes (#10148).
• Template: move header-includes before show doc (#9996, Gordon Woodhull).

LaTeX writer:

• Make table/figure rendering sensitive to caption position settings (#5116).
• Preserve locator labels with --natbib (#10057).

HTML writer/template:

• Make <figcaption> placement sensitive to caption position settings. For tables, <caption> must be the first element, and positioning is determined by CSS, for here we set a variable which the default template is sensitive to.
• Use makeSectionsWithOffsets for writerNumberOffsets, instead of the old, inefficient code.
• Don’t add doc-biblioref role to every link in a citation; only to links to the bibliography (#10156).
• Add data- when rendering label attribute (#10048).

Markdown writer:

• Avoid emitting markdown caption if table has fallen back to raw HTML, which will then contain a <caption> tag (#10094).

• Make math sensitive to tex_math_gfm extension (#9121). This means that in GFM output, the “new style” math will be used by default, e.g.

  $`x=y`$
  
  ```math
  x = y
  ```

  To defeat this and get the older behavior, namely

  $x=y$
  
  $$x=y$$

  one could use -t gfm-tex_math_gfm.

AsciiDoc writer:

• Add link: prefix when needed (#10105). AsciiDoc requires it except for http, https, irc, mailto, ftp schemes (#10105).
• Preserve original base level (#10062). We used to normalize so that the base level is always 1, but asciidoc no longer seems to care about that, and the behavior creates difficulties when we are converting fragments.
• Don’t emit empty figure caption (#10047).

ODT writer:

• Add TableCaption to styles.xml (#10058, Ian Max Andolina).

LaTeX template:

• Fix wrong beamer color in (sub)section page (Jonathan).

Text.Pandoc.Options:

• Add CaptionPosition and new WriterOptions fields writerFigureCaptionPosition and writerTableCaptionPosition [API change].

Text.Pandoc.Opt:

• Change default for optNumberOffset to []. This behaves the same as [0,0,0,0,0].
• Add Opt fields optFigureCaptionPosition and optTableCaptionPosition [API change].

Text.Pandoc.Format: change formatFromFilePaths so that it is smarter about URLs. URLs are parsed, and we take the format from the path component, if present (#10141). This means that https://emacs.org/ will be treated as HTML, while https://emacs.org/sample.org will be treated as Org.

Text.Pandoc.URI:

• Add unofficial gemini: to list of URI schemes (Pau RE).

Text.Pandoc.Shared:

• Add makeSectionsWithOffsets [API change].
• Remove `stripEmptyParagraphs [API change] (Albert Krewinkel). This function is no longer used.

Text.Pandoc.Highlighting: Expose formatANSI [API change] (Evan Silberman).

Text.Pandoc.Writers.Shared: export to{Sub,Super}scriptInline [API change] (Evan Silberman).

Remove use of partial functions (e.g. head) in code.

Use latest skylighting-core, skylighting, doclayout, texmath, typst.

pandoc-lua-engine: Add accessors for several writer options, including some that were added in previous releases.

pandoc-server: Initialize some missing fields in WriterOptions: writerEpubTitlePage, writerChunkTemplate, writerListTables, writerFigureCaptionPosition, writerTableCaptionPosition.

CONTRIBUTING.md: Summarize steps for adding a new cli option.

MANUAL.txt:

• Clarify that the --number-offset option should only directly affect numbering of the first section heading in a document; subsequent headings will increment normally.
• Fix asciidoc link (#10039).
• Fix CSL Docs broken link (#10100, Tristano Ajmone).
• Document the use of luatexja when CJKmainfont is used with lualatex (#3873, Kolen Cheung).
• Add a citations (typst) section to the manual (#9127).
• Clarify that citations affects both input and output for org.
• Add note on --citeproc that you may need to disable citations extension on the output format (e.g., -t markdown-citations) to see the rendered citation (#9127, #10012).

INSTALL.md — reorganise info on static binaries and add conda-forge install options (#10098, #10069, Ian Max Andolina).

New cli option: --link-images. This causes images to be linked rather than embedded in ODT.

Allow --number-sections to take an optional true|false argument.

RTF reader:

• Handle \*\shppict without dropping image (#10025).

TWiki Reader:

• Recognize WikiWords as internal links (#9941).
• Avoid partial function.

Typst reader:

• Ignore ‘pad’ and just parse its body (#9958).
• Use typst 0.5.0.5. Fixes parsing of equations like $1.$.

Docx writer:

• Fix regression with nested lists (#9994). The bug affects e.g. ordered lists with bullet sublists; after the sublist the top-level list reverts to bullets instead of being properly numbered. This is a regression introduced in version 3.2.1.

BibTeX writer:

• Ensure that “literal” names are enclosed in braces (#9987).

Man writer:

• Use default middle header when metadata does not include header (#9943). This change causes pandoc to omit the middle header parameter when header is not set, rather than emitting "". The parameter is optional and man will use a default based on the section if it is not specified.

HTML templates: don’t load polyfill (#9918). This was added in a period when MathJaX required polyfill. MathJaX no longer recommends this and polyfill should no longer be necessary on any reasonably modern browser.

Translations:

• Add ua.yaml (Jens Oehlschlägel).
• Add a script (tools/update-translations.py) and Makefile target (update-translations) to update translation data automatically from babel and polyglossia upstream (Stephen Huan).
• Use this script to update language data, increasing the number of languages we cover (Stephen Huan). Fix a few small bugs in existing translations.

Fix some mistakes with Japanese language code (#9938). In several places we were mistakenly assuming that the BCP 47 code for Japanese language was jp. It is ja.

Text.Pandoc.Options:

• New field in WriterOptions: writerLinkImages [API change] (#9815).

Text.Pandoc.App.Opt:

• New field in Opt: optLinkImages [API change] (#9815).

Lua subsystem:

• Keep lpeg and re as “loaded” modules (Albert Krewinkel). The modules lpeg and re are now treated as if they had been loaded with require. Previously the modules were only assigned to global values, but could be loaded again via require, thereby allowing to use a system-wide installation. However, this proved to be confusing.

  The old behavior can be restored by adding the following lines to the top of Lua scripts, or to the init.lua in the data dir.

  debug.registry()['_LOADED'].lpeg = nil
  debug.registry()['_LOADED'].re = nil

pandoc-cli: Include pandoc copyright in Lua version info (Albert Krewinkel).

pandoc-cli: Refer printing of version info to the Lua interpreter (Albert Krewinkel). The Lua interpreter no longer terminates when called with -v or --version arguments, thus improving compatibility with the default lua interpreter program.

Avoid partial functions in JATS reader, DocBook writer, Haddock reader.

Allow tls 2.1.x.

MANUAL.txt:

• Make documentation of extensions clearer (#9060).
• Fix section level for two Extensions entries.

lua-filters.md: Partially autogenerate docs for module pandoc (Albert Krewinkel). The documentation system isn’t powerful enough to generate the full documentation automatically.

Fix gfm_auto_identifiers to replace emojis with their aliases, as documented (#9876).

CSV reader:

• Turn line breaks into LineBreaks not SoftBreaks (#9797).

Docx reader:

• Support task lists (#8211).
• Fix a small bug in parsing delimiters in numbered lists, which led to the default delimiter being used wrongly in some cases.
• Improve handling of captions.
  • Turn captioned images into Figure elements. Closes #9391.
  • Improve the logic for associating elements with captions (#9358).
  • Ensure that captions that can’t be associated with an element aren’t just silently dropped (#9610).
• Support HorizontalRule. We support both pandoc-style and the style described on a Microsoft support page, an empty paragraph with a bottom border (#6285).
• React to "left" value on jc attribute.
• Handle column and cell alignments (#8551). We take the column alignments from the first body row.
• Fix a bug that caused comments inside insertions or deletions to be ignored (#9833).

HTML reader:

• Better handle non-li elements in ul and ol (#9809). For example, a p after a closed li will be incorporated into the previous li. This mirrors what browsers do with this invalid HTML.

LaTeX reader:

• Fix parsing of dimensions beginning with ., e.g. \kern.1pt (#9902).

Markdown reader:

• Allow author-only textual citations (#7219). E.g. -@reese2002 outside of brackets.

RST reader:

• Tighten up rules for when emphasis can start (#9805).
• Support :cite: role with citeproc (#9904). A subset of the functionality of the sphinxcontrib-bibtex extension to Sphinx is supported.

Textile reader:

• Don’t let spans begin right after a symbol (#9878).

Typst reader:

• Fix an incomplete pattern match (#9807).

• Handle inline bodies ending in a parbreak. E.g.

  `#strong[
  test
  ]

ConTeXt template: remove \setupbackend[export=yes] (#9820).

Docx writer:

• Allow OpenXML templates to be used with docx (#8338, #9069, #7256, #2928). The --reference-doc option allows customization of styles in docx output, but it does not allow one to adjust the content of the output (e.g., changing the order in which metadata, the table of contents, and the body of the document are displayed), or adding boilerplate text before or after the document body. For these changes, one can now use --template with an OpenXML template. (See the default openxml template for a sample.) --include-before-body and --include-after-body can also now be used with docx output. The included files must be OpenXML fragments suitable for inclusion in the document body.
• New unexported module Text.Pandoc.Writers.Docx.OpenXML.
• Omit jc attribute on table cells with AlignDefault (#5662).
• Better formatting for task lists. Task lists are now properly formatted, with no bullet (#5198).
• Replace an expensive generic traverse to remove Space elements, for better performance.
• Wrap figures with id in a bookmark (#8662).
• Add eastAsia font hints to w:r (#9817). We do this when the text in the run contains any CJK characters. This ensures that ambiguous code points (e.g. quotation marks) will be represented as “wide” characters when together with CJK characters.
• Clean up Abstract Title and Subtitle in default reference docx. Center Subtitle, remove color.

HTML writer:

• Ensure URI escaping needed for html4 (#9905). Unicode characters need not be escaped for html5, and still won’t be.
• Don’t emit unnecessary classes in HTML tables (#9325, Thomas Soeiro). Pandoc used to emit a header class on the tr element that forms the table header. This is no longer needed, because head > tr will do the same thing. Similarly, pandoc used to emit even and odd classes on trs, allowing striped styling. This is no longer needed, because one can use e.g. tbody tr:nth-child(2n). Compatibility warning: users who relied on these classes to style tables may need to adjust their CSS.

JATS writer:

• Support supplementary-material in metadata for jats_articlepublishing (#9818).

LaTeX writer:

• New method for ensuring images don’t overflow (#9660). Previously we relied on graphicx internals and made global changes to Gin to force images to be resized if they exceed textwidth. This approach is brittle and caused problems with \includesvg (see #9660). The new approach uses a new macro \pandocbounded that is now defined in the LaTeX template. (Thanks here to Falk Hanisch in https://github.com/mrpiggi/svg/issues/60.) The LaTeX writer has been changed to enclose \includegraphics and \includesvg commands in this macro when they don’t explicitly specify a width or height. In addition, the writer now adds keepaspectratio to the \includegraphics or \includesvg options if height is specified without width, or vice versa. Previously, this was set in the preamble as a global option. Users should attend to the following compatibility issues:
  • If custom templates are used with the new LaTeX writer, they will have to be updated to include the new \pandocbounded macro, or an error will be raised because of the undefined macro.
  • Documents that specify explicit dimensions for an image may render differently, if the dimensions are greater than the line width or page height. Previously pandoc would shrink these images to fit, but the new behavior takes the specified dimensions literally. In addition, pandoc previously always enforced keepaspectratio, even when width and height were both specified, so images with width and height specified that do not conform to their intrinsic aspect ratio will appear differently.
• Task lists must be unordered (#9185).
• Specify language option for selnolig and only include it if english or german is used (#9863). (This includes changes to the LaTeX template.) This should restore proper ligature suppression when lualatex is used.
• Fix --toc-depth with beamer output (#9861). Previously only top-level sections were ever included in the TOC, regardless of the setting of --toc-depth.
• Use \linewidth instead of \columnwidth or \textwidth for resizing figures, table cells, etc. in LaTeX (#9775). \linewidth, unlike the others, is sensitive to indented environments like lists.

LaTeX template: put babel-lang in options to beamer (#9868). This is required to make beamer use proper localized terms for things like “Section.”

Markdown writer:

• Don’t print extra caption when using implicit_figures.
• Ensure blank line after HTML blocks in commonmark-based formats (#9792).
• Fix bug rendering block quotes in lists (#9908).

Texinfo writer:

• Ensure proper escaping in all node/link contexts.
• Target node rather than anchor when possible in internal links.
• Remove illegal characters from internal link anchors (#6177).
• Use two commas not one in @ref.
• Don’t add anchors to headings. We don’t need them, now that we make internal links use the node.
• Avoid duplicate node names.
• Improve menus. Properly handle the case where the node name is different from the descriptive title.

Texinfo template: add variables for filename and version.

Typst writer:

• Support ‘.typst:no-figure’ and ‘typst:figure:kind=kind’ attributes (#9778, Carlos Scheidegger). This extends support for fine-grained properties in Typst. If the typst:no-figure class is present on a Table, the table will not be placed in a figure. If the typst:figure:kind attribute is present, its value will be used for the figure’s kind (#9777). These features are documented in doc/typst-property-output.md.

Typst template:

• Add subtitle (#9747, Mickaël Canouil).
• Use content rather than string for title, author, date, email (#9823). This allows formatting in title, author, date, and email fields. Since the PDF metadata requires a string, and typst only converts the title to a string (not the authors), we use

Textile writer:

• Get rid of header, odd, even classes on tr (#9376).

Text.Pandoc.Class:

• fillMediaBag: Convert IOErrors to warnings when fetching absolute paths (#9859, Albert Krewinkel). This will allow many conversions that would have failed with an error to succeed (albeit without images or other needed resources).

Text.Pandoc.ImageSize:

• Don’t prefer exif width/height when they conflict with image width/height (#9871). That was a mistaken call in #6936. Usually when these values disagree, it is because the image has been resized by a tool that leaves the original exif values the same, so the width/height metadata are more likely to be correct that exif width/height.

Text.Pandoc.SelfContained:

• Strip CRs from XML before base64 encoding for data URI (so tests can work on Windows).
• Only create <svg> elements for SVG images when the image has the class inline-svg. Otherwise just use a data URI as we do with other images (#9787).

Lua subsystem (Albert Krewinkel):

• Split Init module into more modules. The module has grown unwieldy and is therefore split into three internal Haskell modules, Init, Module, and Run.
• Add function pandoc.utils.run_lua_filter (#9803).
• Add function pandoc.template.get (#9854, co-authored by Carsten Gips). The function allows to specify a template with the same argument value that would be used with the --template command line parameter.
• Keep CommonState object in the registry. The state is an internal value and should be treated as such. The PANDOC_STATE global is merely a copy; unsetting the global no longer breaks the Lua engine.
• Allow passing an environment to run_lua_filter. The default is now to use a copy of the global environment when running a filter; this ensures better separation when run_lua_filter is used multiple times. A custom environment can be specified via the optional third parameter.
• Set pandoc.List as default metatable for JSON lists (#9834). Lists created by pandoc.json.decode now behave like lists generated via pandoc.List. This also ensures that pandoc.List tables are encoded as JSON arrays when passed to pandoc.json.encode.

Text.Pandoc.Writers.Shared: export toTaskListItem [API change].

Add unexported module Text.Pandoc.Char. This exports isCJK. Use this instead of locally defined isCJK in T.P.Readers.MediaWiki.

MANUAL.txt:

• Remove false claim that Lua mode does not support -i (#9757, Ian Max Andolina).
• Use level-3 headings for extensions (to avoid gaps).
• Add anchor for tagging extension.
• Remove explicit referencess to generate anchors. These will be linkified automatically.
• Fixed links to option--reference-doc.
• Add a note that column widths aren’t supported in pptx for Divs with class columns (#9890).
• Fix alerts example (#9826, Ian Max Andolina).
• Fix markup of babelfonts example code (Albert Krewinkel).

doc/custom-writers.md:

• Fix usage of Template in example (Albert Krewinkel).
• Document the separator arg of Writer.Blocks (Albert Krewinkel).

doc/lua-filters.md (Albert Krewinkel):

• Fix outdated documentation for math and quoting functions and fields.
• Autogenerate docs for module pandoc.template and pandoc.layout.
• Document operators of the “Doc” type.

pandoc-lua-engine: depend on pandoc >= 3.2 (see #9755).

Allow crypton-connection 0.4, time 1.14.

Allow tasty-quickcheck 0.11.

Use latest emojis, skylighting, skylighting-core, citeproc, djot, commonmark-extensions, typst-hs

Change to --file-scope behavior (#8741): previously a Div with an identifier derived from the filename would be added around the contents of each file. This caused problems for “chunking” files into chapters, e.g. in EPUB. We no longer add the surrounding Div. This cooperates better with chunking. Note, however, that if you have relied on the old behavior to link to the beginning of the contents of a file using its filename as identifier, that will no longer work.

Markdown reader:

• Allow repeated labels in numbered example lists. Previously if you tried to use the same label as an earlier example list item, you’d get a new number, not the old one, and references to the label would go to the second occurrence. Now an existing label will be reused, and no new number will be generated. Caveat: this only works reliably when the re-used example list item occurs by itself in a list, or occurs in a list of previously used example list items that occur in exactly the same order as previously.
• Fix normalCite so it doesn’t consume past a closing ] boundary (#9710). This was causing an exponential performance bug on long lists of links containing potential emphasis characters.
• Generalize inlinesInBalancedBrackets to inBalancedBrackets, with a parameter for the inner parser.
• Auto-close unclosed divs (#9635). This applies to both fenced and HTML-ish varieties. Otherwise we face an exponential performance problem with backtracking. A warning is issued when a div is implicitly closed.

RST reader:

• Fix figclass and align annotations for figures (#7473, Gokul Rajiv).

LaTeX writer:

• Use polytonicgreek instead of polutonikogreek with babel (#9698). polutonikogreek is outdated. Also recognize both in the LaTeX reader.
• Improve treatment of math inside soul commands (#1294, #5529). soul commands (ul, hl, st) are very fragile and the math must be handled specially.

LaTeX reader:

• Fix over-eager macro expansion in conditionals (#9676).
• Parse flalign, flalign* math environments (#9679). We parse these as Math elements with an aligned environment. Semantically it’s not exactly the same, but better than falling back to raw LaTeX.

LaTeX template: add titlegraphicoptions variable (#9207, Guilhem Saurel).

Docx reader:

• Issue warning rather than error when we can’t parse EndNote citations (see #8433).
• Fix anchor in header after anchor (#9626, mbracke).

RTF reader:

• Don’t try to handle non-default code pages (#9683). Emit a warning instead.

OpenDocument writer:

• Implement custom-style for spans (#9657).

Typst writer:

• Add blank line in definition lists with multiple definitions (see #9704).
• Property output (#9648, Gordon Woodhull). The Typst writer will pass on specially marked attributes as raw Typst parameters on selected elements. This allows extensive customization using filters. A separate document (doc/typst-property-output.md) has been added that provides extensive documentation and examples of the use of this feature.

Markdown writer:

• Don’t try to align columns in pipe tables with lines greater than COLUMNS. The alignment just reduces readibility when the lines soft wrap.
• Don’t use raw_attribute syntax for raw blocks, unless there is no other option (see #9677). Macros in a raw_attribute block don’t get interpreted when it is read again by pandoc’s markdown reader.

ConTeXt writer:

• Replace depreciated \sc with \setsmallcaps (#9518, James P. Ascher).

Docx writer:

• Use conventional styles/indents for Word bullet lists (#7280).

reference.docx:

• Use current standard Word theme (#7280). This includes using the sans-serif font Aptos instead of the serif font Cambria, and default colors for headings.
• Remove duplicate DefaultParagraphFont in styles.xml.

New module Text.Pandoc.Transforms [API change] (Albert Krewinkel). This module exports the following functions which were formerly exported from Tetx.Pnadoc.Shared: headerShift, filterIpynbOutput, eastAsianLineBreakFilter, as well as some functions that were previously not exported.

Text.Pandoc.Shared:

• headerShift, filterIpynbOutput, and eastAsianLineBreakFilter are no longer exported from this module; they are now exported from Text.Pandoc.Transforms (Albert Krewinkel).

Text.Pandoc.Error:

• Improve reporting of unsupported extensions errors (#9247, Albert Krewinkel).

Text.Pandoc.App:

• Move “transforms” after filters (#9664). This will mean that --shift-heading-level-by affects a heading added by reference-section-title.

Text.Pandoc.App.CommandLineOptions:

• Simplify output for OptVersion. Omit the information about versions of dependencies. We no longer emit version info at this level anyway; pandoc-cli intercepts and handles --version. This code would only be called if someone used the pandoc library function handleWithOptInfo in their own program.

Text.Pandoc.ImageSize:

• Export ImageSize datatype.

Text.Pandoc.SelfContained:

• Merge class attribute when both img and svg specify it (#9652, Carlos Scheidegger).

Text.Pandoc.Logging:

• Add ScriptingInfo constructor for LogMessage [API change] (Albert Krewinkel).
• Make DocxParserWarning a WARNING, not INFO. [API change].
• Add UnsupportedCodePage constructor to LogMessage [API change].
• Add UnclosedDiv constructor for LogMessage [API change].

Lua subsystem (Albert Krewinkel:

• Add a pandoc.log module.
• Uupdate to pandoc-lua-marshal version 0.2.7 (#8916). This fixes counterintuitive behavior of the content property on BulletList and OrderedList items. Unmarshalling of that field now matches the behavior of the constructor.
• Use newest zip module. This adds a symlink function to Entry objects, allowing to check if an entry represents a symbolic link.
• Improve pandoc.json.decode docs.
• Update and fix docs for pandoc.types.Version and pandoc.utils.type.
• Add new module pandoc.image The module provides basic querying functions for image properties.
• Bump pandoc-lua-engine to 0.2.1.4.

Use latest KaTeX CDN asset (#9707, Salim B).

pandoc-cli: ensure UTF8 when emitting version info.

tools/update-lua-module-docs.lua: improve script-internal docs, cleanup (Albert Krewinkel).

Allow network 3.2.

Use latest versions of texmath, djot, skylighting-core, skylighting.

Fix command test for #9652.

Fix some typos in code comments (#9638, guqicun).

Command tests: include regular PATH after directory with the test executable (ensures that DLLs will be found on Windows).

MANUAL.txt:

• Document handout variable for beamer (#9742).
• Document formats affected by --slide-level (#9745).
• Update the list of required LaTeX packages (#9728, Albert Krewinkel).
• Use more descriptive link text for ODT (#9673).
• Add clarification about toc-title in docx, pptx (#9645).
• Better document truthiness for conditionals (#9661).
• Mention that custom-style works with ODT (Ian Max Andolina).
• Harmonize typographic dashes (#9688, Salim B). Standardize on --- with no space.

INSTALL.md: Minor tweaks (#9705, Leo Heitmann Ruiz).

Org reader:

• Fix treatment of id property under heading (#9639).

DocBook reader:

• Add empty title to admonition div if not present (#9569). This allows admonition elements (e.g. <note>) to work with gfm admonitions even if the <title> is not present.

DokuWiki reader:

• Link text cannot contain formatting (e.g., // is not italics) (#9630).
• An explicitly empty link text ([[url|]]) works the same as an omitted link text (#9632).

Typst reader:

• Support Typst 0.11 table features: col/rowspans, table head and foot (#9588).
• Parse cell col/rowspans.

CSLJson writer:

• Put $ or $$ around math in csljson output (#9616).

ConTeXt writer:

• Fix options order with \externalfigure. The dimensions should come after the class if both are present.

Typst writer:

• Put label after Span, not before. Labels get applied to preceding markup item.
• Support Typst 0.11 table features (#9588): colspans, rowspans, cell alignment overrides, relative column widths, header and footer, multiple table bodies with intermediate headers. Row heads are not yet supported.
• The default typst template has been modified so that tables don’t have lines by default. As is standard with pandoc, we only add a line under a header or over a footer. However, a different default stroke pattern can easily be added in a template.
• More reliable escaping in inline [..] contexts (#9586). For example, we need to escape [\1. April] or it will be treated as an ordered list.
• Handle unnumbered on headings (#9585).

LaTeX writer:

• Fix math inside strikeout (#9597).

Text.Pandoc.Writers.Shared:

• Export isOrderedListMarker [API change].

Change lhs tests so they don’t use --standalone. This will avoid test failures due to minor changes in skylighting versions, e.g. #9589.

Use latest texmath, typst.

Require pandoc-lua-marshal 0.2.6 (#9613, Albert Krewinkel). Fixes an issue arising when the value of content properties on BlockQuote, Figure, and Div elements was an empty list.

Update lua-filters.md (#9611, Carlos Scheidegger).

Markdown reader: Fix bug with footnotes at end of fenced div (#9576).

LaTeX reader:

• Improve tokenization of @ (#9555). Make tokenization sensitive to \makeatletter/\makeatother. Previously we just always treated @ as a letter. This led to bad results, e.g. with the sequence \@. E.g., a\@ b would parse as “ab” and a\@b as “a”.
• Make withRaw work inside parseFromToks (#9517). This is needed for raw environments to work inside table cells.
• Better handling of table colwidths (#9579). Previously the parser just failed if the column width specified in p{} wasn’t a multiple of \linewidth. This led to cases where content was skipped.

Typst writer:

• Add ‘kind’ parameter to figures with tables (#9574).
• Avoid unnecessary box around image in figure (#9236).
• Omit width/height in images unless explicitly specified (#9236). Previously we computed width/heigth for images that didn’t have size information, because otherwise typst would expand the image to fit page width. This typst behavior has changed in 0.11. This change fixes a bug in which images would sometimes overflow page margins, depending on their intrinsic size.
• Don’t add hard-coded inset to tables (#9580). Instead, set this globally in the default template, allowing it to be customized.

LaTeX template: Fix block headings support for unnumbered paragraphs (#9542, #6018, Oliver Fabel).

HTML templates: Replace polyfill provider (#9537, @SukkaW). Replace polyfill.io with cdnjs.cloudflare.com/polyfill. polyfill.io has been acquired by Funnull, and the service has become unstable.

Korean translations: delete colon in translation for ‘to’. This was invalid YAML, and not desired anyway, since a colon is added.

Use latest commonmark, commonmark-extensions. This fixes a 3.12 regression in parsing of commonmark/gfm autolinks (jgm/commonmark-hs#151).

Depend on djot 0.1.1.3, which fixes a serious parsing bug affecting regular paragraphs after lists.

Depend on latest skylighting, skylighting-core, typst-hs, texmath.

MANUAL.txt: Change broken link to IDML cookbook (#9563).

Docx reader:

• Ensure that table captions are counted (#9518).
• Detect caption by style name not id (#9518). The styleId can change depending on the localization.
• Avoid emitting empty paragraph where caption was.

Markdown reader: fix regression in link parsing with wikilinks extensions (#9481). This fixes a regression introduced in 3.1.12.

Org reader/writer: support admonitions (#9475).

Org writer: omit extra blank line at end of quote block.

Typst writer: ensure that -, +, etc. are escaped at beginning of block (#9478). Our recent relaxing of escaping (#9386) caused problems for things like emphasized - characters that were rendered using #strong[-]#. This now gets rendered as #strong[\-].

LaTeX writer: fix bug when a language is specified in two different ways (#9472). If you used lang: de-DE but then had a span or div with lang=de, the preamble would try to load ngerman twice, leading to an error. This fix ensures that a language is only loaded once.

Docx writer: Don’t copy over footnotePr in settings.xml from reference.docx (#9522).

EPUB writer: omit EPUB2-specific meta tag on EPUB3 (#9493). This caused a validation failure in epubs with cover images.

Lua: avoid crashing when an error message is not valid UTF-8 (Albert Krewinkel).

Text.Pandoc.SelfContained:

• Add role="img" to svgs.
• Add aria-label to svg elements with alt text if present. Screen readers ignore alt attributes on svg elements but do pay attention to aria-label (#9525).

Text.Pandoc.Shared: Fix regression in section numbering in makeSections (#9516). Starting with pandoc 3.1.12, unnumbered sections incremented the section number.

Text.Pandoc.Class: fix openUrl TLS negotiation (#9483). With the release of TLS 2.0.0, the TLS library started requiring Extended Main Secret for the TLS handshake. This caused problems connecting to zotero’s server and others that do not support TLS 1.3. This commit relaxes this requirement.

Depend on djot 0.1.1.0 (fixes rendering on multiline block attributes).

Use new releases of skylighting-format-blaze-html (#9520). Fixes auto-wrapping of long source lines in HTML print media.

Use new commonmark-extensions (fixes issue with the rebase_relative_paths extension when used with commonmark/gfm.

Makefile: improve epub-validation target (#9493). Use --epub-cover-image to catch issues that only arise with that.

EPUB writer: omit EPUBv3-specific accessibility features on epub2 (#9469). Fixes a regression in 3.1.12.

More fixes for SVG ids with --self-contained (#9467). This generalizes the fix to #9420 so it applies to things like style="fill(url(#..." and should fix problems with SVGs including gradients.

Powerpoint writer: properly handle math in headings and tables (#9465). This ensures that paragraphs containing math are wrapped in a mc:AlternateContent node as required.

Makefile: make validate-epub check v2 output too.

Add djot as input and output format. Djot is a light markup syntax (https://djot.net).

• New module Text.Pandoc.Readers.Djot [API change]. The function readDjot is also exported by Text.Pandoc.Readers.
• New module Text.Pandoc.Writers.Djot [API change]. The function writeDjot is also exported by Text.Pandoc.Writers.

--number-sections now uses the first digit for the number of the top-level section, no matter what its level. So if the top-level section is level-2, numbers will be 1, 2, etc. rather than 0.1, 0.2, as in the past (#5071). For some backwards compatibility, we revert to the old behavior when the --number-offset option is used.

DocBook reader:

• Better handling of <procedure> and <substeps> (#9341): <procedure> now gets parsed as an ordered list, and <substeps> as a sublist.

Man reader:

• Move spaces outside of emph/strong (#9445).

MediaWiki reader:

• Don’t make leading blanks underscores in image links (#9425).
• Allow lowercase image: (#9424).

BibTeX reader:

• Support pagetotal in converting BibLaTeX.

Markdown reader:

• Fix wikilinks extensions to allow newlines in titles (#9454).

EPUB reader:

• Don’t put # characters in identifiers.

LaTeX reader:

• Improve treatment of \cref, \Cref (#7463). Use the reference-type ref+label and ref+Label. Also, associate with \vref ref instead of ref+page.
• Limited support for \Cref (#7463).
• Generate relative widths for \linewidth, \textheight (#9388).

Typst reader:

• Fix handling of \overline (#9294). Due to a typo, it was being incorrectly rendered as an \underset.
• Improve handling of inline #quote (#9413).
• Fix handling of dot(), tilde(), ddot() (jgm/typst-hs#38).
• Fix character used for norm (jgm/typst-hs#38).

Typst writer:

• Use reference form (e.g. @jones2000[p. 30]) for citations when possible.
• Use #ref or @ for links with reference-type="ref" (#7463). This attribute is added to LaTeX \cref, for example.
• Improve citation support (#9452). Emit form: "prose" or form: "year" qualifiers if the citation is author-in-text or suppress-author. Strip initial comma from suffix, since typst will add an extra one.
• Unescape URI escapes in image paths (#9389).
• Handle labels and citaiton ids with spaces and other special characters (#9387). In these cases, we produce an explicit label() rather than using <> or @.
• Avoid producing illegal labels (#9387).
• Avoid unnecessary escapes (#9386).

LaTeX writer:

• Make writer sensitive to empty_paragraphs extension (#9443).
• Fix beamer highlighting (mh4ckt3mh4ckt1c4s).
• Create valid table even when table is empty (#9350).
• Set font fallback for babel main font (Max Heller).
• Add some kerns where needed between quotes (#9371).

HTML writer:

• Add suffix to multiple footnote section ids, so they are unique (Sam May). This is necessary when --reference-location is block or section.

EPUB writer:

• Add ARA roles for accessibility (#9378, Iacobus1983). Footnote references are given role “doc-noteref”, footnote text gets “doc-footnote”, and nav gets “doc-toc”.
• Ensure that an alt attribute is always added (#9354). This seems to be required by iBooks; even an empty alt attribute will satisfy it.
• Add xml:lang to package element (#9372).
• Add accessibility metadata to EPUB metadata (#9372, #9400, Iacobus1983 and John MacFarlane). Reasonable default values are used to ensure that pandoc’s EPUBs conform to the EU Accessibilty Act requirements, but values can be overridden using metadata.

Docx writer:

• Restore ability to center-justify table (#9393). The fix to #5947 caused all tables to be left indented. This was necessary to avoid extra indentation in table cells when a table appeared in a list item. This change makes the changes conditional, so that they only affect tables in list items.

Man writer:

• Fix bug with long URLs (#9458). URLs with more than 68 characters didn’t display properly because of wrapping.
• Support (limited) syntax highlighting in code blocks (#9446). Currently only boldface and italics are supported. The monochrome style might be of use for those generating man pages.

Org writer:

• Escape special lines in code blocks (#9218, Albert Krewinkel).

Markdown writer:

• Use different width fences for nested divs (#9450). Outer divs have longer fences. This aids clarity for the reader, making it easier to see where the div ends. It also makes the output compatible with some other implementations, e.g. micromark, which require different-width fences for nesting.
• Fix output for pipe tables with a huge number of columns (#9346). Previously we got invalid pipe tables when the number of table columns exceeded the setting of --columns.

Powerpoint writer:

• Fix regression in layout for slides with figures (#9442).
• Use internal column widths in pptx writer tables (#5706, Tomas Dahlqvist). The table writer used to only divide all available width evenly for all columns. In this update the code uses the incoming widths if they are available. If they are not set the earlier even distribution is used. Some of the golden templates are adjusted slightly because of different rounding when using the new calculation model.

Custom writers:

• Fix handling of common state (#9229, Albert Krewinkel). The CommonState (PANDOC_STATE in Lua) may change between the time that a custom writer script is first loaded and when the writer is run. However, the writer was always using the initial state, which led to problems, e.g. when the mediabag was updated in a filter, as those updates where not visible to the writer. The state is now updated right before the writer function runs.

Text.Pandoc.SelfContained:

• Fix id replacements in SVGs with clipping paths (#9420). This fixes --embed-resources when SVGs have clip-path attributes.
• Fix size of duplicated SVGs with --embed-resources (#9439).

ConTeXt template: support font fallback (#9361, Lawrence Chonavel).

Text.Pandoc.Shared:

• addPandocAttributes: use wrapper attribute, not wrap, for Divs and Spans added as wrappers to hold attributes on elements that do not accept them.
• makeSections behavior changes:
  • When the optional base level parameter is provided, we no longer ensure that the sequence of heading levels is gapless (#9398). Instead, we set the lowest heading level to the specified base level, and adjust the others accordingly. If an author wants to skip a level, e.g. from level 1 to level 3, they can do that. In general, the heading levels specified in the source document are preserved; makeSections only puts them into a hierarchical structure.
  • Section numbers are now assigned differently, as described above under --number-sections changes (#5071).
• Improve makeSections code for section number calculation.

Text.Pandoc.Chunks:

• Autogenerate unique ids for sections missing them (#9383). This is needed for TOC generation to work properly. We can’t create TOC links if there are no ids. This fixes some EPUB validation issues we’ve been getting since switching over to Chunks for chunking.
• Improve fixTOCTreePaths. We weren’t adding ids for section headings that don’t head a chunk, but these headings are needed for a TOC.

Lua: catch encoding error in pandoc.read (#9385, Albert Krewinkel). Fixed a bug that could lead to an un-catchable error and program termination when pandoc.read was called with invalid UTF-8 input.

LaTeX template: support font fallback (lawcho). This support is LuaLaTeX-specific. See MANUAL.txt for documentation.

Text.Pandoc.Readers: Add readMan to exports [API change] (George Stagg).

Text.Pandoc.PDF:

• Reliably detect when TOC has changed (#9295). Sometimes the TOC changes but there are no warnings: this happens when no labels are present. In this case we must rerun LaTeX. So we now take the SHA1 hash of the TOC file and rerun LaTeX if it changes between runs.
• Increase maximum number of LaTeX runs to 4 (#9299). On some documents, 4 runs are needed (e.g. when a LastPage reference is used).
• Avoid readFileLazy, which caused improperly cleaned-up temp directories on Windows (#9460).

MANUAL.txt:

• Harmonize spelling of Markdown and MultiMarkdown (#9402, Salim B).
• Add <pre> to list of exceptions for markdown_in_html_blocks extension (#9305).
• Add clarification to docs for --resource-path (#9417).

Makefile: Validate generated EPUB as part of prerelease checks.

Add validation for docx golden files to CI (Edwin Török).

Docx reader:

• Fix HYPERLINK with only switch and no argument (#9246).

Org reader:

• Parse caption and label for grid tables (#9279).

MediaWiki reader:

• Handle multiline math in list items (#9293).

OPML writer:

• Respect --wrap options & --columns in contents of notes (#9297).

ODT/OpenDocument writers:

• Properly handle highlighting styles (#9287). These styles were going into an office:styles element in content.xml, but this is invalid. Instead they must go in styles.xml. The variable highlighting-styles no longer has any effect on the default opendocument template, and highlighting styles are not included in opendocument output.

Markdown writer:

• Add table identifier at end of caption if present (#9279).

Text.Pandoc.PDF:

• Expand list of environment variables to display in verbose output (#9303).
• Ensure that we find all the LaTeX warnings requiring a rerun (#9284). This should fix a regression from 3.1.9 that led to incorrect alignments in tables (and possibly other issues).

Docx writer:

• Ensure that pandoc’s output validates (Edwin Török, #9273, #9269, John MacFarlane, #9265, #9266, #9264).
• Don’t emit empty table rows, which seem to cause problems for Word (#9224).

LaTeX writer:

• Omit superfluous page locator label when used with --natbib or --biblatex (#9275). These will treat a bare number as a page locator, and they will be able to localize it. Note that the recognition of the locator label is locale-sensitive; if lang is de, then S. 33 is a page reference, and p. 33 is not!

Text.Pandoc.Chunks: Fine tune makeChunks (#9281).

• Ensure that chunks not based on sections (those with the “preamble” class) get unique identifiers, by appending chunk number.
• This will also ensure that they get unique path names when the path is generated from the identifier.

Default HTML5 template: remove html5shiv (and support for IE < 9).

Makefile:

• Fix make quick-stack: j was expecting a number (Edwin Török).
• Run built pandoc (instead of pandoc in path).
• Add validate-epub target, using epubcheck to test the golden files.
• Add validate-docx-golden-tests target.

Typst writer:

• Emit ; after typst code, unless followed by space (#9252). Otherwise there’s the potential that the typst code will swallow up a following character.

Text.Pandoc.Logging:

• Add MakePDFWarning constructor to LogMessage [API change].
• Add MakePDFInfo constructor to LogMessage [API change].

Text.Pandoc.PDF:

• LaTeX warnings are passed on to the user as warnings.
• Use report with MakePDFWarning and MakePDFInfo to relay verbose information and warnings, instead of writing directly to stderr.
• Parse logs to determine whether additional runs needed, instead of running a fixed number of times (#9255). (The number of times that was appropriate given pandoc’s default templates didn’t always work for custom templates, and thus pandoc 3.1.10’s change in the number of runs led to some regressions in PDF production.)

Makefile: in make prelease, add checks that pandoc-cli and pandoc have the same version, that pandoc-cli depends on this exact version of pandoc, that there is an entry for this version in the changelog, and that the version numbers in the generated man pages are correct.

Regenerate man pages with pandoc 3.1.10. This properly escapes hyphens and fixes version numbers in man pages for pandoc-server and pandoc-lua.

Depend on texmath 0.12.8.6. This omits unneeded lrs in typst math output.

Depend on typst 0.5. This allows the typst reader to support multiline strings, the version type, and the as keyword with import.

Link pandoc-cli version to pandoc version. Henceforth pandoc-cli’s version will be synchronized with pandoc’s, and pandoc-cli will depend on an exact pandoc version. This will avoid confusion by ensuring that cabal install pandoc-cli-X.Y.Z installs pandoc version X.Y.Z. It will make things more straightforward for upstream packagers (see #9232). This scheme does not follow the Haskell PVP, but that should cause no harm, because this package does not expose a library.

Add alerts markdown extension. This enables GitHub style markdown alerts as a commonmark extension. This extension is now default for gfm. It can’t be used with markdown, only with commonmark and variants.

Markdown reader:

• Preserve newlines in math instead of changing to spaces. Otherwise we can get unwanted results if there’s a % comment (#9193).
• Make attributes work with reference links (#9171).

HTML reader:

• Improve handling of invalidly nested sublists (#9187, cf. #8150).

MediaWiki reader:

• Allow attribute keys with hyphens (#9178).

ODT reader:

• Support attr text:continue-numbering (#8979, Stephan Meijer).

Typst reader:

• Allow references (e.g. @foo) to become citations if there is no corresponding label in the document.
• Collapse adjacent cite elements.
• Handle supplements in cite.
• Change cite (only one key allowed, a label) (typst 0.9 breaking change).
• Support quote element (typst 0.9).

LaTeX reader:

• Handle otherlanguage environment and language-name environments like \begin{french}...\end{french} (#9202).
• Fix theorem label parsing (#8872, Hikaru Ibayashi).

Docx reader:

• Unwrap content of shaped textboxes (Stephan Meijer, #9214).
• Improve handling of w:sym (#9220). We now look up symbols in symbol fonts using the table defined at Text.Pandoc.Readers.Docx.Symbols.
• Add unexported module Text.Pandoc.Readers.Docx.Symbols. This gives us a table to use to resolve characters included in docx via w:sym element.

Man reader:

• Properly handle .sp macro inside lists and block quotes (#9201).

LaTeX writer:

• Fix bug with big footnotes inside emphasis (#8982, Hikaru Ibayashi).
• Handle identifiers inside heading contents. \phantomsection can’t be used in this case, so we need \hypertarget (#9209).

LaTeX template:

• Include bookmark package unconditionally. This package produces better PDF bookmarks than hyperref and does it on the first pass.

Typst writer:

• Use quote for block quotes.
• Support --toc-depth as in other writers (#9242).
• Put inline image dimensions on enclosing box, not image (#9104).
• Better handling of tables with captions (#9194). We now put these in a figure with a caption argument.
• Update typst writer to typst 0.9 citation format (#9188).

Typst template:

• Remove custom definition of blockquote in default template. (We now use built-in quote.)
• Support table of contents.
• Support csl (#9186, Ian Max Andolina). Typst now supports CSL for its native citation engine, so pandoc should use a specified csl style in the template, falling back to bibliographystyle if csl is not specified.

Docx writer:

• Use different style for block quotes in notes (#9243). Using “Footnote Block Text” for the style name, so it can be given a different font size if footnotes are.
• Allow embedded fonts to be used in reference.docx (#6728).

HTML5 writer:

• To conform to validator’s expectations, doc-footnote role is used with aside and doc-endnotes with section.
• aside is used only for notes at ends of sections or blocks; if all the notes come at the end of the document, section is used so we can have the doc-endnotes role.

JATS writer:

• Handle case where there is material after refs div (#9166). Previously in such cases the references were not being moved to back matter.

Ms writer:

• Don’t do normal escapes in filename arguments for PSPIC etc.

T.P.RoffChar: escape - as \-. The groff_man (7) man page indicates that - characters will be treated as typographic hyphens and are not appropriate for cases where the output should be copy-pasteable as an ASCII hyphen-minus character. (E.g. in command line options.) However, until a recent update groff man did not actually do this; it treated - and \- the same. With the new update (1.23.0) the two are distinguished (see https://lwn.net/Articles/947941/ for background), so now it is important that pandoc escape -.

Text.Pandoc.Extension: add Ext_alerts constructor [API change].

Text.Pandoc.PDF: We now default to running LaTeX only once in producing a PDF (instead of twice). This is made possible by the shift to the bookmark package, which does not require a second pass for PDF bookmarks. If a table of contents is present, we still have to run three times to get the page numbers, and if beamer is used we still do a minimum of two runs.

Text.Pandoc.Shared:

• renderTags': use minimized tag for rect.
• Allow svg path element to be minimized.
• Export combineAttr [API change].
• Improve isTightList so that it recognizes an item containing only a list which is itself tight as potentially an item in a tight list (#9161).

Text.Pandoc.MIME: Ensure we use .svg not .svgz as extension for image/svg+xml mime type. This fixes issues with embedded SVG images in docx output, among other things (#9195).

Text.Pandoc.Class: openURL improvements for data uris. Only treat data URI as base64 if ‘;base64’ is specified. Otherwise treat as UTF-8 (not 100% reliable but should cover most other cases). Strip off ;base64 (or ;charset=... or whatever) from mime type (#9195).

Text.Pandoc.SelfContained: Improve treatment of embedded SVGs (#9206, #8948).

• Ensure unique ids for elements by prefixing SVG id.
• Ensure SVG id attribute except when use element is used.
• Remove width, height attributes from svg element when use element is used. Instead, add width and height 100% to the use element. This seems to get the sizing right.

Text.Pandoc.Citeproc: Don’t link citations if suppress-bibliography specified, for there will be nothing to link to (#9163).

epub.css: add styling for sup and sub (#9160).

Switch from base64 to base64-bytestring (#9233).

Use newest versions of commonmark, commonmark-extensions, commonmark-pandoc, texmath, typst, skylighting, skylighting-core.

Benchmark: use standalone documents for reader tests. Otherwise typst reader benchmark fails. Note: this means that we are now parsing longer documents, so bench results on readers won’t be comparable to before.

MANUAL.txt: update defaults file docs for bibliography fields (#9173). Recommend using top-level bibliography csl, etc. instead of a nested metadata field. Reason: ${USERDATA} and ${HOME} are only expanded in these contexts, not in metadata.

Move man pages to pandoc-cli package (#9245).

Make reference-section-title work with jats+element_citations (#9021).

Add bits as synonym of jats as input format.

JATS reader:

• Modify JATS reader to handle BITS too (#9138, Julia Diaz). Add provision for title-group, book, book-part-wrapper, book-meta, book-part-meta, book-title, book-title-group, index, toc, legend, title, collection-meta
• Fix handling of alt-text (#9130, Julia Diaz). Previously we were looking for an attribute that doesn’t exist in JATS; alt-text is provided by a child element.

CommonMark reader:

• Handle Ext_tex_math_gfm (#9121). Parse GFM-specific math constructions when tex_math_gfm enabled.

DokuWiki reader:

• Allow autolinks to be avoided using e.g. https:%%//%%... (#9153).
• Parse <code> and <file> as block-level code (#9154). Previously we treated them as inline code in some contexts, but that is not how DokuWiki works.

LaTeX reader:

• Better handle spacing commands \hfill, \vfill, \hskip, \vskip, etc. (#9150).
• Fix incorrect abbreviation for astronomical unit (#9125, Michael McClurg).

Markdown reader:

• Fix blindspot with superscript in links (#8981). Previously [^super^](#ref) wasn’t parsed as a link, due to code that was meant to prevent footnote markers from being recognized as reference links. This commit tightens up that code to avoid this bad effect. We have also added a new restriction on footnote labels: they cannot contain the characters ^, [, or ]. Though this is technically a breaking change, we suspect that the impact will be minimal, as it’s very unlikely people would be using these characters in their note labels.
• Don’t apply --default-image-extension to data URIs (#9118).
• More accurate check that a normalCite is not a link, bracketed span, or reference (#9080).

HTML reader:

• Allow th to close td and vice versa (#9090).
• Parse task lists using input elements (#9047, Seth Speaks).

Creole reader:

• Handle empty cells correctly (#9141, Sascha Wilde).

Org writer:

• Escape literal *, |, # at beginning of line with ZWS (#9159).

ICML writer:

• Prevent doubled attributes (#9158).

Powerpoint writer:

• Fix a corruption error caused when the document used both a regular png and a png in a data URI (#9113). (Similarly for any other image format.) The problem was that duplicate entries in [Content Types].xml were being created, one for the mime type image/png, one for image/png;base64.

LaTeX writer:

• Fix rowspans in tables so they use the width of the column (= as the width parameter) (#9140).
• Don’t treat table as “simple” if they have col widths. This should help fix a problem wherein some grid tables with colspans were overly wide (#9140).
• Fix uneven indents in line block output (#9088).

JATS writer: fix 3.1.4 regression in handling block-level metadata (#9092).

Ms writer: improvements in image handling (#4475).

• PDFPIC is now used for PDF images in figures.
• Inline images that are postscript or PDF are rendered using PSPIC or PDFPIC. This isn’t ideal, because they will still be rendered as if in a separate paragraph, but it’s probably better than just printing the image name.
• Units are included in height.

HTML writer:

• If raw format is an HTML side deck format, emit it (James J Balamuta).

Typst writer:

• Add #box around image to make it inline. (#9104) An #image by itself in typst is a block-level element. To force images to be inline (as they are in pandoc), we need to add a box with an explicit width. When a width is not given in image attributes, we compute one from the image itself, when possible.
• Don’t allow long heading to wrap (#9132).
• Escape ( (#9137). If unescaped ( occurs in certain contexts, it can be parsed as function application.

Man writer:

• Fix some spacing issues around links (#9120). We need to use \c before a .UR or .MT, to avoid an extra space, and also after. To ensure that a space at the beginning of the following line doesn’t get swallowed up, we escape it with \.
• Use UR, MT macros for URLs, emails (#9120).

Text.Pandoc.Extensions:

• Add Ext_tex_math_gfm constructor to Extension (#9121). [API change]. This handles two GitHub-specific syntaxes for math. This is now default for gfm, in addition to tex_math_dollars.
• Remove duplicates for Ext_raw_html and Ext_pipe_tables in some of the lists (Tim Stewart).

Text.Pandoc.Metadata: Add helpful message on some metadata YAML errors (#9155).

Text.Pandoc.Shared:

• splitSentences: don’t split after initials. This improves the man and ms writer output, preventing sentence breaks after initials.
• Add addPandocAttributes function [API change]. This is meant to simplify addition of attributes to Pandoc elements: for elements that don’t have a slot for attributes, an enclosing Div or Span is added to hold the attributes.

MANUAL.txt:

• Clarify that formatting can’t cross line boundaries in line blocks (#9119).
• Fix legacy option for citation (#8737, 3w36zj6)

Update et translations (priiduonu).

Updated no translations (Stephan Daus). Renamed no.yaml (macrolanguage Norwegian) to nb.yaml (Norwegian Bokmål). Created soft symbolic link from no.yaml pointing to nb.yaml.

Lua subsystem: Use the newest LPeg version (lpeg-1.1.*) (#9107, Albert Krewinkel).

Default epub.css: Apply style to h6, format styles, and combine identical styles under shared selectors (samuel-weinhardt).

Update nix flake with dependencies (piq9117).

LaTeX template: fix \CSLBlock vertical space (John Purnell).

Allow tasty 1.5 and Diff 0.5.

Require commonmark-extensions 0.2.4, commonmark 0.2.4.

Require texmath 0.12.8.4. This should improve math in powerpoint, fixing empty boxes around roots in some cases.

Require typst 0.3.2.1

JATS reader:

• Ignore <processing-meta> element (#9057, Julia Diaz).
• Fix conversion of date to ISO 8601 format (#8865).

LaTeX template:

• Add code allow \cite to break across lines (#9050).
• Fix regression with CSL display="block" (#7363). This restores the line break before the block.
• Rewrite CSLReferences environment to avoid depending on enumitem, which plays badly with beamer. Instead we use a regular list environment. Thanks to @jpcirrus for the concept (#9053).
• Restore the pre-3.1.7 format of the CSLReferences environment, which again has two parameters. The first determines whether a hanging indent is used (1 = yes, 0 = no), and the second is the entry line spacing (0 = none).
• Add a strut to avoid inconsistencies in spacing (#9058).
• Remove a break at the end of CSLRightInline to avoid inconsistencies in spacing. It shouldn’t be necessary because the paragraph should extend to the right margin (#9058).

LaTeX writer:

• Fix regression with figure labels (#9045). In 3.1.7, pandoc added two labels to LaTeX figure environments, one with a phantomsection.
• Fix default citeproc entry-spacing. According to the CSL manual, the default entry spacing is 1. We were treating it as 0 (#9058).

HTML writer:

• Use the ID prefix in the ID for the footnotes section (#9044, Benjamin Esham).
• Fix CSL entry-spacing default (#9058).

Text.Pandoc.Citeproc: always include an entry-spacing attribute in the Div if the bibliography element contains an entry-spacing attribute (previously we omitted it when it was 0) (#9058).

Clean up pandoc’s own man pages by regenerating with pandoc 3.1.7.

pandoc-lua-engine: bump lower bound for pandoc (#9046).

Depend on texmath 0.12.8.2, fixing binom in typst writer (#9063).

Org reader:

• Don’t parse alphabetical lists unless the fancy_lists extension is enabled (#9042).
• Allow escaping commas in macro arguments (Amneesh Singh).

JATS reader:

• Support for <permissions> metadata (#9037, Julia Diaz). metadata objects with multiple fields are created, matching the structure in JATS.
• Correct name of JATS element attrib.

Markdown reader:

• Support images with wikilink syntax, e.g. ![[foo|bar]], when one of the wikilinks extension is enabled (#8853).
• Allow a citation or reference link to be parsed after a ! (#8254).
• Fix dropped ! before nonexistent reference (#9038).

LaTeX writer:

• Fix regression in escaping URLs (#9043).
• Use \cite and \bibitem to link up citations, even with citeproc. (#9031). This will give us better accessibility; when tagging is enabled, the citation can be linked to the bibliography entry. This changes some of the details of the layout and the default template. We now make CSLReferences a special enumitem list that will contain \bibitems. Internal links inside citations to ids beginning in ref- are creating using \cite instead of \hyperref.
• Use \phantomsection and \label instead of \hypertarget (#9022).
• Use \hyperref for LaTeX internal links, \hyperlink for beamer (since \hyperref doesn’t seem to work) (#9022).
• Backslash-escape % and # in URLs (#9014).

JATS writer:

• Fix placement of ref-list when no title is specified for the reference section (#9017). (In this case we place it in back with an empty title.)

Man writer:

• Avoid a .PP right after a section heading (#9020). This is at best a no-op (in groff man and mandoc) and at worst (in some formatters) may create extra whitespace.
• We revert the fanciness introduced in #7506, which employs a custom font name V and a macro that makes this act like boldface in a terminal and monospace in other formats. Unfortunately, this code uses a mechanism that is not portable (and does not work in mandoc) (#9020).
• Instead of using V for inline code, we simply use CR. Note that \f[CR] is emitted instead of plain \f[C], because there is no C font in man. (This produces warnings in recent versions of groff, #9020.)
• For code blocks, we now use the .EX and .EE macros, together with .IP for spacing and indentation. This gives more standard code that can be better interpreted e.g. by mandoc (#9020).

Man template: don’t emit .hy, regardless of setting of hyphenate variable (#9020).

LaTeX template: special redefinition of \st for CJK (#9019). soul’s version raises on error on CJK text.

Use latest skylighting-format-blaze-html (#7248). This works around a longstanding iOS Safari bug that caused long lines to be displayed in a different font size in highlighted code.

Allow skylighting 0.14 (and require it in pandoc core).

Allow text 2.1.

Org reader: allow example lines to end immediately after the colon (Brian Leung).

Docx reader:

• Omit “Table NN” from caption (#9002).
• Avoid spurious block quotes in list items (#8836).

JATS reader: Fix display of block elements (#8889, Julia Diaz). A number of block elements, like disp-quote, list, and disp-formula, were always treated as inlines if appearing inside paragraphs, even if their usage granted a separate block.

HTML reader: avoid duplicate id on header and div (#8991).

Typst writer:

• Use ~ for nonbreaking space, and escape literal ~ (#9010).
• Put the label in right place for Div, use #block (#8991). Previously we were putting the label at the beginning of the Div’s contents, but according to the documentation such a label gets attached to the preceding element. We now use an explicit #block and add the label at the end.

LaTeX writer:

• Improve escaping of URIs in href, url (#8992).
• Improve internal links and targets (#8744). We no longer wrap section headings in a \hypertarget. This is unnecessary (hyperref creates an anchor based on the label) and it interferes with tagging. In addition, we now use \hyperref rather than \hyperlink for internal links. Currently \hypertarget is still being used for link anchors not on headings. Thanks to @u-fischer.

HTML format templates (style.html): Fix typo in clause for svg (Jackson Schuster).

Use lastest texmath, typst-symbols, typst. Targets typst 0.7.

HTML reader: properly calculate RowHeadColumns (#8984). This fixes a bug in the calculation of the number of header columns in table row. It also changes the algorithm for determining the table body’s RowHeadColumns based on the numbers of head columns in each row. Previously we used the max, and #8634 switched to the min, which led to bad results. Now we only set RowHeadColumns to a non-zero value if all rows have the same number of head columns.

OpenDocument writer:

• Implement syntax highlighting for inline and block code (#6710).
• Support highlighted text in ODT/OpenDocument writers for Span with class mark (#8960). The color can be adjusted by modifying the Highlighted style.

Typst writer: escape // so it doesn’t get interpreted as a comment (#8966).

ChunkedHTML writer: Fix regression including MathJax script (#8967). The fix for #8620 caused the script to be included when the table of contents but not the body text of a page contains math. But it broke the case where the table of contents doesn’t contain math but the page does. This patch fixes the issue.

Text.Pandoc.SelfContained:

• Retain attributes in SVG tag when referring to another SVG’s content using <use> (#8969).
• Allow units in width and height for SVG. Units are optional but allowed.
• Don’t coerce calculated SVG dimensions to Int.
• fix calculation of SVG width and height. We were computing width and height from viewBox incorrectly (#8969).
• Add clause for SVG to default CSS for HTML (#8969).
• Ensure that width and height attributes don’t get specified twice is both the img tag and the svg include them (#8965).
• Omit unnecessary attributes xmlns, xmlns:xlink, and version on SVG element (#8965).
• Use 20 character rather than 40 character hashes for generated IDs (#8965).

Use pandoc-types 1.23.1. This fixes a regression with toJSONFilter (#8976), which in 1.23.0.1 no longer worked on pure values of type a -> [a].

Use ghc 9.6 for release builds (#8947).

Fix some links in FAQs (Diogo Almiro).

Fix CVE-2023-38745, a variant of the vulnerability in CVE-2023-35936. Guilhem Moulin noticed that the fix to CVE-2023-35936 was incomplete. An attacker could get around it by double-encoding the malicious extension to create or override arbitrary files.

--embed-resources: Use inline SVG instead of data uris for SVG images in HTML5 (#8948). Note that SelfContained does not have access to the writer name, so we check for HTML5 by determining whether the document starts with <DOCTYPE! html>. This means that inline SVG won’t be used when generating document fragments.

Fix regression on short boolean arguments (#8956). In 3.1.5 boolean arguments were allowed an optional argument (true|false). This created a regression for uses of fused short arguments, e.g. -somyfile.html, which was equivalent to -s -omyfile.html, but now raised an error because pandoc attempted to parse o as a boolean true or false. This change allows the fused short arguments to be used again. Note that -strue will be interpreted as -s with an argument true, not as -s -t -rue. It is best to use long option names with the optional boolean values, to avoid confusion.

Make --epub-title-page’s argument optional. It takes a boolean argument, and now that all of our boolean flags take such an argument, we can make this one optional for consistency.

Improve errors for illegal output formats. Previously if you did pandoc -s -t bbb, it would give you an error about the missing bbb template instead of saying that bbb is not a supported output format.

Improve errors for incorrect command-line option values (#8879). Always give the name of the relevant argument.

Fix typo on error message for incorrect --preserve-tabs argument. Thanks @fsoedjede

Docx reader: use SVG version of image if present (#7244). Previously the backup PNG was exported even if an SVG was present, but the SVG should be preferred.

Typst reader: fix regression in recognition of display math (#8949). The last release caused all math to be parsed as inline math.

JATS writer: don’t use <code> for inline code (#8889). It is intended for block-level code.

HTML writer: don’t make line blocks sensitive to --wrap (#8952).

RST writer: fix figure handling (#8930, #8871). This fixes a number of regressions from pandoc 2.x. Properly handle caption, alt attribute in figures. No longer treat a paragraph with a single image in it as a figure (we have a dedicated Figure element now).

Docx writer: Copy “mirror margins” property from reference.docx (#8946).

Text.Pandoc.UTF8: Deprecate decodeArg which is now a no-op. This was needed for old base versions which we no longer support.

Use released skylighting, typst.

Allow latest commonmark-extensions. This allows entities in wikilinks.

Switch back to using ghc 9.2 for linux and Windows binary releases (#8947, #8955). With ghc 9.4+, we were getting AVX instructions in the amd64 binary, which aren’t supported on older hardware. For maximum compatibility we switch back to ghc 9.2, which doesn’t cause the problem. (As documented, ghc should not be emitting these instructions, so we aren’t clear on the diagnosis, but the cure has been tested.)

Change Windows release build to use cabal instead of stack.

Allow all boolean flags to take an optional true or false value (#8788, Sam S. Almahri). The default is true if no value is specified, so this is fully backwards-compatible.

Support --id-prefix for markdown output (#8878)

Markdown reader:

• Add strictness annotations to fix a memory leak (#8762).

Typst reader:

• Use typst-hs 0.3.0.0, which is more robust, fixes many bugs, and targets typst 0.6.
• Package loading is now supported, as long as the package has been cached or is local.
• Rewrite Typst reader in a way that makes it easier to extend.
• Filter out CR in raw.
• Handle block content for link element.
• Handle block-level content in text element.
• Handle style, align, place in inline contexts too.
• Improve info message for skipped elements.

Add typst reader tests (#8942).

MediaWiki reader:

• Revise treatment of “link trail.” Previously we only included ASCII letters. That is correct for English but not for, e.g., Spanish (see comment in #8525). A safer approach is to include all letters except those in the CJK unified ideograph ranges.

AsciiDoc writer:

• Make modern AsciiDoc the target for asciidoc (#8936). The AsciiDoc community now regards the dialect parsed by asciidoctor as the official AsciiDoc syntax, so it should be the target of our asciidoc format. The asciidoc output format now behaves like asciidoctor used to. asciidoctor is a deprecated synonym. For the old asciidoc behavior (targeting the Python script), use asciidoc_legacy. The templates have been consolidated. Instead of separate default.asciidoctor and default.asciidoc templates, there is just default.asciidoc.
• Text.Pandoc.Writers.AsciiDoc API changes:
  • writeAsciiDoc now behaves like writeAsciiDoctor used to.
  • writeAsciiDoctor is now a deprecated synonym for writeAsciiDoc.
  • New exported function writeAsciiDocLegacy behaves like writeAsciDoc used to.
• Update line-through for asciidoc writer to custom inline style (#8933, Kevin Broch).

Typst writer:

• Support unlisted class in headings (#8941).
• Consolidate bibliography files into one #bibliography command (#8937).
• Improve handling of autolinks (#8931).

Docx writer:

• Make relative widths work in tables. This didn’t work before because we were missing an attribute that tells Word to used fixed widths rather than computing optimal ones.

DokuWiki writer: fix lists with Div elements (#8920). The DokuWiki writer doesn’t render Divs specially, so their presence in a list (e.g. because of custom-styles) need not prevent a regular DokuWiki list from being used. (Falling back to raw HTML in this case is pointless because no new information is given.)

LaTeX writer:

• Fix babel name for fa (should be persian).
• Prevent babel language from being imported twice (#8925).

Text.Pandoc.Class:

• Add toTextM [API change]. This is like Text.Pandoc.UTF8.toText, except:

  • it takes a file path as first argument, in addition to bytestring contents
  • it raises an informative error with source position if the contents are not UTF8-encoded

This replaces utf8ToText whenever we have the filename and are in a PandocMonad instance. This will lead to more informative error messages for UTF8-encoding, indicating the file path and byte offset where the error occurs (#8884).

Remove invalid term “Subject” from Turkish translations (#8921).

stack.yaml: add pkg-config to nix packages (#8927, pacien).

Allow aeson 2.2.

MANUAL: Add clarification on –section-divs. Closes #8882.

Fix a security vulnerability in MediaBag and T.P.Class.IO.writeMedia. This vulnerability, discovered by Entroy C, allows users to write arbitrary files to any location by feeding pandoc a specially crafted URL in an image element. The vulnerability is serious for anyone using pandoc to process untrusted input. The vulnerability does not affect pandoc when run with the --sandbox flag. [CVE-2023-35936]

Allow epub-title-page to be used in defaults files (#8908).

Issue Extracting info message (in --verbose mode) when using --extract-media or extracting media temporarily in PDF production.

HTML reader: Update TableBody RowHeadColumns caculation (#8634, Ruqi). This change sets RowHeadColumns to the minimum value of each row, which gives better results in cases where rows have different numbers of leading th tags.

Dokuwiki reader: retain image query parameters as attributes (#8887, echo0).

Textile reader: Add support for link references (#8706, Stephen Altamirano). Textile supports what it calls “link alias”, which are analogous to Markdown’s reference-style links.

LaTeX reader: support alt text on images (#8743, Albert Krewinkel).

Commonmark reader: Make implicit_figures work again. Support for this (introduced in #6350) disappeared when we made an architectural change.

JATS reader:

• Add footer and multiple body parsing to table reader (#8765, Noah Malmed).
• Parse references title from ref-list (#8365).

JATS writer:

• Make --number-sections work.
• Include title in ref-list (#8364). Previously the reference title ended up in a separate section at the back of the body instead of in the ref-list in the back matter.

Mediawiki writer: allow highlighting to work for F# language (Adelar da Silva Queiróz).

LaTeX writer: Fix escaping of & in \href and \url (#8903).

Docx writer:

• Fix localization of “Abstract” title (#8702).
• Allow abstract-title to be specified in docx metadata (#8794).

ChunkedHTML writer: Make math work in top-level page (#8915).

Text.Pandoc.Logging: add new log message type ScriptingWarning [API change] (Albert Krewinkel).

Lua: report warnings from Lua scripts (Albert Krewinkel). Lua’s warning system is plugged into pandoc’s reporting architecture. Warnings that are raised with the Lua warn function are now reported together with other messages.

Use crypton-connection instead of connection (#8896, Felix Yan). Follows the change introduced in tls 1.7.0.

Bump versions for skylighting-core, skylighting.

Include lua/module/sample.svg in cabal extra-source-files (Felix Yan).

Add Nynorsk (New Norwegian) translations (Per Christian Gaustad).

Add tests for fillMediaBag/extractMedia.

INSTALL.md:

• Mention alternatives to LaTeX to generate PDF (Norwid Behrnd).
• Update Linux install links (harabat).

pandoc-extras.md: add to “Academic publishing workflows” (#8696, Vladimir Alexiev).

New output format: typst.

New module: Text.Pandoc.Readers.Typst [API change].

DocBook reader:

• Support more emphasis roles (Albert Krewinkel). The role “bf” is taken to indicate “bold face”, i.e., “strongly emphasized” text, while “underline” leads to underlined text.

JATS reader:

• Improve title and label parsing in the JATS reader (#8718, Noah Malmed.)
• Add rowspan, colspan and alignment to cells in jats table reader (#8408, Noah Malmed)

Org reader (Albert Krewinkel):

• Require abstract environment to use lowercase.
• Treat #+NAME as synonym for #+LABEL (#8578).

ODT reader:

• Allow lists in table cells (#8892).
• Allow frames inside spans (#8886).

RST reader:

• Fix sorting on anonymous keys (#8877). This fixes a link resolution bug bug affecting RST documents with anonymous links.

HTML reader:

• Fix iframe with data URI of an image (#8856). In this case we don’t want to try to parse the data at the URL. Instead, create an image inside a div.

RTF reader:

• Fix bug in table parsing (#8767). In certain cases, text before a table was being incorporated into the table itself.

Docx reader:

• Introduce support for Intense Quote (Stephan Meijer).

Markdown reader:

• Disallow escaping of ~ and " in markdown_strict (#8777, Albert Krewinkel). This matches the behavior of the legacy Markdown.pl as well as what is described in the manual.

LaTeX reader: ignore args to column type in \multicolumn (#8789).

HTML writer:

• Use first paragraph in task item as checkbox label (#8729, Albert Krewinkel).

Ms writer:

• Coerce titles to inlines (#8835). Block-level formatting is not allowed inside .TL.

LaTeX writer:

• Fix width for multicolumn simple table (#8831).

Jira writer:

• Use first code block class as highlighting language (#8814, Albert Krewinkel). The writer no longer searches the list of classes for a known programming language but always uses the first class in that list as the language identifier.

OpenDocument writer:

• Handle row header column cells as header cells (#8764, Michael Stahl).
• Fix invalid text:p inside text:p from meta (#8256).

ODT writer:

• Don’t add settings.xml (Michael Stahl). This will cause defaults to be used, which is what we want.
• Don’t add unnecessary Configurations2 directory (Michael Stahl).
• Don’t add thumbnail (Michael Stahl).
• Put manifest.version on directory file-entry (Michael Stahl). See ODF 1.3 part 2, 4.16.14.1.
• Stop validator complaints by producing ODF 1.3 (Michael Stahl).

MediaWiki writer:

• Remove links from inside links in mediawiki writer (#8739, Wout Gevaert).

Typst writer:

• Omit bibliography if citations not enabled (#8763). With this change, the typst writer will omit the #bibliography command when citations is not enabled. (If you want to use pandoc’s own --citeproc, you should combine it with -t typst-citations to disable native typst citations.
• Use <..> for labels, create internal links.
• Use #footnote for notes (#8893).
• Fix alignment issue in lists. It’s an aesthetic issue only; the first line had an extra space indent after the list marker.

Commonmark writer:

• Use shortcut reference links: commonmark supports these.

EPUB template: add lang attribute to <html> (Gabriel Lewertoski).

Template styles.html: fix task-list styling in reveal.js (#8731, Albert Krewinkel).

LaTeX template: Fix \babelfont (#8728).

Text.Pandoc.Parsing:

• Remove unnecessary ‘spaces’ in parseFromString.

Text.Pandoc.ImageSize: Drop BOM at start of SVG if present. Otherwise our code can fail to determine image size.

Lua subsystem:

• Fix value of PANDOC_SCRIPT_FILE for custom readers & writers (#8781, Albert Krewinkel). The value did not hold the actual file path for scripts in the custom folder of the datadir.

Fix YAML in translation files for cs and pl (#8787).

Fix pdf output via typst (#8754). One must now use typst compile rather than typst.

MANUAL.txt:

• Added note that the user will need to create the user data dir (#8727).
• Add wikilinks to non-default extensions (Ilona).
• Update link to custom djot writer (Albert Krewinkel).
• Better link to citation syntax.
• Fix typo (sdhoward).
• Note that # fancy list markers don’t work with commonmark (#8772, William Lupton).
• Add commonmark fenced_div note (#8773, William Lupton).
• Move highlighting documentation, with minor adjustments (William Lupton).
• Fix inaccurate statement about spaces and tabs in template syntax (Frank Seifferth).

Update documentation for org-mode (Christian Christiansen, #8716).

doc/lua-filter.md:

• Fix typos (#8734, perro tuerto).
• Fix anchor (Toni Dietze).
• Use full field name in example (#8857, Matt Dodson).
• Fix copy-paste error (#8798, thron7).

CONTRIBUTING.md: update info on ghc versions.

INSTALL.md:

• Fix cabal install instructions (Albert Krewinkel).
• Use more relevant link to NetBSD/pkgsrc entry (Charlotte Koch).
• Fix Windows install instructions for winget (#8799).

Tests: Rename test/docx/block_quotes_parse_indent.native for consistency (Stephan Meijer).

Add tls constraint on cabal.project. This is needed to avoid problems caused by the transition to crypton.

Require texmath 0.12.8.

Add a Lua REPL (Albert Krewinkel). This can be started with pandoc lua -i. It is also possible to instruct a filter to open the REPL at a certain point, for debugging (see pandoc.cli.repl).

Support typst as a --pdf-engine.

Add typst writer (#8713). New module Text.Pandoc.Writers.Typst, exporting writeTypst [API change].

Org reader:

• Allow zero width space as an escape character (#8716, Christian Christiansen). Allow the character U+200B to be used as an escape character as described in the Org-mode documentation (https://orgmode.org/manual/Escape-Character.html).

DocBook reader:

• Handle “book” for xref references (#8712, Andres Freund) This also adds a test xref to book and part.
• Handle <part> (#8712).

HTML reader:

• Fix behavior with -native_spans-raw_html (#8711). Previously with this configuration, <span>s were not treated as inline elements at all.

HTML writer:

• Avoid duplicate classes (#8705).
• Use img element instead of embed for .svg.gz and .png.gz etc. (#8699).
• HTML writer footnotes changes (#8695): when --reference-location=section or =block, use an aside element for the notes rather than a section. When --reference-location=section, include the aside element inside the section element, rather than outside. (In slide shows, this option causes footnotes on a slide to be displayed at the bottom of the slide.)

EPUB writer:

• Use different structure for epub footnotes (#8676, see #8672, #5583). Many EPUB readers are thrown off by pandoc’s current footnote output. Both the ol and the fact that the footnote backlink is at the end of the note seem to pose problems. With this commit, we now create a list of aside (or div) elements, instead of an ordered list. Each element begins with a note number that is linked back to the note reference. (So, the backlink occurs at the beginning rather than the end.) Thanks to @Porges and @lewer.

Docx writer:

• Include abstract title (#8702). Uses localized term for abstract.

Markdown writer:

• Use implicit figures if there’s a caption but no alt (#8689, Albert Krewinkel).

Jira reader (Albert Krewinkel):

• Add panel title as nested div (#8681).
• Require jira-wiki-markup 1.5.1 (#8680). This fixes a bug in the parser that caused text between two exclamation marks to be parsed as an image. The first ! of image markup must now be followed by a non-space character; otherwise, the enclosed text is parsed as normal content.

Ms writer:

• Fix handling of Figure (#8660).

ICML writer:

• Fix images with data (#8675). The Contents element should be inside Properties.

LaTeX writer:

• Add Chinese to Babel languages.
• Fix background image in Beamer when there are figure environments (#8671, Martín Pozo).

LaTeX template:

• Add babelfonts variable to default LaTeX template. This allows specifying certain fonts to be used with certain babel languages. Thanks to Frederik Elwert.
• Fix highlight/underline with lualatex (#8707). We need the lua-ul package instead of soul, which doesn’t work with lualatex.

Lua (Albert Krewinkel):

• Add pandoc.cli.repl function
• Fix json.encode for nested AST elements. Ensures that objects with nested AST elements can be encoded as JSON.
• Auto-generate docs for pandoc modules.
• Load text module as pandoc.text. This only affects the name in the Lua-internal documentation. It is still possible to load the modules via require 'text', although this is deprecated.
• Move docs from module text to pandoc.text The latter is easier to use and more consistent with the other modules.
• Keep the Lua stack clean A metatable used during initialization was not properly removed from the stack. Likewise, accessing the CommonState from Lua previously led to the pollution of the Lua stack with a left-over value.
• Add function pandoc.format.from_path.
• Allow to get the JSON encoding of log messages.

Text.Pandoc.Format: Add new function formatFromFilePaths [API change] (#8710, Albert Krewinkel).

The old Text.Pandoc.App.FormatHeuristics module has been removed.

In --version, use Windows %APPDATA% variable to describe user data dir (#8686, Pablo Rodríguez).

Text.Pandoc.App.CommandLineOptions: don’t lowercase arg to --from/--read (Albert Krewinkel). This prevented users to use custom writers with uppercase characters in their filenames. Format-normalization, including lower-casing of format identifiers, happens during format parsing.

Documentation:

• Add doc/nix.md.
• Add doc/extras.md. This was formally in the website repo.
• doc/lua-filters.md: improve docs for pandoc.zip.

Factor out make_macos_release.sh from the release candidate workflow. Use cabal instead of stack to build the macos binary.

Modify linux/make_artifacts.sh so it will work on cirrus.

Switch to hslua-2.3

Depend on latest releases of texmath, doclayout.

EPUB reader: Give additional information in error if the epub zip container can’t be unpacked.

TSV reader: don’t gobble tabs as whitespace (#8661).

Org reader: accept empty tables (#8659).

LaTeX reader: fix multiplication syntax for tabular (#8658). We recognized *{6}{...} but not *6{...} or *6c.

Docx reader: parse image alt texts in LibreOffice generated files. LibreOffice tags images slightly differently than Word; this change lets the parses take that difference into account when looking for an image description (alt text).

DocBook reader:

• Fix <xref> references to tables in DocBook files (#8626, Pavol Otto).
• Parse figure as a Figure element in the AST (#8668).

JATS reader: avoid generating duplicate figure captions (#8669).

RST reader: align with spec in syntax for role names (#8653). In particular, we now allow colons in row names.

Add note on converting from .doc format to FAQs (#8654).

Trap error in getAppUserDataDirectory (#8648). This can raise an error if pandoc is run in a non-user environment.

LaTeX writer: do not use longtable foot with Beamer (#8638, Albert Krewinkel). The table foot is made part of the table body, as otherwise it won’t show up in the output. The root cause for this is that longtable cannot detect page breaks in Beamer.

LaTeX template: Add CJKsansfont and CJKmonofont for XeLaTeX (#8656, Yudong Jin). CJKsansfont and CJKmonofont will be set for xelatex only if CJKmainfont is also provided.

URL style in ConTeXt (#8612, Thomas Hodgson). Previously, a URL like this would be in monospace text: \useURL[url1][https://example.com]. Now, it will match the main text unless the linkstyle variable is set, which controls the styling of all links. Closes #8602.

Asciidoc writer: Properly escape | in table cells (#8665).

asciidoc{,tor} template: fix revision date when author is unset (#8637, arcnmx). Revision line syntax is only valid in combination with an author line, so the date attribute must be set explicitly when the author is missing

HTML writer: allow “track” element to be treated as block-level HTML (#8629).

Include needed polyfill when MathJaX is used (#8625).

JATS writer: include alt-text in <graphic>, <inline-graphic> elements (#8631, Albert Krewinkel).

Chunked HTML writer: Retain metadata in processing sections for chunked HTML (#8620). Previously we suppressed metadata in all but the top page, in order to prevent the title block from being printed on every page. This prevented use of custom variables set by metadata fields. This commit moves to a better solution: a conditional in the default template restricts the title block to the top page.

Lua API:

• Add new function pandoc.system.cputime (Albert Krewinkel). The function returns the CPU time consumed by pandoc and can be used to benchmark Lua computations.
• Add module pandoc.json to handle JSON encoding (#8605, Albert Krewinkel).

Use pandoc-lua-marshal 0.2.1 (Albert Krewinkel). All major AST elements now have __tojson metamethods that return the JSON representation of an element. This allows to JSON-encode these elements with libraries that respect the __tojson metamethod, including dkjson.

Use latest zip-archive. This allows pandoc to open certain epubs that it could not open before.

Use commonmark-extensions 0.2.3.4. This fixes some bugs involving definition lists and inline formatting.

Use latest skylighting-format-context

MANUAL.txt:

• Document chunk-template in defaults file.
• Remove obsolete “raw content in a style” section.
• Revise documentation for --mathml to reflect support in all major browsers (#8667).

docs/custom-readers.md: Update JSON parsing example. The example now uses the built-in pandoc.json library to parse the API output.

doc/press.md: Add article on CiTO in J Cheminform by @egonw.

doc/lua-filters.md: fix typo in run_json_filter (Morgan Willcock).

Fix regression with --print-highlight-style option (#8586).

Add new --chunk-template option (#8581), allowing more control over the filenames in chunked HTML output.

Text.Pandoc.App: Add optChunkTemplate constructor to Opt [API change].

Text.Pandoc.Options: add writerChunkTemplate constructor to WriterOptions [API change].

Text.Pandoc.Chunks: add Data, Typeable, Generic, ToJSON, FromJSON instances for PathTemplate [API change].

Text.Pandoc.Citeproc: Fix bug in metaValueToReference (#8611). This bug caused us to get some repeated content when converting MetaBlock to Inlines.

Textile reader:

• Support footnote backlinks (#8585, Stephen Altamirano).
• Don’t allow brackets in URLs (#8582).

ODT reader: fix blockquote indent detection (#3437, Daniel Kessler).

LaTeX writer: include short figure/table caption if one is given (Albert Krewinkel). Short captions are used by LaTeX when generating the list of figures or list of tables. Adding a short caption will now overwrite the full caption in these lists.

Powerpoint writer: fix handling of simple figures (#8565, Albert Krewinkel). This ensures that simple figures are displayed in the same way as before the introduction of a dedicated Figure constructor in the AST.

Improve handling of % in bib(la)tex parsing (#8597, #8595).

Use released skylighting 0.13.2.1

INSTALL.md: direct people to cabal install pandoc-cli.

doc/lua-filters.md: document ‘Figure’ type and constructor (Albert Krewinkel). Fix typos (Martin Joerg).

Fix link in manual (#8583, Salim B).

Fix use of extensions with custom readers (#8571).

Text.Pandoc.Writers.Shared: export setupTranslations [API change]. Use this in HTML and OpenDocument writers, to ensure that translations are set up properly even when we don’t go through convertWithOpts.

LaTeX reader: fix regression in macro resolution for environments (#8573).

Chunked HTML writer: Fix handling of images with absolute URLs (#8567).

HTML writer:

• Don’t omit newlines in task lists.
• Don’t disable checkboxes in task lists (#8562).

Ensure that automatically set variables pandoc-version, outputfile, title-prefix, epub-cover-image, curdir, dzslides-core can be overridden by --variable on the command line. Previously they would create lists in the template Context, which is not desirable.

Fix man page copying in linux/make_artifacts.sh (#8566). Previously we were copying the pandoc-server.1 pandoc page to pandoc-lua.1.

pandoc.cabal: remove pandoc.cabal, stack.cabal from extra-source-files (#8560). The problem is that if these are in extra-source-files, then they get put in the tarball, and then anyone trying to build the source from an unpacked tarball will run into the problem that cabal.project and stack.yaml refer to pandoc-server, pandoc-lua-engine, and pandoc-cli, which aren’t in the tarball.

Require texmath 0.12.6 for better MathML output.

Fix typo in Lua filter documentation (Carlos Scheidegger).

Fix formatting of link in pandoc-server.md (James Scott-Brown).

Minor changelog fixups.

Split pandoc-server, pandoc-cli, and pandoc-lua-engine into separate packages (#8309). Note that installing the pandoc package from Hackage will no longer give you the pandoc executable; for that you need to install pandoc-cli.

Pandoc now behaves like a Lua interpreter when called as pandoc-lua or when pandoc lua is used (#8311, Albert Krewinkel). The Lua API that is available in filters is automatically available to the interpreter. (See the pandoc-lua man page.)

Pandoc behaves like a server when called as pandoc-server or when pandoc server is used. (See the pandoc-server man page.)

A new command-line option --list-tables, causes tables to be formatted as list tables in RST (#4564, with Francesco Occhipinti).

New command line option: --epub-title-page=true|false allows the EPUB title page to be omitted (#6097).

--reference-doc can now accept a URL argument (#8535) and load a remote reference doc.

--version output no longer contains version info for dependent packages. Instead, it contains a “Features” line that indicates whether the binary was compiled with support for acting as a server, and for using Lua filters and Custom writers.

A new option --split-level replaces --epub-chapter-level and affects both EPUB and chunked HTML output. --epub-chapter-level will still work but is deprecated.

Multiple input files with --file-scope: fix case where the links are URL-encoded, e.g. with %20 (#8467).

Produce error if --csl is used more than once (#8195, Prat).

Remove deprecated --atx-headers option.

Remove deprecated option --strip-empty-paragraphs.

In --verbose mode add message when running citeproc (as with other filters).

Add new mark extension for highlighted text in Markdown, using == delimiters (#7743).

Add new extensions wikilinks_title_after_pipe and wikilinks_title_before_pipe for commonmark and markdown. (#2923, Albert Krewinkel). The former enables links of style [[Name of page|Title]] and the latter [[Title|Name of page]]. Titles are optional in both variants, so this works for both: [[https://example.org]], [[Name of page]]. The writer is modified to render links with title wikilink as a wikilink if a respective extension is enabled. Pandoc will use wikilinks_title_after_pipe if both extensions are enabled.

Add prefixes to identifiers with --file-scope (#6384). This change only affects the case where --file-scope is used and more than one file is specified on the command line. In this case, identifiers will be prefixed with a string derived from the file path, to disambiguate them. For example, an identifier foo in contents/file1.txt will become contents__file1.txt__foo. Links will be adjusted accordingly: if file2.txt links to file1.txt#foo, then the link will be changed to point to #file1.txt__foo. Similarly, a link to file1.txt will point to #file1.txt. A Div with an identifier derived from the file path will be added around each file’s content, so that links to files will still work.

New output format: chunkedhtml. This creates a zip file containing multiple HTML files, one for each section, linked with “next,” “previous,” “up,” and “top” links. (If -o is used with an argument without an extension, it is treated as a directory and the zip file is automatically extracted there, unless it already exists.) The top page will contain a table of contents if --toc is used. A sitemap.json file is also included. The option --split-level determines the level at which sections are to be split.

Support complex figures (Albert Krewinkel, Aner Lucero). There is now a dedicate Figure block constructor for figures. The old hack of representing a figure as Para [Image attr [..alt..] (source, "fig:title")] has been dropped. Here is a summary of figure support in different formats:

• Markdown reader: paragraphs containing just an image are treated as figures if the implicit_figures extension is enabled. The identifier is used as the figure’s identifier and the image description is also used as figure caption; all other attributes are treated as belonging to the image.
• Markdown writer: figures are output as implicit figures if possible, via HTML if the raw_html extension is enabled, and as Div elements otherwise.
• HTML reader: <figure> elements are parsed as figures, with the caption taken from the respective <figcaption> elements.
• HTML writer: the alt text is no longer constructed from the caption, as was the case with implicit figures. This reduces duplication, but comes at the risk of images that are missing alt texts. Authors should take care to provide alt texts for all images. Some readers, most notably the Markdown reader with the implicit_figures extension, add a caption that’s identical to the image description. The writer checks for this and adds an aria-hidden attribute to the <figcaption> element in that case.
• JATS reader: The <fig> and <caption> elements are parsed into figure elements, even if the contents is more complex.
• JATS writer: The <fig> and <caption> elements are used write figures.
• LaTeX reader: support for figures with non-image contents and for subfigures.
• LaTeX writer: complex figures, e.g. with non-image contents and subfigures, are supported. The subfigure template variable is set if the document contains subfigures, triggering the conditional loading of the subcaption package. Contants of figures that contain tables are become unwrapped, as longtable environments are not allowed within figures.
• DokuWiki, Haddock, Jira, Man, MediaWiki, Ms, Muse, PPTX, RTF, TEI, ZimWiki writers: Figures are rendered like Div elements.
• Asciidoc writer: The figure contents is unwrapped; each image in the the figure becomes a separate figure.
• Classic custom writers: Figures are passed to the global function Figure(caption, contents, attr), where caption and contents are strings and attr is a table of key-value pairs.
• ConTeXt writer: Figures are wrapped in a “placefigure” environment with \startplacefigure/\endplacefigure, adding the features caption and listing title as properties. Subfigures are place in a single row with the \startfloatcombination environment.
• DocBook writer: Uses mediaobject elements, unless the figure contains subfigures or tables, in which case the figure content is unwrapped.
• Docx writer: figures with multiple content blocks are rendered as tables with style FigureTable; like before, single-image figures are still output as paragraphs with style Figure or Captioned Figure, depending on whether a caption is attached.
• DokuWiki writer: Caption and “alt-text” are no longer combined. The alt text of a figure will now be lost in the conversion.
• FB2 writer: The figure caption is added as alt text to the images in the figure; pre-existing alt texts are kept.
• ICML writer: Only single-image figures are supported. The contents of figures with additional elements gets unwrapped.
• OpenDocument writer: A separate paragraph is generated for each block element in a figure, each with style FigureWithCaption. Behavior for single-image figures therefore remains unchanged.
• Org writer: Only the first element in a figure is given a caption; additional block elements in the figure are appended without any caption being added.
• RST writer: Single-image figures are supported as before; the contents of more complex images become nested in a container of type float.
• Texinfo writer: Figures are rendered as float with type figure.
• Textile writer: Figures are rendered with the help of HTML elements.
• XWiki: Figures are placed in a group.

Changes in custom readers/writers:

• It is now possible to have a custom reader and a custom writer for a format together in the same file. The file may also define a custom template for the writer.
• Pandoc now checks the folder custom in the user’s data directory for a matching script if it can’t find one in the local directory. Previously, the readers and writers data directories were searched for custom readers and writers, respectively. Scripts in those directories must be moved to the custom folder.
• Custom readers used to implement a fallback behavior that allowed to consume just a string value as input to the Reader function. This has been removed, the first argument is now always a list of sources. Use tostring on that argument to get a string.

New module Text.Pandoc.Writers.ChunkedHTML, exporting writeChunkedHtml [API change].

We now set the pandoc-version variable centrally rather than in the writers. One effect is the man writer now emits a comment with the pandoc version.

pandoc-server:

• Add simple CORS support to pandoc-server (#8427).
• Print message to stderr when starting the server.

Docx reader:

• Mark unnumbered headings with class unnumbered (#8148, Albert Krewinkel). This change ensures good conversion results when converting with --number-sections.
• Support parsing of highlighted text.
• Fix handling of oMathPara in w:p with other content (#8483).

ODT reader:

• Fix relative links. ODT adds a ../ to relative links (see #3524); this needs to be removed when converting from ODT.
• Handle “section” elements (#8409).
• Rename Text.Pandoc.Readers.Odt -> Text.Pandoc.Readers.ODT, for consistency with Writers.ODT. Rename readOdt -> readODT. [API change]

DocBook reader:

• Support href on link even in a fragment (#8437). (We now just look for an href attribute without worrying about the namespace.)
• Parse title from imageobject/objectinfo (#8437).

JATS reader:

• Handle uri element in references (#8270).

Ipynb reader:

• Add cell id to attachment filename when storing in MediaBag (#8415). Otherwise attachments with the same name can overwrite each other.

LaTeX reader:

• Skip parenthenized args of toprule, midrule, etc (#8242).
• Handle ## macro arguments properly (#8243).
• Remove unused function toksToString in Parsing module.
• Support more soul commands, including \hl.
• Add unnumbered class for \part* (#8447)
• Fix TEXINPUTS handling (#8392). If TEXINPUTS ends with :, then the system default TEXINPUTS is added. We handle this by just adding the working directory in this case.
• Parse short table caption (see jgm/pandoc-types#103). This is not too useful yet, because writers don’t do anything with the short caption.

MediaWiki reader:

• Parse table cell with attributess, to support rowspan, colspan (#8231, Ruqi).
• Refine “blending” rules for MediaWiki links (#8525, Ruqi). The rules for “blending” characters outside a link into the link are described here: https://en.wikipedia.org/wiki/Help:Wikitext#Blend_link These pose a problem for CJK languages, which generally don’t have spaces after links. However, it turns out that the blending behavior, as implemented on Wikipedia, is (contrary to the documentation) only for ASCII letters. This commit implements that restriction, which fixes the problem for CJK.

HTML reader:

• Fix regression for <tt> (#8330). It was no longer being parsed as Code (Justin Wood).

RST reader:

• Support mark role for round-trip.

Textile reader:

• Support linked images (#8541).
• Fix strong emph ending with link (#8540).
• Adding a Parser to look for ordered list start attribute numbers if any (#2465, vkraven).
• Handle empty paragraphs (#8487). Also, if attributes are added explicitly to a paragraph, put it in a Div with the attributes.

Markdown reader:

• Allow fenced code block “bare” language to be combined with attributes (#8174, Siphalor), e.g.

  ```haskell {.class #id}
  ```

• Allow table caption labels to start with lowercase t (#8259).

• Grid tables: allow specifying a table foot by enclosing it with part separator lines, i.e., row separator lines consisting only of + and = characters (#8257, Albert Krewinkel). E.g.:

  +------+-------+
  | Item | Price |
  +======+=======+
  | Eggs | 5£    |
  +------+-------+
  | Spam | 3£    |
  +======+=======+
  | Sum  | 8£    |
  +======+=======+

• Fix implicit_header_references with duplicate headings (#8300). Documentation says that when more than one heading has the same text, an implicit reference [Heading text][] refers to the first one. Previously pandoc linked to the last one instead. This patch makes pandoc conform to the documented behavior.

• Parse highlighted text inside ==..== if mark extension enabled.

Org reader:

• Allow org-ref v2 citations with & prefix (#8302).
• Make #+pandoc-emphasis-pre work as expected (#8360, Amir Dekel).

BibTeX reader:

• Fix handling of % in url field (#7678). % does not function as a comment character inside url (where URL-encoding is common).
• Allow url field in bibtex as well as biblatex (#8287). This field is not officially supported for BibTeX, but many styles can handle it (https://www.bibtex.com/f/url-field/), and others will ignore it.
• Support software type in biblatex <-> CSL conversions (#8504).
• Make sure version field comes through in biblatex (#8504).

BibTeX writer:

• Pass through url even for bibtex (#8287).

Org writer:

• Pass through unknown languages in code blocks (#8278), instead of producing begin_example.
• Use span attributes tag-name in headers as tags (#8513, Albert Krewinkel). This enables round-tripping of tags in Org headings.

EndNote reader:

• Better error when parsing EndNote references fails.

DocBook writer:

• Rename Text.Pandoc.Writers.Docbook -> Text.Pandoc.Writers.DocBook. Rename writeDocbook -> writeDocBook, for consistency with the DocBook reader’s naming. [API change]
• Fix position of textobject (#8437). It is a child of inlinemediaobject, not imageobject.
• Add regression tests for #8437.
• Render image alt text using textobject element (#8437).
• Don’t indent contents of title element.
• Store “unnumbered” class in DocBook role attribute (#1402, lifeunleaded).

ConTeXt writer (Albert Krewinkel):

• Support syntax highlighting for code.
• Always use \type for inline code. Inline codes that contained curly braces where previously rendered with \mono; this led to unexpected results when the presentation of \type was customized, as those changes would not have been applied to code rendered with \mono.
• Add support for unlisted, unnumbered headings (#8486).
• Support tagging extension (Albert Krewinkel). Paragraphs are enclosed by \bpar and \epar commands, and highlight commands are used for emphasis. This results in much better tagging in PDF output.

LaTeX writer:

• Do not repeat caption on headless tables (Albert Krewinkel). The caption of headless tables was repeated on each page that contained part of the table. It is now made part of the “first head”, i.e. the table head that is printed only once.
• Add separator line between table’s body and its foot (Albert Krewinkel).
• Ignore languages with no babel equivalent, instead of generating an invalid command in the preamble (#8325).
• Use \includesvg for SVGs and include the svg package (#8334).
• Use soul instead of ulem for strikeout, underline (#8411). This handles things like hyphenation, line breaks, and nonbreaking spaces better.
• Use \toprule\noalign{} instead of \toprule() in tables, and similarly for \midrule and \bottomrule (#8223). This facilitates redefining \toprule, \midrule, and \bottomrule without needing to gobble the ()s. (Those who redefine these macros on the assumption that they will be followed by () may need to change their definitions.)
• Support highlighted text for Span with class mark.

JATS writer:

• Use <break/> for LineBreak in the limited contexts that accept it (#8344).
• Officially deprecate writeJATS in favor of writeJatsArchiving.

RTF writer:

• Add space after unicode escape commands (#8264). This fixes a bug that caused characters to disappear after unicode escapes.

RST writer:

• Render tables as list tables when the --list-tables option is specified (writerListTables) (#4564, Francesco Occhipinti).
• Improve inline escaping rules (#8380).
• Use special mark role for Span with class mark.

Commonmark writer:

• Ensure that we don’t have blank lines in raw HTML (#8307).

HTML writer:

• Only add role attribute in HTML5 (#8241). It is not valid in HTML4.
• Avoid aria-hidden in code blocks for HTML4 (#8241).
• Only treat . . . as a slide pause in slides, and not in regular HTML output (#8281).
• Properly merge classes for headings of level > 6 (#8363).
• Prevent <a> inside <a> (#7585). If a link text contains a link, we replace it with a span.
• Replace deprecated aria roles for bibliography entries (#8354). doc-biblioentry -> listitem, doc-bibliography -> list.
• Remove obsolete stuff about mathml-script. This was a shim we used to include for mathml support. We don’t do anything with this any more, so this is dead code.
• Include math links if there are raw commands or environments that can be interpreted as math e.g. by MathJax (#8469).
• Add prooftree to list of math environments (#8462). This will cause raw LaTeX prooftree environments to be rendered appropriately when --mathjax is used.

HTML, Markdown writers: filter out empty class attributes (#8251). These should not be generated by any pandoc readers, but they might be produced programmatically.

Markdown writer:

• Avoid HTML fallbacks in the generated TOC (Albert Krewinkel, #8131). The generated table of contents usually has IDs for each TOC link, allowing to link back to specific parts of the TOC. However, this leads to unidiomatic markup in formats like gfm, which do not support attributes on links and hence fall back to HTML. The IDs on TOC items are now removed in that case, leading to more aesthetic TOCs.
• Escape ! before [ (#8254).
• Support mark extension.

AsciiDoc writer:

• In link text, only replace commas with entities when they’re in Str elements. If a link contains an image, it may have attributes, and the commas there should not be converted (see #8437, #8070).

ODT writer:

• Fix relative links (#3524).

Docx writer:

• Better handling of tables in lists (#5947). Previously the content of each list cell was indented when the table belonged to a list item.
• Indent tables in list items (#5947).
• Adjust correct attribute on lang element (#7022). For East Asian languages, we need to adjust w:eastAsia rather than w:val. This allows normal fonts to be used for any Latin-font text. Similarly, for bidi languages, we need to adjust w:bidi rather than w:val. We treat he and ar as bidi languages, zh, ja, ko as East Asian languages.
• Support relative image widths (Albert Krewinkel). Image widths given in percent are interpreted to be relative to the text width. Previously, percent widths were taken relative to the image’s native size, inconsistently with other writers.
• Avoid using ‘error’ for unassigned table cells (#8468). Instead, throw a regular pandoc error.
• Render a Span with class mark as highlighted. Currently yellow is hardcoded.

MediaWiki writer:

• Use the ‘new’ table structure, so that colspan and rowspan are supported (Wout Gevaert).

Man writer:

• Use UTF-8 by default for non-ascii characters (#8507). Only use groff escapes if --ascii has been specified on the command line (writerPreferAscii).

ICML writer:

• Use Contents element for images with raw data instead of a link with a data: uri (#8398).

EPUB writer:

• Refactor to use Text.Pandoc.Chunks.
• Refactored and simplified code.
• Make title page optional (#6097).

Ms writer:

• Properly format display equations (#8308).
• Remove -C option on PSPIC. Some old versions don’t support this option, and since it’s the default it shouldn’t be necessary.

XWiki writer:

• Use template if it is specified (#8296). Previously templates were ignored.

LaTeX template:

• Set fonts after Beamer theme (Jeremie Knuesel). Beamer themes such as metropolis and saintpetersburg change the default fonts. This change gives precedence to the user font settings by moving them after the loading of the Beamer theme.
• Set \babelfont when mainlang and lang are specified and pdflatex is not being used (#8538). This is needed for good results in Arabic.
• Add variable urlstyle (#8429, Amar Al-Zubaidi). This is set to same by default, so users should not experience any change.

HTML template:

• Remove default font size, line height and font family in default inline css (#8423). mainfont, fontsize, and linestretch can still be used as before; the only difference is that we no longer provide opinionated defaults. This commit also adds a maxwidth variable that sets max-width; if not set, 36em is used as a default.
• Add code { hyphens: manual; }.
• Use styles.citations.html partial in styles.html.
• Fix class name hanging -> hanging-indent in styles.citations.html.
• Put Consolas before Lucida Console for code font (#8543). This is to prevent Lucida Console from being used on Windows, where it causes spacing issues in some applications, with boldface glyphs wider than regular ones.

EPUB CSS changes: Reduce the amount of inline CSS used for EPUBs (#8379). Almost everything is now in the default EPUB CSS (data/epub.css), which can be overridden either by putting epub.css in the user data directory or by using --css on the command line. Inline styles are only used for syntax highlighting (which depends on the style specified, and is only included on pages with highlighted code) and for bibliography formatting (which can depend on the CSL style, and is only used in the page containing the bibliography).

Note that, for compatibility with older readers, we don’t use flexbox to style column/columns divs by default, as we do in HTML. Instead, we use an older method which only works when there are two column divs inside a columns

504. If you need more than two columns and aren’t worried about support for older EPUB readers, you can modify the default CSS (there is a comment in the CSS telling you what to do).

Reveal.js template: prevent line-wrapping of parallax options (#8503, Albert Krewinkel).

reference.pptx: Remove unsupported element (#8342, #6338, Link Swanson). The default template contained text above the header, which can mislead users into thinking there is a way to put text there using pandoc.

Text.Pandoc.Readers.Metadata:

• Fix metadata parsing corner case (#8465).
• Don’t fail on inline metadata beginning with newline (#8358).

Text.Pandoc.App:

• Move initial input-to-Pandoc code to internal submodule (Albert Krewinkel).
• Change parseOptionsFromArgs and parseOptions (#8406) They now return Either OptInfo Opt. [API change]
• Add OptInfo type [API change].
• Add handleOptInfo function. This performs the IO actions for things like --version that were previously done in parseOptionsFromArgs [API change].
• convertWithOpts: add argument for a ScriptingEngine [API change].
• Unify check for standalone output (Albert Krewinkel).
• New optEpubTitlePage field on Opt [API change] (#6097).
• Remove optEpubChapterLevel, add optSplitLevel [API change].
• Export IpynbOutput(..) [API change].

Text.Pandoc.App.OutputSettings:

• Remove unused field outputWriterName in OutputSettings.

Text.Pandoc.Citeproc:

• Check both extension and mime type to determine bibliography type when the bibliography is fetched remotely (#7151).
• CslJson: allow an object with items property in addition to an array of references. This is what is returned by e.g. https://api.zotero.org/groups/904125/items?v=...&format=csljson
• Require a digit for an implicit “page” locator inside explicit locator syntax {...} (#8288). Previously a locator specified as {} would be rendered as p. with nothing after it.
• Update sub verbo to sub-verbo (#8315). This is a change in the term’s canonical name in citeproc. As a result of this change, sub verbo locators have not worked in pandoc since citeproc 0.7.
• Text.Pandoc.Citeproc.MetaValue: remove unused function metaValueToPath.
• Add internal module Text.Pandoc.Citeproc.Name (#8345). This exports toName, which previously had been part of T.P.Citeproc.BibTeX, and allows for cleaner module dependencies.

Export module Text.Pandoc.Slides [API Change] (Albert Krewinkel).

Add new module Text.Pandoc.Format [API change] (Albert Krewinkel). The module provides functions and types for format spec parsing and processing. The function parseFormatSpec was moved from Text.Pandoc.Extensions to the new module and renamed to parseFlavoredFormat. It now operates in a PandocMonad and is based on the updated types.

Text.Pandoc.Sources:

• Add UpdateSourcePos instances for String and strict and lazy ByteString [API change].

Text.Pandoc.Extensions:

• Fix JSON decoding of Extensions (#8352, Albert Krewinkel).
• Add new exported function readExtension [API change].
• Remove parseFormatSpec [API change]. This has been moved to Text.Pandoc.Format and renamed as parseFlavoredFormat (Albert Krewinkel).
• Simpler implementation of Extensions based on Set (benchmarks show no performance penalty).
• Add CustomExtension constructor to Extension [API change].
• Remove Bounded, Enum instances for Extension.
• Add extensionsToList function.
• Revise readExtension so it can handle CustomExtension, and so that it returns a Text rather than Maybe Text.
• Add showExtension [API change].
• Add Ext_mark extension [API change].
• Add Ext_tagging constructor [API change] (Albert Krewinkel).
• Add Ext_wikilinks_title_after_pipe, Ext_wikilinks_title_before_pipe [API change] (Albert Krewinkel).

Text.Pandoc.PDF:

• Fix papersize on PDF generation via ms (#8403). We need to set an option in pdfroff in addition to including a macro in the ms file. With this fix, -Vpapersize=a4 should be sufficient to produce A4 PDF via ms.
• Change default background color of PDFs generated via HTML (#8422, Marcin Serwin).

Text.Pandoc.MIME:

• Base module on package mime-types, which is already a transitive dependency (#8277, Albert Krewinkel).
• Remove deprecated overrides (#8292).

Text.Pandoc.XML:

• Re-export lookupEntity from commonmark-hs [API change].

Text.Pandoc.Parsing:

• Remove gratuitious renaming of Parsec types. We were exporting Parser, ParserT as synonyms of Parsec, ParsecT. There is no good reason for this and it can cause confusion. Also, when possible, we replace imports of Text.Parsec with Text.Pandoc.Parsing. The idea is to make it easier, at some point, to switch to megaparsec or another parsing engine if we want to. New (re-)exports: Stream(..), updatePosString, SourceName, Parsec, ParsecT. Removed exports: Parser, ParserT [API change].
• Export errorMessages, messageString [API change].
• Export fromParsecError, which can be used to turn a parsec ParseError into a regular PandocParseError (#8382) [API change].
• Remove nested [API change]. It was not being used, and in fact it was a bad idea from the beginning, as it had no hope of solving the problem it was introduced to solve.
• Change characterReference, charsInBalanced. characterReference so they now return a Text (some named references don’t correspond to a single Char). Use the the lookupEntity function from commonmark-hs instead of the slow one from tagsoup [API change].
• charsInBalanced now takes a Text parser rather than a Char parser as argument [API change].

Text.Pandoc.Shared:

• Export textToIdentifier [API change].
• Remove deprecated crFilter. [API change]
• Remove deprecated deLink. [API change]
• Deprecate notElemText.
• Deprecate makeMeta.
• Remove pandocVersion (now available in Text.Pandoc.Version as pandocVersionText).
• Remove findM [API change]. This was only used in one place, and can be replaced with simpler code.
• Remove deprecated makeMeta [API change].
• Remove ordNub [API change]. This is just nubOrd from Data.Containers.ListUtils.
• Remove mapLeft [API change]. This is just a synonym for Bifunctor.first.
• Remove elemText, notElemText [API change].
• Drop export of pandocVersion and pandocVersionText, which are now exported by Text.Pandoc.Version.
• Remove escapeURI, isURI. These are now exported by Text.Pandoc.URI, and removing them from Shared helps make the module structure more straightforward.
• Use LineBreak as default block sep in blocksToInlines. (#8499, Albert Krewinkel). This change also affects the pandoc.utils.blocks_to_inlines Lua function.
• defaultUserDataDir is no longer exported (it has been moved to T.P.Data) [API change].
• New function figureDiv, offering offers a standardized way to convert a figure into a Div element (Albert Krewinkel) [API change].

Text.Pandoc.Writers.Shared:

• Export htmlAddStyle, htmlAlignmentToString and htmlAttrs [API change] (Wout Gevaert).
• Use ‘literal tag’ instead of ‘text (T.unpack tag)’ in tagWithAttrs (Wout Gevaert).
• toTableOfContents: handle nested Divs better (#8402).

Rename Text.Pandoc.Network.HTTP -> Text.Pandoc.URI. This is still an unexported internal module. Export urlEncode, escapeURI, isURI, schemes, uriPathToPath. Drop exports of schemes and uriPathToPath.

Text.Pandoc.URI isURI: don’t require non-ASCII characters to be escaped (#8508).

Rename Text.Pandoc.Readers.LaTeX.Types -> Text.Pandoc.TeX (internal module).

Text.Pandoc.Options:

• WriterOptions now has a field writerListTables, specifying that list tables be used in RST output [API change].
• New writerEpubTitlePage field on WriterOptions (#6097) [API change].
• Remove writerEpubChapterLevel, add writerSplitLevel [API change].

Text.Pandoc.Filter:

• Export applyFilters [API change].
• Export applyJSONFilter [API Change] (Albert Krewinkel).
• Parameterize applyFilters over scripting engine [API change] (Albert Krewinkel).

New exported module Text.Pandoc.Chunks [API change]. This module provides functions to split Pandoc documents into chunks to be rendered in separate files, e.g. one per section. Internal identifiers are rewritten appropriately to point to the new locations (#6122).

Text.Pandoc.Readers:

• Change argument type of getReader, so it takes a FlavoredFormat instead of a Text [API change] (Albert Krewinkel).

Text.Pandoc.Writers:

• Change argument type of getWriter, so it takes a FlavoredFormat instead of a Text [API change] (Albert Krewinkel).

Text.Pandoc.Templates:

• Do not try to normalize input to getDefaultTemplate (Albert Krewinkel). The function getDefaultTemplate no longer splits off extension modifers from the given format, as that conflicts with using custom writers as formats. Haskell library users should use getDefaultTemplate <=< (fmap formatName . parseFlavoredFormat) if the input format can still contain extensions. The same is true for compileDefaultTemplate, which calls getDefaultTemplate internally
• Add Wrapper type documentation (#8490, William Rusnack).

New exported module Text.Pandoc.Scripting (Albert Krewinkel). The module contains the central data structure for scripting engines (e.g., Lua) [API change].

Text.Pandoc.Error:

• Add new PandocError constructor PandocNoScriptingEngine [API change] (Albert Krewinkel).
• Add new PandocError constructor PandocFormatError [API change] (Albert Krewinkel). The new error is used to report problems with input or output format specifications.
• Add new PandocError constructor PandocNoTemplateError (Albert Krewinkel).
• Remove PandocParsecError constructor from PandocError (#8385). Henceforth we just use PandocParseError.

New module Text.Pandoc.Version, exporting pandocVersionText and pandocVersion [API change]. pandocVersion returns a Version instead of a Text, which is consistent with pandocTypesVersion.

Text.Pandoc.Class:

• Make getPOSIXTime, getZonedTime sensitive to SOURCE_DATE_EPOCH environment variable if set (#7093). (getTimestamp was already sensitive.) This ensures that EPUB builds are reproducible.
• Text.Pandoc.Class no longer exports readDataFile, readDefaultDataFile, setTranslations, and translateTerm [API change].
• Text.Pandoc.Class now exports checkUserDataDir [API change].

T.P.Class.IO: export function writeMedia [API change] (Albert Krewinkel). This is useful for the pandoc.mediabag module.

Separate out Text.Pandoc.Data and Text.Pandoc.Translations from Text.Pandoc.Class (#8348). This makes Text.Pandoc.Class more self-contained.

• Text.Pandoc.Data is now an exported module, providing readDataFile and readDefaultDataFile (both formerly provided by Text.Pandoc.Class), and also getDataFileNames (formerly unexported in Text.Pandoc.App.CommandLineOptions) and defaultUSerDataDir (formerly provided by Text.Pandoc.Shared). [API change]
• Text.Pandoc.Translations is now an exported module (along with Text.Pandoc.Translations.Types), providing readTranslations, getTranslations, setTranslations, translateTerm, lookupTerm, readTranslations, Term(..), and Translations [API change].

Text.Pandoc now exports Text.Pandoc.Data and setTranslations and translateTerm {API change].

Export module Text.Pandoc.Class.IO [API change]. The module is useful when defining instances of class PandocMonad for types that are also instances of MonadIO.

Remove modules Text.Pandoc.Writers.Custom and Text.Pandoc.Readers.Custom [API Change] (Albert Krewinkel). The functions writeCustom and readCustom are available from module Text.Pandoc.Lua.

Text.Pandoc.Server:

• Split this module into a separate package, pandoc-server, allowing the pandoc library to be compiled without server support.

• Return object if JSON is accepted. Previously we just returned a JSON-encoded string. Now we return something like:

  {
      "output": "<p>hello</p>"
      "base64": false,
      "messages": [
          {
              "message": "Not rendering RawInline (Format \"tex\") \"\\\\noe\"",
              "verbosity": "INFO"
          }
      ],
  }

  This is a change in the pandoc-server JSON API.

• Set translations in the writer based on lang metadata.

• Return error in JSON object if response is JSON.

• Remove parseServerOpts. [API change]

Text.Pandoc.Lua:

• This module has been moved to a separate package, pandoc-lua-engine.

• Export applyFilter, readCustom, and writeCustom. No longer export the lower-level function runFilterFile [API change].

• Change type of applyFilter [API Change] (Albert Krewinkel). The module Text.Pandoc.Filter.Lua has been merged into Text.Pandoc.Lua. The function applyFilter now has type

  applyFilter :: (PandocMonad m, MonadIO m)
              => Environment-> [String]-> FilePath-> Pandoc-> m Pandoc

  where Environment is defined in Text.Pandoc.Filter.Environment.

• Export new function getEngine [API Change]. The function returns the Lua scripting engine.

• Add unexported modules T.P.Lua.Reader, T.P.Lua.Writer. These contain the definitions of readCustom and writeCustom that were previously in T.P.Readers.Custom and T.P.Writers.Custom.

• Cleanup module dependencies, for a cleaner module dependency graph.

• The writeCustom function has changed to return a Writer and an ExtensionsConfig [API change]. This allows ByteString writers to be defined.

• The readCustom function has changed to return a Reader and an ExtensionsConfig [API change].

Lua subsystem (Albert Krewinkel):

• The whole Lua subsystem has been moved to a separate package, pandoc-lua-engine. pandoc does not depend on it. convertWithOpts has a new parameter that can be used to pass in the scripting engine defined in pandoc-lua-engine (or a different one, in theory).
• Fix the behavior of Lua “Version” objects under equality comparisons (#8267).
• Support running Lua with a GC-collected Lua state.
• Ensure that extensions marshaling is consistent.
• Produce more informative error messages for pandoc errors. Errors are reported in Lua in the same words in which they would be reported in the terminal.
• Add new module pandoc.format. The module provides functions to query the set of extensions supported by formats and the set of extension enabled per default.
• Add function pandoc.template.apply.
• Add function pandoc.template.meta_to_context. The functions converts Meta values to template contexts; the intended use is in combination with pandoc.template.apply.
• Allow Doc values in WriterOptions.variables. The specialized peeker and pusher function for Context Text values does not go via JSON, and thus keeps Doc values unchanged during round-tripping.
• Fix rendering of Lua errors in Lua, so that the Error running Lua message is not prepended multiple times.
• Add new module pandoc.zip.
• Allow strings in place of compiled templates (#8321). This allows to use a string as parameter to pandoc.template.apply and in the WriterOptions template field.
• Rename reader_extensions/writer_extensions globals as Extensions (#8390).
• Add pandoc.scaffolding.Writer (#8377). This can be used to reduce boilerplate in custom writers.
• Fix peeker for PandocError (Albert Krewinkel). String error messages were incorrectly popped of the stack when retrieving a PandocError.
• Add functions pandoc.text.toencoding, pandoc.text.fromencoding (#8512, Albert Krewinkel).
• Add pandoc.cli module. Allow processing of CLI options in Lua.
• Support -D CLI option for custom writers. A new error PandocNoTemplateError (code 87) is thrown if a template is required but cannot be found.
• Allow table structure as format spec. This allows to pass structured values as format specifiers to pandoc.write and pandoc.read.
• Add function pandoc.mediabag.write (Albert Krewinkel).
• Add module pandoc.structure (Albert Krewinkel). The function make_sections has been given a friendlier interface and moved to the new module; the old pandoc.utils.make_sections has been deprecated.

Custom writers:

• The global variables PANDOC_DOCUMENT and PANDOC_WRITER_OPTIONS are no longer set when the writer script is loaded. Both variables are still set in classic writers before the conversion is started, so they can be used when they are wrapped in functions.

• Deprecate classic custom writers.

• Add function pandoc.write_classic. The function can be used to convert a classic writer into a new-style writer by setting it as the value of Writer:

  Writer = pandoc.write_classic

  or to fully restore the old behavior:

  function Writer (doc, opts)
    PANDOC_DOCUMENT = doc
    PANDOC_WRITER_OPTIONS = opts
    load(PANDOC_SCRIPT_FILE)()
    return pandoc.write_classic(doc, opts)
  end

• Support extensions in custom writers. Custom writers can define the extensions that they support via the global writer_extensions. The variable’s value must be a table with all supported extensions as keys, and their default status as values. For example, the below specifies that the writer supports the extensions smart and sourcepos, but only the smart extension is enabled by default:

  writer_extensions = {
    smart = true,
    sourcepos = false,
  }

• Custom writers can define a default template via a global Template function; the data directory is no longer searched for a default template. Writer authors can restore the old lookup behavior with

  Template = function ()
    local template
    return template.compile(template.default(PANDOC_SCRIPT_FILE))
  end

Custom readers:

• Support extensions in custom readers. Custom readers, like writers, can define the set of supported extensions by setting a global. E.g.:

  reader_extensions = {
    smart = true,
    citations = false,
  }

Use latest versions of commonmark-extensions, texmath, citeproc, gridtables, and skylighting.

Use pandoc-types 1.23. This adds the Figure Block constructor and removes the Null Block constructor.

Require aeson >= 2.0.

Use jira-wiki-markup 1.5.0 (#8511, Albert Krewinkel). Fixes issues with icon-like sequences at the beginning of words.

Use doctemplates 0.11, avoiding a transitive dependency on HsYAML.

Use skylighting 0.13.1.2.

Allow mtl 2.3.1 (Alexander Batischev).

Use latest skylighting-format-context.

Allow building with mtl 2.3.

Remove lua53 flag. We now only support Lua 5.4.

Add hie.yaml for haskell language server.

Add tools/latex-package-dependencies.lua.

Update default CSL with latest chicago-author-date.csl.

make_artifacts.sh: various small improvements.

Remove sample.lua from data files (#8356).

Documentation:

• Deprecate PANDOC_WRITER_OPTIONS in custom writers (Albert Krewinkel).
• Document pandoc.write_classic (Albert Krewinkel).
• Document new table features (Albert Krewinkel).
• Clarify what background-image does in reveal.js (#6450).
• Documentation improvements for blank_before_blockquote (#8324, Pranesh Prakash).
• Update grid table documentation (#8346).
• Add note about MathJax fonts to --embed-resources.
• Use cabal’s –package-env more (#8317, Artem Pelenitsyn).
• Modify Zerobrane instructions to use Lua 5.4 (#8353, Ian Max Andolina).
• Fix documentation for highlight-style in pandoc-server.md.
• Fix link to fedora package site (#8246, Akos Marton).
• Rephrase paragraph on format extensions (#8375, Ilona Silverwood).
• Update README.template (#8496, Sven Wick).
• Fix a tiny typo in lua-filters.md (TomBen).
• Clarify that --css should be used with -s.
• Clarify font selection for pdf -t ms (#8421, nbehrnd).
• Clarify docs for --metadata-file (#8459).
• Fix typo in epub.md (Vladimir Alexiev).
• Add missing backtick in filters.md (R. N. West).
• doc/lua-filters.md: add documentation for pandoc.format (Albert Krewinkel).
• Fix epub-embed-font documentation (#8455, Terence Eden).
• Removed obsolete Templates section in CONTRIBUTING.md.
• Add manual section on accessible PDFs, archiving standards (#8312, Albert Krewinkel).

Tests.Command: remove unused runTest.

Add pandoc-lua.1 man page.

Improve shell.nix.

Add tools/moduledeps.lua for inspecting the internal module dependency tree.

Fix macOS zip so pandoc-server is a symlink. This cuts its size by 2x.

CI: Improve CI speed by caching more, eliminating macos builds, and splitting benchmarks into a separate action, run by manual dispatch. (We still test that benchmarks build in the regular CI.) The cache can be expired manually by modifying the secret CACHE_VERSION.

Remove the unnecessary Setup.hs from pandoc. Cabal does not need this with build-type ‘simple’.

Add pandoc-lua and pandoc-server (symlinks) and their man pages to releases.

Use hslua-cli package for pandoc-lua interface (Albert Krewinkel).

Add server flag to pandoc-cli, allowing it to be compiled without server support.

pandoc-cli: Allow building a binary without Lua support (Albert Krewinkel). Disabling the lua cabal flag will result in a binary without Lua.

Move --version handling to pandoc-cli. We need it here in order to print information about whether server and Lua support have been compiled in.

Move nightly flag from pandoc to pandoc-cli (#8339).

Makefile changes:

• make help will now print all the targets and what they do.
• Add targets: coverage, weeder, moduledeps, prerelease, ghcid, repl, linecounts, hie.yaml, binpath.
• Note that you can alias pandoc=`make binpath` for convenient local testing of a build.
• Rename quick-cabal -> build, quick-test -> test.
• Exclude tests from SOURCEFILES.

Factor out xml-light into an internal library.

Add CITATION.cff (#8434).

Move trypandoc to a separate repository, jgm/trypandoc.

Fix regression with data uris in 2.19.1 (#8239). In 2.19.1 we used the base64URL encoding rather than base64.

pandoc-server: handle citeproc parameter as documented (#8235).

Org reader: treat emacs-jupyter src blocks as code cells (#8236, Albert Krewinkel). This improves support for notebook-like org files that are intended to be used with emacs-jupyter package.

HTML writer and templates: revert to using width property for column widths (Albert Krewinkel). The default flex and overflow-x properties of a column are set to auto. In combination, these changes allow to get good results when using columns with or without explicit widths.

Org writer (Albert Krewinkel):

• Add support for jupyter nodebook cells (#6367).
• Prefix code language of ipynb code blocks with jupyter-. This is the convention used by the emacs-jupyter package.
• Keep code block attributes as header args. This allows to keep more information in the resulting src blocks, making it easier to roundtrip from or through Org. Org babel ignores unknown header arguments.
• Add code block identifier as #+name to src blocks.

Fix some typos in the codebase (luz paz).

Require hslua-module-path 1.0.3 (#8228, Albert Krewinkel).

Add server capabilities.

• New exported module Text.Pandoc.Server [API change].
• The pandoc executable now starts up a web server when renamed or symlinked as pandoc-server, and functions as a CGI program when renamed or symlinked as pandoc-server.cgi. See the man page for pandoc-server for full documentation.

Text.Pandoc.App.Opts: Redo FromJSON for Opt so that optional values can be omitted (in which case the values from defaultOptions are used).

Org reader: treat “abstract” block as metadata (Albert Krewinkel, #8204). A block of type “abstract” is assumed to define the document’s abstract. It is transferred from the main text to the metadata.

Org template: add abstract from metadata as block of type “abstract” (#8204).

HTML writer: use flex property for column widths (Albert Krewinkel, #8232).

LaTeX writer:

• Add label to tables that have an identifier (Albert Krewinkel, #8219). Tables with an identifier are marked with a \label. A caption is always included in this case, even if the caption is empty.
• Use \textquotesingle for straight quotes in text.
• Fix widths of multicolumn cells (#8218).

LaTeX template: fix behavior of colorlinks variable (Albert Krewinkel, #8226). Fixes a regression in 2.19 that required the boxlinks variable to be set in addition to the usual link coloring variables. Otherwise links were never colored in LaTeX PDF output.

Text.Pandoc.Highlighting: Export lookupHighlightingStyle [API change]. Previously this lived in an unexported module Text.Pandoc.App.CommandLineOptions, under the name lookupHighlightStyle.

Text.Pandoc.App:

• Remove unneeded MonadIO constraints in readSources.
• Factor out convertWithOpts' from convertWithOpts. This runs in any PandocMonad, MonadIO, MonadMask instance. So far it is not exported, but it might find a use later.

Support --strip-comments in commonmark/gfm (#8222). This change makes the commonmark reader sensitive to readerStripComments.

Lua: add function pandoc.utils.citeproc (Albert Krewinkel). The function runs the citeproc processor on a Pandoc document. Exposing this functionality to Lua allows to make citation processing part of a filter or writer, simplifies the creation of multiple bibliographies, and enables the use of varying citation styles in different parts of a document.

Refactor linux/make_artifacts.sh.

Update INSTALL.md installation from source instructions.

Use base64 package instead of base64-bytestring. It is supposed to be faster and more standards-compliant.

trypandoc improvements:

• Add dropdown with canned examples.
• Add citeproc support.
• Support csv, bibliographic and binary formats.
• Add load from file.
• Add permalink. Don’t always reload page.
• Use vanilla JS and CSS + the new pandoc-server.cgi.

Allow haddock-library-1.11.0.

Convert tool/extract-changes.hs to a Lua filter.

Add --embed-resources flag (Elliot Bobrow, #7331). This can be used to embed resources without implying --standalone. Deprecate --self-contained in favor of --embed-resources --standalone.

Allow environment variable interpolation in highlight-style and pdf-engine fields in defaults files (#8061; Jaehwang Jung, #8073).

Allow placing custom readers and writers in user data directory (Albert Krewinkel, #8112) (readers and writers subdirectories).

Add tsv (tab separated values) as an input format (#7974). [API change]: Text.Pandoc.Readers.CSV now exports readTSV. Internal change: In Text.Pandoc.CSV, CSVOptions has changed so that csvQuote takes a Maybe value.

Add tex_math_dollars to gfm default extensions (reflecting gfm’s new support for math).

RST, Org, Markdown readers: support rowspans and colspans in grid tables (#8202, Albert Krewinkel). Note: the writers does not yet support these more complex grid table features, so these complex grid tables will not round-trip.

HTML, LaTeX, and MediaWiki readers: use formatCode (#8162, #8129, Elliot Bobrow). This moves formatting from inside inline code elements to the outside, since pandoc’s Code element only takes string content.

Markdown reader:

• Don’t parse inline notes with blank lines inside (#8028).
• Allow attributes in special spans (e.g. smallcaps, underline) (Albert krewinkel, #4102). These spans are parsed as SmallCaps or Underline elements, but any attributes are included in a wrapping Span.

HTML reader:

• Allow sublists that are not marked as items (Albert Krewinkel, #8150). This is technically invalid HTML, but it can be found in the wild and browsers handle it.

Org reader (Albert Krewinkel):

• Recognize absolute paths on Windows (Albert Krewinkel, #8201).
• Recognize {webp,jxl} files as images (YI).
• Allow attrs for Org tables (Albert Krewinkel, #8049). Tables with attributes are no longer wrapped in Div elements; attributes are added directly to the table element.
• Support line selection in INCLUDE directives (Brian Leung, #8060).
• Fix Post / Pre mixup when setting emphasis chars (Amir Dekel, #8134).

LaTeX reader:

• Support \includesvg (#8027).
• Unescape characters in \lstinline inside \passthrough (#8179).
• Improve mathEnvWith (#8122). When converting e.g. an align environment to an aligned environment inside a Math element, we need to include a newline before the \end{aligned}, since the previous line might end in a comment.
• Fix treatment of extensions for \input in LaTeX reader (#8092). Previously we required a .tex extension, but TeX allows any extension for \input (as opposed to \include).

RTF reader:

• support \nosupersub (#8170).

TikiWiki reader:

• Support underlined text

DocBook reader:

• Improved reading <xref> elements (Frerich Raabe, #8065).

JATS reader:

• Strip ref- prefix from ref id in xref (#8007).
• Support edition in references (#8087).

RIS reader:

• Make parser more forgiving (#8034). Allow blank lines after entries. Allow entries with no space after the -, provided they just have a newline, e.g. DB -\n.
• Get right order of names (#8055).

MediaWiki reader:

• Allow HTML comment after row start (#8110).

DokuWiki reader:

• The tex_math_dollars extension is now supported for dokuwiki (but off by default) (#8178).
• Content inside <latex>...</latex> is parsed as raw LaTeX inline, and inside <LATEX>..</LATEX> as raw LaTeX block (#8178).
• The behavior of <php>...</php> is changed, so that instead of producing a code block, it produces raw HTML with <?php ... ?>.

LaTeX writer:

• Improve grouping with autocites (#8088).
• Extend list of book documentclasses (Wentau Han, #8053).
• Fix width of multicolumn cells (Albert Krewinkel, #8090). Cells spanning multiple columns must be given an explicit width, calculated from the table properties.
• Beamer: allow containsverbatim as alternative to fragile (#8080).

HTML writer:

• Add ‘footnotes’ identifier to footnotes section (#8043).
• Fix bug with --number-offset. This formerly caused section divs to be produced, even when --section-divs was not specified (#8097).
• Use CSS flexboxes for columns (Albert Krewinkel). This allows an arbitrary number of columns, while the previous approach assumed exactly two columns.
• Allow “spanlike” classes to be combined (see #8194). Previously classes like “underline” and “marked” had to be the first class in a span in order for the span to be interpreted as a “ul” or “mark” element. This commit allows these special classes to be “stacked,” e.g. [test]{.mark .underline}; in addition, the special classes are no longer required to come first in the list of classes.
• Avoid doubled style attribute when height and width are added to style because of an image, but the image already has a style attribute (#8047).
• Do not include the deprecated doc-endnote role (#8030). doc-endnote was deprecated in DPUB-ARIA 1.1.
• Remove extra soft break for tasklist (black-desk, #8142). Browser will display the extra newline character between checkbox and text as a space, which make tasklist items cannot be aligned.

EPUB writer:

• Allow choice of math method for v3 (#8164). Previously we always used MathML for math in EPUB3, because the spec includes MathML. But this is not widely supported by readers, so it seems better to allow users to choose their math method as they can with EPUB2 or HTML. NOTE: Existing workflows that produce EPUBv3 documents including math will be affected by this change. You must add --mathml to your command line if you want to continue producing MathML.

RST writer:

• Fix missing spaces with nested inlines (#8182).
• Always escape literal backslash (#8178).

Ms writer:

• Add comment in preamble stating generator.
• Fix roff ms syntax highlighting definitions (#8175, thanks to Branden Robinson).

ConTeXt writer:

• Support complex table structures (Albert Krewinkel, #8116). The following table feature are now supported in ConTeXt:

  • colspans,
  • rowspans,
  • multiple bodies,
  • row headers, and
  • multi-row table head and foot.

  The wrapping placetable environment is also given a reference option with the table identifier, enabling referencing of the table from within the document.

• Unify link handling (Albert Krewinkel, #8096). Autolinks, i.e. links with content that’s the same as the linked URL, are now marked with the \url command. All other links, both internal and external, are created with the \goto command, leading to shorter, slightly more idiomatic code. As before, autolinks can still be styled via \setupurl, other links via \setupinteraction.

• Use “sectionlevel” environment for headings (Albert Krewinkel, #5539). The document hierarchy is now conveyed using the \startsectionlevel/\stopsectionlevel by default. This makes it easy to include pandoc-generated snippets in documents at arbitrary levels. The more semantic environments “chapter”, “section”, “subsection”, etc. are used if the --top-level-division command line parameter is set to a non-default value.

Docx writer:

• Add w:lang to rPr for Span and Div with lang attribute, so that Word can know that “Apfel” is not a spelling error (#8026).
• Prevent crashing when handling invalid tables (Albert Krewinkel, #8102). Tables with different numbers of cells per row would sometimes crash pandoc. This fix prevents this by cutting off overlong rows.

ICML writer:

• Support custom-style attribute on Table (#8079).

AsciiDoc writer:

• Fix commas in link text (#8070). Commas in link text trigger interpretation of attributes. To block this, we replace them with numeric entities.
• Fix underline. We were rendering it as +++text+++; this is now changed to [.underline]#text#. See comment at https://github.com/jgm/pandoc/issues/8070#issuecomment-1126883824.

FB2 writer:

• Fix handling of non-section Divs (#8123).

Markdown writer:

• Disable soft wrapping when hard_line_breaks enabled (#8035). We were already doing this for markdown; this commit does the same thing for markua and commonmark and gfm.
• Avoid excessive indentation on bullet lists for commonmark, markua, gfm. They are now nested by 2 spaces instead of 4 (#8011).

Text.Pandoc.Class:

• Add new function findFileWithDataFallback [API Change] (Albert Krewinkel).
• fillMediaBag: Keep attributes of original image on Span (Albert Krewinkel, #8099). Images that cannot be fetched are replaced with a Span that contains the image’s description. The span now also retains all original image attributes and inherits all attributes of the image. Furthermore, the classes image and placeholder are added, and path and title are store in attributes original-image-src and original-image-title, respectively.

Text.Pandoc.Shared:

• makeSections: don’t make a section for a div with class “fragments” (#8098).
• Ensure that Nulls are ignored by makeSection and in segmenting slides (#8155).
• Add formatCode function to Text.Pandoc.Shared [API change] (Elliot Bobrow, #8129).
• taskListItemToAscii: handle asciidoctor’s characters (#8011). Asciidoctor uses different unicode characters for task lists; we should recognize them too and be able to convert them to ascii task lists in formats like gfm.
• Deprecate deLink and mark for later removal.

Text.Pandoc.Writers.Shared:

• toTableOfContents: Don’t replace links with empty spans in TOC (#8020).

Text.Pandoc.Readers.Metadata:

• Ensure that metadata values w/o trailing newlines are parsed as inlines, as the manual states. Previously, they were parsed as inlines if they would otherwise have been a single Plain or Para, but otherwise left unchanged. This led to some quirky results (e.g. #8143). We now use the general function blocksToInlines from T.P.Shared.

Text.Pandoc.Parsing:

• Simplify gridTableWith', gridTableWith [API Change] (Albert Krewinkel). The functions gridTableWith and gridTableWith' no longer takes a boolean argument that toggles whether a table head should be parsed: both, tables with heads and without heads, are always accepted now.

Lua subsystem (Albert Krewinkel):

• Extend pandoc.system module (Albert Krewinkel, #8184). The module now has the additional functions list_directory, make_directory, and remove_directory. This makes it easier to write cross-platform scripts that need to inspect or modify the file system.
• Require pandoc-lua-marshal 0.1.7. Adds a clone methods to Pandoc objects and allows to pass Blocks in instead of full Caption elements.
• Add fields pandoc.readers and pandoc.writers (#8177). The set of supported input and output formats is made available to Lua users.
• Ensure that tables marshaled via JSON arrays behave like Lists. This allows to invoke methods like map and includes on lists like PANDOC_WRITER_OPTIONS.extensions.
• Require hslua-2.2.1, unless lua53 flag is set, and do not reset foreign encoding before running Lua. This fixes a problem where the encoding used for Lua filenames would sometimes mismatch the encoding used by the OS.
• Simplify module loading code. Modules are now loaded directly; the special pandoc Lua package searcher is no longer necessary and has been removed.
• Add function pandoc.mediabag.fill (#8104). The function allows to fill the mediabag with all images in a given document. Images that cannot be fetched are replaced with a Span containing the image description.

Populate mediabag after filters have run (Albert Krewinkel, #8099). The mediabag is filled with document resources after the filters have run. This allows, for example, filter authors to modify image paths before pandoc tries to fetch the images. Lua filters that rely on a filled mediabag can use the new pandoc.mediabag.fill function to perform that action in the filter.

Ms template: redefine rather than removing .CH macro (#8175).

JATS template (Albert Krewinkel, except as noted):

• Include particles, prefix, suffix in names.
• Mark authors with cor-id as corresponding authors. Corresponding authors are marked by setting the attribute corresp="yes" in their respective <contrib> element.
• Unconditionally include permissions element (#8040). Fixes a bug that caused license information to be omitted when no copyright information was provided.
• Follow JATS4R recommendation and PudMed Central for license URI (Castedo Ellerman, #8041).

LaTeX template:

• Rename \textormath to \TextOrMath (Hos Es, #8036).
• Fix links-as-notes (Albert Krewinkel, #8077).

HTML template styles:

• Remove span.underline rule. This is superfluous now that we render Underline as <u>.
• Improve CSS for task lists (#8151).

LaTeX template: Add boxlinks variable for LaTeX/PDF output (#8198). If boxlinks is set but colorlinks is not, then boxes will be printed around links (hidelinks will not be set in hypersetup).

--self-contained: Handle url() in <style> elements (#8193).

Text.Pandoc.PDF: use SHA1 hash of filename when converting SVG. The previous code threw away the directory component of the filename in constructing a new one. This led to surprising results if you had e.g. foo/pic.svg and bar/pic.svg; in the final PDF they’d be the same image, because the latter would overwrite the former in the temp directory.

Remove Muse reader round-trip tests. These are nondeterministic and have repeatedly failed on strange edge cases.

Update fonts MIME following IANA recommendation (Gabriel Lewertowski, #8127).

Future compat change for liftA2 export from Prelude (Georgi Lyubenov, #8132).

Update default.csl from the latest chicago-author-date.csl.

Update manfilter.hs for greater portability (#8045). The tables in our man pages were not rendering correctly with mandoc, now used by default with macOS. mandoc doesn’t allow man formatting inside table cells. For maximum portability, we now render the tables in plain format and include them as code blocks in the man page.

CI: update macos container version (#8197, Sukka).

Add nightly flag. This causes a -nightly-COMPILEDATE suffix to be added the the output of --version (#8016). This is used in the nightly CI builds.

Update dependencies (aeson, skylighting, pandoc-lua-marshall, citeproc, texmath).

Documentation improvements (thanks to Jiří Wolker, Castedo Ellerman, Albert Krewinkel, Bastien Dumont, Cezar Drożak, Benjamin Wuethrich, Ivan Panchenko, Sukil Etxenike, Masataka Ogawa).

New input formats: endnotexml (EndNote XML bibliography), ris (RIS bibliography).

A RIS bibliography file may now be used with --citeproc.

Citeproc: Allow a formatted bibliography to be placed in metadata fields via a Div with class refs (#7969, #526). Thus, one can include a metadata field, say refs, whose content is an empty div with id refs, and the formatted bibliography will be put into this metadata field. It may then be interpolated into a template using the variable refs.

Ensure that you don’t get PDF output to terminal. -t pdf now behaves like -t docx and gives an error unless the output is redirected.

--version now prints hslua version (#7929) and Lua version (#7997, Albert Krewinkel).

Change --metadata-file parsing so that, when the input format is not markdown or a markdown variant, pandoc’s markdown is used (#6832, #7926). When the input format is a markdown variant, the same format is used. Reason for the change: it doesn’t make sense to run the markdown parser with a set of extensions designed for a non-markdown format, and this dramatically limits what people can do in metadata files.

Trim whitespace from math in --webtex (#7892). This fixes problems with –webtex and markdown output, when display math starts or ends with a newline.

--self-contained: issue warning rather than failing with an error if a resource can’t be found (#7904).

New exported module Text.Pandoc.Readers.EndNote, exporting readEndNoteXML and readEndNoteXMLCitation [API change].

New exported module, Text.Pandoc.Readers.RIS, exporting readRIS (#7894) [API change].

LaTeX reader:

• Handle subequations as inline math environment (#7883).
• Rudimentary support for vbox (#7939).
• Support \today (#7905).
• Handle \label and \ref for footnotes (#7930).
• Allow inline groups starting with \bgroup (#7953).
• Use custom TokStream that keeps track of whether macros are expanded. This allows us to improve performance a bit by avoiding unnecessary runs of the macro expansion code (e.g. from 24 ms to 20 ms on our standard benchmark).
• Further optimizations for inline parsing.
• Better handling of \usepackage. If the package is local but causes parse errors, parse everything up to the error and skip the rest. Issue a CouldNotParseIncludeFile warning indicating that parsing failed at that point.
• Text.Pandoc.Readers.LaTeX.Parsing: Monoid and Semigroup instances for TokStream.

HTML reader:

• Give warnings and emit empty note when parsing <a epub:type="noteref"> and the identifier doesn’t correspond to anything in the note table (#7884). Previously we just silently skipped these cases.
• Fix parsing of epub footnotes (#7884).

DocBook reader:

• Handle complete set of entities as specified at https://www.w3.org/2003/entities/2007doc/byalpha.html (#7938).
• Handle abstract in info section (#7747).
• Improve info parsing.
• Simplify metadata parsing code (#7747). Handle abstract as block-level content. Report skipped info elements with --verbose.
• Handle address and coyright in metadata (#7747).

DokuWiki reader:

• Add DokuWiki table alignment (#5202, damon-sava-stanley).

RST reader:

• Fix treatment of headerless simple tables (#7902).
• Wrap math in Span to preserve attributes (#7998, Albert Krewinkel). Math elements with a name, classes, or other fields are wrapped in a Span with these attributes.

JATS reader:

• Improve handling of fn-group elements (#6348, Albert Krewinkel). Footnotes in <fn-group> elements are collected and re-inserted into the document as proper footnotes in the place where they are referenced.
• Handle pub-date (#8000).
• Support PMID, DOI, issue in citations (#7995).
• Improve refs parsing. Handle issn and isbn; use simpler form for issued date.
• Strip ‘ref-’ from ref id in constructing CSL id. This allows better round-tripping, because the JATS writer adds the ref- prefix to the citation id to get the ref element’s id.

Org reader:

• Allow “:” in property drawer keys (Lucas V. R). Any non-space character is allowed as property drawer key, including “:” itself (so it is not really a delimiter). The real delimiter is a space character, so in a drawer like

  :PROPERTIES:
  ::k:ey:: value
  :END:

  “:k:ey:” is a key with value “value”.

• Allow comments above property drawer.

• More flexible LaTeX environments (Lucas V. R).

• Handle #+bibliography: as metadata so that it can work with --citeproc.

• Parse #+print_bibliography: as Div with id refs.

• Allow multiple #+bibliography:.

Markdown reader:

• Allow one-column pipe tables with pipe on right (#7919).
• Remove restriction on identifiers, so they no longer need to begin with a letter (#7920).

Docx reader:

• Enable citations extension for docx reader (#7840). When enabled, Zotero, Mendeley, and EndNote citations embedded in a docx are parsed as native pandoc citations. (When disabled, the generated citation text and bibliography are passed through as regular text.) The bibliography generated by the plugin is suppressed. Instead, bibliographic data embedded in citation items is added to the references metadata field so that it can be used with --citeproc.

Docbook writer:

• Interpret links without contents as cross-references (#7360, Jan Tojnar). Links without text contents are converted to <xref> elements. DocBook processors will generate appropriate cross-reference text when presented with an xref element.

Docx writer:

• Single numbering ID for examples (#7895, mjfs). This change ensures that example list items all belong to a single number sequence, so that if items are added or deleted in a word processor, the other items will renumber automatically.
• Add bookmark with table id to table (#7989, Nikolai Korobeinikov, #7285). This allows tables with ids to be linked to.

Ipynb writer:

• Handle metadata better (#7928). Previously we used the markdown writer to render metadata. This had some undesirable consequences (e.g. en dash expanded to -- when smart enabled), so now we use the plain writer.

LaTeX writer:

• Avoid extra space before \CSLRightInline (#7932).
• Add scrreport to chaptersClasses (#6168, ivardb).
• Support page,trim,clip attributes on images (#7181).
• Add () after booktabs rules (#8001). These commands take optional arguments with () and [], which can lead to problems if the content of the table cell begins with these characters.

RST writer:

• Support all standard metadata (“bibliographic”) fields.

HTML writer: performance improvements.

Org writer:

• Stop indenting property drawers, quote blocks (#3245, Albert Krewinkel). This follows the current default org-mode behavior.

Markdown writer:

• Move table-related code into submodule (Albert Krewinkel).
• Don’t produce redundant header identifier when the gfm_auto_identifiers extension is set (#7941).
• Update escaping rules for \. We now escape \ only if raw_tex is enabled or it is followed by a non-alphanumeric.

JATS writer:

• Encode author “others” as <etal/> (Albert Krewinkel). Citeproc adopted the BibTeX convention to use the author name “others” when there are additional authors that are not named. JATS uses the <etal> element for this.
• Avoid doubled ref-list element (#7990). Previously when generating JATS with the element_citations extension enabled, the references were put in a doubly-nested ref-list element (<ref-list><ref-list>...).
• Keep edition info in element citations (#7993, Albert Krewinkel).
• Fix handling of CSL variable ‘page’ (not ‘pages’ as we had before). It should go to ‘lpage’ and ‘rpage’, not ‘page-range’.

EPUB writer: refactor for clarity (#7991, Jonathan Dönszelmann, Ola Wolska, Ivar de Bruin, Jaap de Jong).

Custom writer (Albert Krewinkel):

• Support new-style Writer function (Albert Krewinkel). See the documentation for custom writers for details.
• Produce stacktrace if Writer function fails

Text.Pandoc.Logging: add CouldNotParseIncludeFile constructor for LogMessage [API change].

Text.Pandoc.Shared:

• Put id attributes on TOC entries (#7907, damon-sava-stanley). Naming scheme of id is “toc-” + id of linked to header/section. Effects HTML, Markdown, Powerpoint, and RTF.
• Define ordNub as alias for nubOrd from containers package (#7963, Albert Krewinkel).
• Export ensureValidXmlIdentifiers. This function changes identifiers that don’t start with letters, and internal links to these identifiers, making them compatible with XML standards. The change is simple: we add id_ to the front. There is potential for duplication if there are already id_... identifiers defined, but this seems rare enough not to worry too much about.

Ensure that valid XML identifiers are used in Docbook, EPUB, FB2, HTML4, S5, Slidy, Slideous, ICML, ODT, TEI writers. Thus, if you convert [anchor]{#1} and [link to](#1), id_1 will be used instead of 1 for the identifier.

Lua (Albert Krewinkel).

• Add module pandoc.layout to format and layout text.
• Move custom writer code into Lua hierarchy.
• Use pandoc-lua-marshal 0.1.5.
• Allow any type of callable object as argument to List functions filter, map, and find_if. These previously required the argument to be of type function, which was too restrictive.
• Inline: the type of Image captions is now Inlines instead of List.
• Allow passing Sources to pandoc.read (#8002).

Text.Pandoc.PDF:

• Restore wkhtmltopdf as default pdf engine for HTML (Albert Krewinkel).
• Allow custom writer as format if engine is explicitly specified (#7898, Albert Krewinkel). Note that it may be necessary to explicitly specify a template on the command line.

Text.Pandoc.MediaBag: improve detection of absolute paths (#7881).

Text.Pandoc.Extensions:

• Remove raw_tex extension from list of commonmark extensions, and from the commonmark_x defaults. commonmark doesn’t parse raw TeX, and it doesn’t make sense to write it if we don’t parse it.
• Remove native_divs from allowed gfm extensions (#7965). This allows <div> to be suppressed using -raw_html. Previously native_divs was enabled but could not be suppressed, because it was not in the list of available extensions for commonmark-based formats.

Text.Pandoc.Parsing:

• Partition module into (internal) submodules (#7962, Albert Krewinkel).

• Unify grid table parsing (#7971, Albert Krewinkel). Grid table parsing in Markdown and RST are updated use the same functions. Functions are generalized to meet requirements for both formats. This change also lays the ground for further generalizations in table parsers, including support for advanced table features. [API change] in Text.Pandoc.Parsing:

  • Parse results of functions tableWith' and gridTableWith' are now a mf TableComponents instead of a quadruple of alignments, column widths, header rows and body rows.
  • Additional exports from Text.Pandoc.Parsing: tableWith', TableComponents, TableNormalization, toTableComponents, toTableComponents'.

Text.Pandoc.XML.Light: add versions of the parsers that allow specifying a custom entity map. Exports new functions: parseXMLElementWithEntities, parseXMLContentsWithEntities [API change].

Text.Pandoc.Writers.GridTable: improve module documentation (Albert Krewinkel).

Text.Pandoc.Parsing.GridTable: simplify column handling code (Albert Krewinkel).

Text.Pandoc.MIME: Add mime type for mkv extension (#7181).

Text.Pandoc.Asciify: Fix regression with ascii_identifiers and Turkish undotted i (#8003).

Relax upper bound for hslua, allow hslua-2.2. (#7929) Lua 5.4 is used by default after this is merged. Packagers may still include Lua 5.3 instead by building pandoc with --constraint='hslua <2.2'. Differences between 5.3 and 5.4 should not generally affect pandoc Lua filters. See list of incompatible changes here: https://www.lua.org/manual/5.4/manual.html#8.1

JATS template: allow multiple licenses (Albert Krewinkel).

LaTeX template:

• Skip \babelprovide if babel-lang is empty (#7945) to avoid an error.
• Move hyperref near end of preamble (#5811). It now comes after header-includes and right before title, author, date, abstract. Note: Users who presuppose hyperref in their header-includes will now have to add \usepackage{hyperref} to their header-includes to make it available there. (The redundant \usepackage will do no harm in this case.)
• Remove special redefinition of \sout. This used to be necessary to avoid problems with hyperref, when headings contain strikeout text, but it does not seem to be necessary any more (tested).

Tests: improve location reporting of failing tests (Albert Krewinkel).

Add tests for idempotency of makeSections (#7950).

Add armhf support in linux build script (#7944, Tony).

Use latest pandoc-types, so that toJSONFilter will work with Meta and MetaValue.

Use latest doclayout, texmath, commonmark, citeproc.

INSTALL.md: add reference to install via winget (#7951, Guriy Samarin).

MANUAL.txt:

• Document way to get list in block quote in slide shows (#7916).
• Minor changes to security section.
• Note that smart works for html input.
• Fix typos (#7934, Dimitris Apostolou).
• Fix documentation for citations and org mode.
• Remove claim that --self-contained does not work with --mathjax (#682).

doc/org.md: remove obsolete citations section. This mostly described citation formats we no longer support.

doc/lua-filters.md: Fix typo (#7981, Mario Lang).

Makefile: Use cabal for default build

Add lua53 cabal flag. It is false by default. If set to true, compile with hslua 2.1 and Lua 5.3, otherwise hslua 2.2 and Lua 5.4.

Fix regression in 2.17.1 which caused problems finding default files in the default user data directory. (Reverts the item “logic bug in fullDefaultsPath”, which was misguided.)

Sample custom writer: use single quotes for strings (#7487, Albert Krewinkel).

Support pagedjs-cli as pdf engine (#7838, Albert Krewinkel). PagedJS is a polyfill and supports the Paged Media standards by the W3C. https://www.pagedjs.org/

CommonMark reader: fix source position after YAML metadata (#7863).

LaTeX reader:

• Remove retokenizing in rawLaTeXParser.

• Ensure that \raggedright doesn’t gobble an argument (#7757).

• Improve descItem. For some reason we were skipping arbitrary blocks before \item. This is now changed to “skip whitespace and comments.”

• Improve handling of \newif. Adding a pair of braces around the second argument of \def prevents LaTeX from an emergency stop on input like the following (#6096).

  \newif\ifepub
  \epubtrue
  \ifepub
  hi
  \fi

Docx reader: Parse both Zotero citation and bibliography as FieldInfo (#7840).

LaTeX writer:

• Allow arbitrary frameoptions to be passed to a beamer frame, using the frameoptions attribute (#7869).
• Add s and squeeze to recognized beamer frameoptions (#7869).

Markdown writer: handle explicit column widths with pipe tables (#7847). If a table has explicit column width information and the content extends beyond the --columns width, we need to adjust the widths of the pipe separators to encode this width information.

Docx writer: Separate tables even with RawBlocks between (#7724, Michael Hoffmann). Adjacent docx tables need to be separated by an empty paragraph. If there’s a RawBlock between tables which renders to nothing, be sure to still insert the empty paragraph so that they will not collapse together.

Man writer: use custom font V for inline code (#7506). The V font is defined conditionally, so that it renders like CB in output formats that support that, and like B in those that don’t (e.g. the terminal). Aliases also defined for VI, VB, VBI.

Asciidoc writer: Support checklists in asciidoctor writer (#7832, Nikolai Korobeinikov, ricnorr). The checklist syntax (similar to task_list in markdown) seems to be an asciidoctor-only addition.

HTML writer:

• Avoid duplicate “style” attributes on table cells (#7871).
• Don’t break lines inside code elements. With the new (default) line wrapping of HTML, in conjunction with the default CSS which includes code { whitespace: pre-wrap; }, spurious line breaks could be introduced into inline code (#7858).

Custom writer: preserve order of element attributes (#7489, Albert Krewinkel). Attribute key-value pairs are marshaled as AttributeList, i.e., as a userdata type that behaves both like a list and a map. This allows to preserve the order of key-value pairs.

Switch to hslua-2.1 (Albert Krewinkel). This allows for some code simplification and improves stability.

Don’t read files outside of user data directory (Even Brenden). If a file path does not exist relative to the working directory, and it does exist relative to the user data directory, but outside of of the user data directory, do not read it. This applies to readDataFile and readMetadataFile in PandocMonad and, by extension, any module that uses these by passing them relative paths.

Text.Pandoc.Class.makeCanonical: Correctly handle consecutive “..”s at the beginning of a path (Even Brenden). Prior to this commit, ../../file would evaluate to file, when it should be unchanged.

Search for metadata files in $DATADIR/metadata (#7851, Even Brenden). If files specified with --metadata-file are not found in the working directory, look in $DATADIR/metadata (#5876).

Text.Pandoc.Class: export readMetadataFile [API change] (#5876).

Text.Pandoc.Error: export new PandocCouldNotFindMetadataFileError constructor for PandocError [API change] (#5876).

Avoid putting a frame around speaker notes in beamer (#7857). If speaker notes (a Div with class ‘notes’) occur right after a section heading, but above slide level, the resulting \note{..} caommand should not be wrapped in a frame, as that will cause a spurious blank slide.

CSS in HTML template: adjust #TOC and h1 on mobile (#7835, Mauro Bieg).

Text.Pandoc.Readers.LaTeX.Parsing: don’t export totoks. Make the first param of tokenize a SourcePos instead of SourceName, and use it instead of totoks.

Text.Pandoc.Shared: Modify stringify so it ignores [Citation] inside Cite (#7855). Otherwise we’ll sometimes get two copies of things, one from the citationPrefix or citationSuffix and another from the embedded fallback text. When there is no fallback text, we’ll get no content. However, it really isn’t an alternative to just rely on the result of running query on the embedded Citations; this will result in a jumble of text rather than anything structured.

Omit --enable-doc in the cabal haddock invocation in tools/build-and-upload-api-docs.sh.

Text.Pandoc.App.Opt: fix logic bug in fullDefaultsPath. Previously we would (also) search the default user data directory for a defaults file, even if a different user data directory was specified using --data-dir. This was a mistake; if --data-dir is used, the default user data directory should not be searched.

Text.Pandoc.Shared: defaultUserDataDir behavior change (#7842). If the XDG data directory is not defined (e.g. because it’s not supported in the OS or HOME isn’t defined), we return the empty string instead of raising an exception.

Update command tests to distinguish stderr and test exit status.

MANUAL: add that speaker notes can be used with beamer (#7856).

Update build-and-upload-api-docs.sh.

Document --trace option. Document no-check-certificate in defaults files. Document ‘sandbox’ option for defaults files. (#7873).

Fix pattern syntax in sample readability custom reader.

doc/custom-readers.lua: add example for “readable HTML.”

Fix message in man page about where code can be found.

manfilter.lua: remove extra indent in table cells with code blocks.

Fix lua-filters documentation for table column widths (#7864).

epub.doc: Update links to KindleGen (#7846, Benson Muite, Mauro Bieg). KindleGen has been deprecated and we need to link to archived versions.

Use tables in defaults files documentation, so each default option is paired with the corresponding command-line option (Carsten Allefeld).

Use skylighting 0.12.2.

Add pandoc-lua-marshal to Nix shell (#7849, Even Brenden).

Require pandoc-lua-marshal 0.1.3.1 (#7831, Albert Krewinkel). Fixes a problem with List.includes and List.find that caused a Lua stackoverflow and subsequent program crash.

HTML template: load header-includes before math (#7833, Kolen Cheung). MathJax expect the config comes before loading the MathJax script. This change of order allows one to config MathJax via header-includes, which loads before the MathJax script. Cf. #2750.

When reading defaults file, stop at a line .... This line signals the end of a YAML document. This restores the behavior we got with HsYaml. yaml complains about content past this line. See https://github.com/jgm/pandoc/issues/4627#issuecomment-1012438765

Text.Pandoc.Citeproc: allow notes-after-punctuation to work with numerical styles that use superscripts (e.g. american-medical-association.csl), as well as with note styles. The default setting of notes-after-punctuation is true for note styles and false otherwise. This restores a behavior of pandoc-citeproc that wasn’t properly carried over to Citeproc (#7826, cf. jgm/pandoc-citeproc#384).

Use commonmark-pandoc 0.2.1.2 (#7769).

Add FAQ on images in ipynb containers (#7749, Kolen Cheung).

Support markua as an output format (#1871, Tim Wisotzki and Saumel Lemmenmeier). Markua is a markdown variant used by Leanpub.

Add text wrapping for HTML output (#7764). Previously the HTML writer was exceptional in not being sensitive to the --wrap option. With this change --wrap now works for HTML. The default (as with other formats) is automatic wrapping. Note that the contents of script, textarea, and pre tags are always laid out with the flush combinator, so that unwanted spaces won’t be introduced if these occur in an indented context in a template.

Don’t read sources until in/out format are verified (#7797).

Issue error with --list-extensions for invalid formats (#7797).

Make --citeproc recognize .yml as well as .yaml extensions as YAML bibliography files (#7707, Jörn Krenzer).

Use latest version of KaTeX with --katex.

Fix parsing of footnotes in --metadata-file (#7813). Previously non-inline footnotes were not being parsed.

ODT reader:

• Parse list-header as a list item (Tuong Nguyen Manh).

Commonmark reader:

• Put sourcepos attribute on header, not enclosing div with -f commonmark+sourcepos (#7769).

Markdown reader:

• Don’t allow ^ at beginning of link or image label (#7723). This is reserved for footnotes. Fixes regression from 0a93acf.
• Fix parsing of “bare locators” after author-in-text citations. Previously @item [p. 12; @item2] was incorrectly parsed as three citations rather than two. This is now fixed by ensuring that prefix doesn’t gobble any semicolons.
• Revert changes to inlinesInBalancedBrackets (commit fa83246), which caused regressions.
• Improve detection of pipe table line widths (#7713). Fixed calculation of maximum column widths in pipe tables. It is now based on the length of the markdown line, rather than a “stringified” version of the parsed line. This should be more predictable for users. In addition, we take into account double-wide characters such as emojis.

Custom (Lua) readers:

• First argument is now a list of sources instead of the concatenated text (Albert Krewinkel). The list structure can easily be converted to a string by applying tostring, but it is also possible to access the elements (each with a text and name). A small example is added to the custom reader documentation, showcasing its use in a reader that creates a syntax-highlighted code block for each source code file passed as input. Existing readers will still work through a fallback mechanism, issuing a deprecation notice.

Org reader:

• Parse official org-cite citations (#7329). We also support the older org-ref style as a fallback. We no longer support the “markdown style” or “Berkeley style” citations.
• Support alphabetical (fancy) lists (Lucas Viana). When the fancy_lists extension is enabled, alphabetical list markers are allowed, mimicking the behaviour of Org Mode when org-list-allow-alphabetical is enabled.
• Support counter cookies in lists (Lucas Viana). Such cookies are used to override the item counter in ordered lists. In org it is possible to set the counter at any list item, but since Pandoc AST does not support this, we restrict the usage to setting an offset for the entire ordered list, by using the cookie in the first list item.
• Allow trailing spaces after key/value pairs in directives (Albert Krewinkel). Ensures that spaces at the end of attribute directives like #+ATTR_HTML: :width 100% (note the trailing spaces) are accepted.

LaTeX reader:

• Omit visible content for \label{...}. Previously we included the text of the label in square brackets, but this is undesirable in many cases. See discussion in https://github.com/jgm/pandoc/issues/813#issuecomment-978232426.
• Improve references (#813). Resolve references to theorem environments. Remove the Span caused by “label” in figure, table, and theorem environments; this had an id that duplicated the environments’ id.
• Fix semantics of \ref. We were including the ams environment type in addition to the number. This is proper behavior for \cref but not for \ref. To support \cref we need to store the environment label separately.
• Add babel mappings for Guajati (gu) and Oriya (or) (#7815).
• Fix typo panjabi -> punjabi in babel mappings (#7814).

HTML reader:

• Parse attributes on links and images (#6970).

Docx reader:

• Handle multiple pic elements inside a drawing (#7786).
• Change elemToParPart to return [ParPart] instead of ParPart. Also remove NullParPartconstructor, as it is no longer needed. This will allow us to handle elements that contain multiple ParParts, e.g. w:drawing elements with multiple pic:pic.

DocBook reader:

• Collapse internal spaces in literal and other similar tags (#7821), as the standard docbook toolchain does.
• Be sensitive to spacing=“compact” in lists (#7799). When spacing="compact" is set, Para elements are turned into Plain, so we get a “tight” list.

Markdown writer:

• Add new exported function writeMarkua from Text.Pandoc.Writers.Markdown [API change] (#1871, Tim Wisotzki and Saumel Lemmenmeier).
• Fix indentation issue in footnotes (#7801).
• Avoid extra space before citation suffix if it already starts with a space.
• Ensure semicolon between the locator and the next citation when an author-in-text citation has a locator and following citations.
• Improve escaping for # (#7726).

Custom (Lua) writers:

• Allow variables to be set via second return value of Doc (#6731, Albert Krewinkel). New templates variables can be added by giving variable-value pairs as a second return value of the global function Doc. Example:

  function Doc (body, meta, vars)
    vars.date = vars.date or os.date '%B %e, %Y'
    return body, vars
  end

• Provide global PANDOC_WRITER_OPTIONS (#6731, Albert Krewinkel).

• Assign default Pandoc object to global PANDOC_DOCUMENT (Albert Krewinkel). The default Pandoc object is now non-strict, i.e., only the parts of the document that are accessed will be marshaled to Lua. A special type is no longer necessary. This change also makes it possible to use the global variable with library functions such as pandoc.utils.references, or to inspect the document contents with walk().

LaTeX writer:

• Fix typo panjabi -> punjabi in babel mappings (#7814).

MediaWiki writer:

• Remove redundant display text for wiki links (Jesse Hathaway).

Docx writer:

• Handle bullets correctly in lists by not reusing numIds (#7689, Michael Hoffmann). This fixes a bug in which a Div in a list item would receive bullets on its contained paragraphs.

Org writer:

• Fix list items starting with a code block or other non-paragraph content (#7810).
• Avoid blank lines after tight sublists (#7810).
• Fix extra blank line inserted after empty list item (#7810).
• Don’t add blank line before lists (#7810).
• Support starting number cookies (Lucas Viana). This is necessary for lists that start at a number other than 1.
• Support the new org-cite syntax (#7329).

Haddock writer:

• Avoid blank lines after tight sublists (#7810).

Ipynb writer:

• Ensure deterministic order of keys.
• Handle cell output with raw block of markdown (#7563, Kolen Cheung). Write RawBlock of markdown in code-cell output. This is designed to fit the behavior of #7561, which makes the ipynb reader parse code-cell output with mime “text/markdown” to a RawBlock of markdown. This commit makes the ipynb writer writes this RawBlock of markdown back inside a code-cell output with the same mime, preserving this information in round-trip.
• In choosing between multiple output options, always favor those marked with the output format over images (Kolen Cheung). Previously, both fmt == f case and Image have a rank of 1.

Ipynb reader & writer: properly handle cell “id” (#7728). This is passed through if it exists (in Nb4); otherwise the writer will add a random one so that all cells have an “id”.

Ms writer:

• Properly encode strings for PDF contents (#7731).

JATS writer:

• Keep quotes in element-citations (Albert Krewinkel). Fixed a bug that lead to quote characters being lost in element-citations.

RTF writer:

• Properly handle images in data URIs (#7771).

Commonmark writer:

• Allow ‘)’ delimiters on ordered lists.

RST writer:

• Avoid extra blank line after empty list item (#7810).

HTML writer:

• Make line breaks more consistent. With --wrap=none, we now output line breaks between block-level elements. Previously they were omitted entirely, so the whole document was on one line, unless there were literal line breaks in pre sections. This makes the HTML writer’s behavior more consistent with that of other writers. Also, regardless of wrap settings, put newline after <dd> and after block-level elements in the footnotes section. And add a line break between an img tag and the associated figcaption.
• reveal.js: Make sure images with r-stretch are not in p tags. They must be direct children of the section. There was previously code to make this work with the older class name stretch, but the name has changed in reveal.js.
• reveal.js: don’t add r-fit-text class to section. It must go on the header only.

AsciiDoc writer:

• Improve detection of intraword emphasis (#7803).

OpenDocument writer:

• Fix vertical alignment bug with display math (#7777). Previously some displayed formulas would be floated above a preceding text line. This is fixed by setting vertical-rel to text rather than paragraph-content.

JATS template (Albert Krewinkel):

• Fix position of contrib affiliations in authoring set. Any <aff> element must come before any <email> element.
• Fix affiliation tagging in articleauthoring output. Affiliations were xlinked even in the articleauthoring tag set, but <aff> are not allowed as children of contrib-group elements in that tag set. Each affiliation must be listed directly in the contrib element.
• Add support for article subtitles.

EPUB template:

• Include abstract in default template.
• Ensure that the essential styles needed by pandoc (styles.html partial) are included in the templates. This is important for correct formatting of CSL bibliographies. Note that much of the styling in styles.html will be ignored for EPUB, because of the conditional on document-css. Setting the document-css variable will cause it not to be ignored.

HTML template: Add abstract (#7588, Jannik Buhr, John MacFarlane). By default, a localized title (the word “Abstract”) will be used, unless the variable abstract-title is set.

ConTeXt template: Make title appear in PDF title bar. This is recommended for accessibility reasons. Note: doesn’t work with macOS Preview.app.

reference.pptx: change to use 16:9 aspect ratio, Powerpoint’s default.

Text.Pandoc.Writers:

• Do not export writeCustom [API change]. This ensures that all writers exported in T.P.Writers are parameterized and work with any PandocMonad type. This is consistent with T.P.Readers, as readCustom is not exported from that module either.

Text.Pandoc.Writers.Shared:

• endsWithPlain now returns True if the list ends with a list that ends with a Plain, and so on recursively (#7810).

Text.Pandoc.Class.IO:

• writeMedia: unescape percent-encoding in creating file path. This addresses a problem with spaces in image filenames when creating PDFs (#7819); it also affects --extract-media.

New internal module Text.Pandoc.Writers.Blaze, exporting layoutMarkup. This converts a blaze Html structure into a doclayout Doc Text.

Text.Pandoc.Extensions:

• parseFormatSpec: cleaner error message for invalid extensions.

Text.Pandoc.MediaBag:

• Fix bug in extractMedia, which caused the test for .. in paths to fail, with the result that images with .. in the path name could be extracted outside of the directory specified by extractMedia. It also led a check for media in resource paths to fail in the docx reader.

Text.Pandoc.Citeproc:

• Avoid adding comma before an author-in-text citation in a note if it begins with a title (no author) (#7761).
• Text.Pandoc.Citeproc.Locator now exports toLocatorMap, LocatorInfo, and LocatorMap. The type of parseLocator has changed, so it now takes a LocatorMap rather than a Locale as parameter, and returns a LocatorInfo instead of a tuple.
• Fix citation locator detection for German. toLocatorMap now stores keys case-folded. We want to do a case-insensitive comparison when parsing locators, so that e.g. both Chap. and chap. work. Previously we case-folded terms when doing the lookup, but they weren’t case-folded in the map itself, which led to locator-detection breaking for German (where the terms have uppercase letters).

Lua (Albert Krewinkel):

• Allow single elements as singleton MetaBlocks/MetaInlines. Single elements should always be treated as singleton lists in the Lua subsystem.
• Add pandoc.template module. The module provides a compile function to use strings as templates.
• Add pandoc.WriterOptions constructor.
• Add function pandoc.write.
• Provide global PANDOC_WRITER_OPTIONS (#5221).
• The function Text.Pandoc.Filter.applyFilters now takes a filter environment of type Environment, instead of a ReaderOptions value [API change].
• The Environment type is exported from Text.Pandoc.Filter and allows to combine ReaderOptions and WriterOptions in a single value [API change].
• Global, exported from Text.Pandoc.Lua, has a new type constructor PANDOC_WRITER_OPTIONS [API change].
• Add constructors pandoc.Blocks and pandoc.Inlines. The functions convert their argument into a list of Block and Inline values, respectively. When applied to a string, they split the string into Str elements separated by Space or SoftBreak (#7712).
• Support topdown traversals The traversal order of filters can now be selected by setting the key traverse of the filter to either 'topdown' or 'typewise'; the default remains 'typewise'. Topdown traversals can be cut short by returning false as a second value from the filter function. No child-element of the returned element is processed in that case.
• Marshal ReaderOptions field extensions, track_changes via JSON. Extensions are now available as a list of strings; the track-changes settings are given as the kebab-case representation used in JSON.
• Allow binary (byte string) readers to be used with pandoc.read.
• Use global state when parsing documents in pandoc.read. The function pandoc.read is updated to use the same state that was used while parsing the main input files. This ensures that log messages are preserved and that images embedded in the input are added to the mediabag.
• Cleanup stack in peekReadOptionsTable. A ReaderOptions element was left on top of the stack when the peekReadOptionsTable function was invoked.
• walk methods are added to Pandoc, Block, Inline, Blocks, Inlines values; the methods are similar to pandoc.utils.walk_block and pandoc.utils.walk_inline, but apply to filter to the element’s contents.
• Functions of name Doc are no longer accepted as alternatives for Pandoc filter functions. This functionality was undocumented.
• Improve handling of empty caption, body by from_simple_table #7776). Create truly empty table caption and body when these are empty in the simple table.
• Change representation of TableHead, TableFoot, and Row values (#7718). The objects now also follow the principle that element attributes are accessible through the .attr field. Rows in TableHead and TableFoot are available via the .rows field. Row objects have a .cells field, containing the list of table cells.
• Simplify code of pandoc.utils.stringify. Minor behavior change: plain strings nested in tables are now included in the result string.
• Simplify and deprecate function pandoc.utils.equals. The function is no longer required for element comparisons; it is now an alias for the == operator.
• Add function pandoc.utils.references (#7752).
• Add new library function pandoc.utils.type. The function behaves like the default type function from Lua’s standard library, but is aware of pandoc userdata types. A typical use-case would be to determine the type of a metadata value.
• Fix return types of blocks_to_inlines, make_sections. Ensures the returned lists have the correct type (Inlines and Blocks, respectively).
• Use more natural representation for Reference values Omit false boolean values, push integers as numbers.

Lua: use package pandoc-lua-marshal (#7719, Albert Krewinkel). The marshaling functions for pandoc’s AST are extracted into a separate package. The package comes with a number of changes:

• Pandoc’s List module was rewritten in C, and error messages were improved.
• Lists of Block and Inline elements are marshaled using the new list types Blocks and Inlines, respectively. These types currently behave identical to the generic List type, but give better error messages. This also opens up the possibility of adding element-specific methods to these lists in the future.
• Elements of type MetaValue are no longer pushed as values which have .t and .tag properties. This was already true for MetaString and MetaBool values, which are still marshaled as Lua strings and booleans, respectively. Affected values:
  • MetaBlocks values are marshaled as a Blocks list;
  • MetaInlines values are marshaled as a Inlines list;
  • MetaList values are marshaled as a generic pandoc Lists.
  • MetaMap values are marshaled as plain tables and no longer given any metatable.
• Cell values are now marshaled as userdata objects; a constructor function for table cells is provided as pandoc.Cell.
• The test suite for marshaled objects and their constructors has been extended and improved.
• A bug in Citation objects, where setting a citation’s suffix modified it’s prefix, has been fixed.
• Inlines, Blocks, and List objects now have an __eq metamethod, testing equality by comparing two lists element-wise.

Powerpoint tests: shorten lines by grouping tests (Albert Krewinkel). This makes the test output more pleasant to read in narrow terminal windows.

make check: check for unreleased dependencies.

Add tools/build-and-upload-api-docs.sh.

Update cabal description.

MANUAL.txt: Add section on EPUB styling.

MANUAL.txt: clarify “standard Markdown” as “original Markdown” (#7802, Martin Fischer).

doc/custom-writers.md: use filter to include source of example.

Add an example to doc/custom-readers.md.

Fix typo in custom-readers.md (#7722, Mauro Bieg).

doc/jats.md: add link to JATS documentation (Martin Fischer).

doc/lua-filters.md: many improvements (Albert Krewinkel, John MacFarlane).

Use commonmark-extensions 0.2.3. This allows a bare-word class attribute on fenced divs.

Use ipynb 0.2.

Use citeproc 0.6.0.1.

Use texmath 0.12.4.

Use doctemplates 0.10.0.1.

Add interface for custom readers written in Lua (#7669). Users can now do -f myreader.lua and pandoc will treat the script myreader.lua as a custom reader, which parses an input string to a pandoc AST, using the pandoc module defined for Lua filters. A sample custom reader can be found in data/creole.lua. Also see documentation in doc/custom-readers.md.

New module Text.Pandoc.Readers.Custom, exporting readCustom [API change].

Allow plain to be used in raw attribute syntax.

Accept empty --metadata-file (#7675). This was a regression from 2.15 behavior.

Markdown reader: Improve inlinesInBalancedBrackets. This is just a small improvement in terms of performance, but it’s simpler and more direct code. Also, we avoid parsing interparagraph spaces in balanced brackets, as the original did.

BibTeX reader: Properly handle commented lines in BibTeX/BibLaTeX (#7668).

RST reader: handle class attribute for for custom roles (#7699, willj-dev). Previously the class attribute was ignored, and the name of the role used as the class.

DocBook reader:

• Add <titleabbr> support (Rowan Rodrik van der Molen).
• Support for <indexterm> (#7607, Rowan Rodrik van der Molen).

LaTeX reader:

• Add rudimentary support for \autoref (#7693).
• Add ‘uri’ class when parsing \url, for consistency with treatment of autolinks in other formats (#7672).

JATS reader: Capture alt-text in figures (#7703, Aner Lucero).

MediaWiki writer: use HTML spans for anchors when header has id (#7697). We need to generate a span when the header’s ID doesn’t match the one MediaWiki would generate automatically. Note that MediaWiki’s generation scheme is different from pandoc’s (it uses uppercase letters, and _ instead of -, for example). This means that in going from markdown to mediawiki, we’ll now get spans before almost every heading, unless explicit identifiers are used that correspond to the ones MediaWiki auto-generates. This is uglier output but it’s necessary for internal links to work properly.

Markdown writer: don’t create autolinks when this loses information (#7692). Previously we sometimes lost attributes when rendering links as autolinks.

Text.Pandoc.Readers.Metadata: allow multiple YAML documents when parsing YAML for yamlBsToRefs. Some people use --- as the end delimiter in YAML bibliography files, which causes the yaml library to emit an error unless we explicitly allow multiple YAML documents (and just consider the first).

JATS writer:

• Ensure figures are wrapped with <p> in list items (Albert Krewinkel). This prevents the generation of invalid output.
• Add URL to element citation entries (Albert Krewinkel). The URL of a reference, if present, is added in tag <uri> to element-citation entries.

HTML writer: Don’t create invalid data- attribute for empty attribute key (#7546).

LaTeX writer:

• Babel mappings: use ancientgreek for grc.
• With -t latex-smart, don’t generate \ldots from ellipsis (#7674). Instead just use unicode ellipsis.

JATS template: fix equal-contrib attribute (Albert Krewinkel). The standard requires the value to be either yes or no, but is was set to true for authors who contributed equally.

reveal.js template: Add disableLayout variable (Christophe Dervieux).

Text.Pandoc.Error: sort errors in handleError by exit code (Albert Krewinkel).

Text.Pandoc.Writers.Shared: Improve toLegacyTable (#7683, Christian Despres).

Lua subsystem:

• Include lpeg module (#7649, Albert Krewinkel). Compiles the lpeg library (Parsing Expression Grammars For Lua) into the program. Package maintainers may choose to rely on package dependencies to make lpeg available, in which case they can compile the with the constraint lpeg +rely-on-shared-lpeg-library. lpeg and re are always made available in global variables, without the need for a require.

• Set lpeg and re as globals; allow shared lib access via require. The lpeg and re modules are loaded into globals of the respective name, but they are not necessarily registered as loaded packages. This ensures that

  • the built-in library versions are preferred when setting the globals,
  • a shared library is used if pandoc has been compiled without lpeg, and
  • the require mechanism can be used to load the shared library if available, falling back to the internal version if possible and necessary.

• Fix argument order in constructor pandoc.Cite (Albert Krewinkel). This restores the old behavior; argument order had been switched accidentally in pandoc 2.15.

• Add Pushable instance for ReaderOptions (Albert Krewinkel).

• Allow to pass custom reader options to pandoc.read as an optional third argument (#7656, Albert Krewinkel). The object can either be a table or a ReaderOptions value like PANDOC_READER_OPTIONS. Creating new ReaderOptions objects is possible through the new constructor pandoc.ReaderOptions.

• Display Pandoc values using their native Haskell representation (Albert Krewinkel).

• Require latest hslua (2.0.1) (#7661, #7657, Albert Krewinkel). This fixes issues with

  • misleading error messages when a required function parameter is omitted;
  • absent properties still being listed in the output of pairs; and
  • alias accessing leading to errors instead of returning nil, e.g. with (pandoc.Str '').identifier.

• Add missing space in “package not found” message (#7658, Albert Krewinkel).

Update build files (#7696, Fabián Heredia Montiel). Drop old windows 32-bit constraints. Update cabal tested-with field to correspond to ci.yml matrix

Remove unneeded package dependencies from benchmark target.

Require ghc >= 8.6, base >= 4.12. This allows us to get rid of the old custom prelude and some crufty cpp. But the primary reason for this is that conduit has bumped its base lower bound to 4.12, making it impossible for us to support lower base versions.

Require Cabal 2.4. Use wildcards to ensure that all pptx tests are included (#7677).

Update bash_completion.tpl (S.P.H.).

Add data/creole.lua as sample custom reader.

Add doc/custom-readers.md and doc/custom-writers.md.

doc/lua-filters.md: add section on global modules, including lpeg (Albert Krewinkel).

MANUAL.txt: update table of exit codes and corresponding errors (Albert Krewinkel).

Use latest texmath.

Docx reader: don’t let first line indents trigger block quotes (#7655). This fixes a regression introduced in pandoc 2.15.

Docx writer: use getTimestamp for modification times in reference.docx (#7654). This ensures that when SOURCE_DATE_EPOCH is set, the modification times of files taken from the reference.docx will be set deterministically, allowing for reproducible builds.

Lua subsystem (Albert Krewinkel):

• Load module pandoc.path on startup (#7524). Previously the module always had to be loaded via require 'pandoc.path'.
• Fix typo in SoftBreak constructor.
• Re-add content property to Strikeout elements. Fixes a regression introduced in 2.15.
• Be more forgiving when retrieving the Image caption property. Fixes a regression introduced in 2.15.
• Display Attr values using their native Haskell representation.
• Allow omitting the 2nd parameter in pandoc.Code constructor. Fixes a regression introduced in 2.15 which required users to always specify an Attr value when constructing a Code element.
• Allow to compare, show Citation values. Comparisons of Citation values are performed in Haskell; values are equal if they represent the same Haskell value. Converting a Citation value to a string now yields its native Haskell string representation.
• Restore List behavior of MetaList (#7650). Fixes a regression introduced in 2.16 which had MetaList elements lose the pandoc.List properties.
• Restore content property on Header elements.
• Ensure Block elements have all expected properties.
• Ensure Inline elements have all expected properties.

Allow tasty-bench 0.3.x.

Switch back from HsYAML to yaml for parsing YAML metadata (#6084). HsYAML is around 20 times slower in parsing large YAML bibliographies. In addition, HsYAML is not being actively maintained. This sets us back in our attempts to free ourselves from C dependencies (#4535). But I don’t see a good alternative until a faster pure Haskell parser is available. Notes:

• We’ve removed the FromYAML instances for all types that had them, since this is a HsYAML-specific typeclass [API change]. (The yaml package just uses From/ToJSON instead of having a dedicated From/ToYAML class.)
• Unlike HsYAML (in the configuration we were using), yaml parses ‘Y’, ‘N’, ‘Yes’, ‘No’, ‘On’, ‘Off’ as boolean values. Users may need to quote these when they are meant to be interpreted as strings. Similarly, ‘null’ is parsed as a YAML null value (and will be treated as an empty string by pandoc rather than the string ‘null’). Quoting it will force it to be interpreted as a string.
• Some tests had to be adjusted accordingly.
• Pandoc now behaves in a more useful way when the YAML metadata contains escaping errors: instead of just failng silently and falling back to some other interpretation of the section, it raises a YAML parsing error.

Markdown writer: Ensure that special values are quoted in YAML metadata. These include “Y”, “yes”, “on”, and “off”, which are now (with yaml library) considered boolean values, as well as “null”.

Change JSON encodings of some types.

• For LineEnding use lowercase constructors, e.g. crlf, native.
• For HTMLSlideVariant use lowercase constructors.
• For ReaderOptions use e.g. default-image-extension instead of readerDefaultImageExtension for field names.
• For Extension, use e.g. tex_math_dollars instead of Ext_tex_math_dollars as constructor.
• For Extensions, use an array of Extensions, instead of an object wrapping the tag Extensions and an integer. (The integer representation is not supposed to be part of the public API.)
• For Opt, use field names like tab-stop instead of optTabStop.

Docx writer:

• Add IDs to native_numbering test (Tristan Stenner).
• Move “:” out of the caption bookmark (Tristan Stenner). This is needed so that native references to the figure are included as “As seen in Figure X, it is…” instead of “As seen in [Figure: X, it is…”

Lua (Albert Krewinkel, except as noted):

• Use hslua module abstraction where possible.

• Fix placement of tests for Block elements in pandoc module tests

• Increase strictness when getting attribute keys

• Re-add t and tag property to Attr values. Removal of these properties from Attr values was a regression.

• Fix pandoc.utils.stringify regression. The pandoc.utils.stringify function returned empty strings when called with a string argument.

• Fix a copy/paste bug in Lua marshalling code (John MacFarlane, #7639). This caused links to be changed to figures when Lua filters changed link properties.

• Re-add content property to Link elements (#7647). This was a regression introduced in version 2.15.

• Generate constants in module pandoc programmatically.

• Marshal SimpleTable, ListAttributes, Citation, and Block values as userdata objects. Properties of Block values are marshalled lazily, which generally improves performance considerably. Script users may also notice the following differences:

  • Block element properties can no longer be accessed by numerical indexing of the .c field. The .c property now serves as an alias for .content, so some filter that used this undocumented method for property access may continue to work, while others will need to be updated and use proper property names.
  • The marshalled Block elements now have a show method, and a __tostring metamethod. Both return the Haskell string representation of the element.
  • Block values now have the Lua type userdata instead of table.

Add a short guide to pandoc’s sources (Albert Krewinkel).

Fix epub files in epub reader tests, so that they are valid according to epubcheck (#7586).

Allow time 1.13.

Require latest skylighting (0.12.1).

Fix build on GHC 9.2 (Joseph C. Sible).

Fix trypandoc so it builds with aeson > 2.

Add --sandbox option (#5045).

• Add sandbox feature. When this option is used, readers and writers only have access to input files (and other files specified directly on command line). This restriction is enforced in the type system.
• Filters, PDF production, custom writers are unaffected. This feature only insulates the actual readers and writers, not the pipeline around them in Text.Pandoc.App.
• Note that when --sandboxed is specified, readers won’t have access to the resource path, nor will anything have access to the user data directory.

--self-contained: Fix bug that caused everything to be made a data URI (#7635, #7367). We only need to use data URIs in certain cases, but due to a bug they were being used always.

Pandoc will now fall back to latin1 encoding for inputs that can’t be read as UTF-8. This is what it did previously for content fetched from the web and not marked as to content type. It makes sense to do the same for local files. In this case a NotUTF8Encoded warning will be issued, indicating that pandoc is interpreting the input as latin1.

Markdown reader:

• Don’t parse links or bracketed spans as citations (#7632). Previously pandoc would parse [link to (@a)](url) as a citation; similarly [(@a)]{#ident}. This is undesirable. One should be able to use example references in citations, and even if @a is not defined as an example reference, [@a](url) should be a link containing an author-in-text citation rather than a normal citation followed by literal (url).
• Fix interaction of --strip-comments and list parsing (#7521). Use of --strip-comments was causing tight lists to be rendered as loose (as if the comment were a blank line).
• Fix parsing bug for math in bracketed spans and links (#7623). This affects math with unbalanced brackets (e.g. $(0,1]$) inside links, images, bracketed spans.
• Fix code blocks using --preserve-tabs (#7573). Previously they did not behave as the equivalent input with spaces would.

DocBook reader:

• Honor linenumbering attribute (Samuel Tardieu). The attribute DocBook linenumbering="numbered" on code blocks maps to the numberLines class internally.

LaTeX reader:

• Implement siunitx v3 commands (#7614). We support \unit, \qty, \qtyrange, and \qtylist as synonyms of \si, \SI, \SIrange, and \SIlist.
• Properly handle \^ followed by group closing (#7615).
• Recognize that \vadjust sometimes takes “pre” (#7531).
• Ignore (and gobble parameters of) CSLReferences environment (#7531). Otherwise we get the parameters as numbers in the output.
• Restrict \endinput to current file (Simun Schuster).

RST reader: handle escaped colons in reference definitions (#7568).

HTML reader:

• Handle empty tbody element in table (#7589).

Ipynb reader (Kolen Cheung):

• Get cell output mime from raw_mimetype in addition to format. (format is what the spec calls for, but raw_mimetype is often used in practice; see jupyter/nbformat#229).
• Add more formats that can be handled as “raw” cells.
• Fix mime type for rst.
• Support text/markdown, which is now a supported mime type for raw output (#7561).

RTF reader:

• Support \binN for binary image data.
• If doc begins with { … } only parse its contents. Some documents seem to have non-RTF (e.g. XML) material after the {\rtf1 ... } group.
• Ignore \pgdsc group. Otherwise we get style names treated as test.
• Better handling of \* and bookmarks. We now ensure that groups starting with \* never cause text to be added to the document. In addition, bookmarks now create a span between the start and end of the bookmark, rather than an empty span.

Docx reader:

• Avoid blockquote when parent style has more indent (Milan Bracke). When a paragraph has an indentation different from the parent (named) style, it used to be considered a blockquote. But this only makes sense when the paragraph has more indentation. So this commit adds a check for the indentation of the parent style.
• Fix handling of empty fields (Milan Bracke). Some fields only have an instrText and no content, Pandoc didn’t understand these, causing other fields to be misunderstood because it seemed like a field was still open when it wasn’t.
• Implement PAGEREF fields (Milan Bracke). These fields, often used in tables of contents, can be a hyperlink.
• Fix handling of nested fields (Milan Bracke). Fields delimited by fldChar elements can contain other fields. Before, the nested fields would be ignored, except for the end, which would be considered the end of the parent field.
• Add placeholder for word diagram instead of just omitting it (Ezwal).

Org reader:

• Don’t parse a list as first item in a list item (#7557).
• Allow an initial :PROPERTIES: drawer to add to metadata (#7520).

Docx writer:

• Make id used in native_numbering predictable (#7551). If the image has the id IMAGEID, then we use the id ref_IMAGEID for the figure number. This allows one to create a filter that adds a figure number with figure name, e.g. <w:fldSimple w:instr=" REF ref_superfig "><w:r><w:t>Figure X</w:t> </w:r></w:fldSimple>. If an image lack an id, an id of the form ref_fig1 is used.

Ensure we have unique ids for wp:docPr and pic:cNvPr elements (#7527, #7503).

Handle SVG images (#4058). This change has several parts:

• In Text.Pandoc.App, if the writer is docx, we fill the media bag and attempt to convert any SVG images to PNG, adding these to the media bag. The PNG backups have the same filenames as the SVG images, but with an added .png extension. If the conversion cannot be done (e.g. because rsvg-convert is not present), a warning is omitted.
• In Text.Pandoc.Writers.Docx, we now use Word 2016’s syntax for including SVG images. If a PNG fallback is present in the media bag, we include a link to that too.

Powerpoint writer (Emily Bourke):

• Add support for more layouts (#5097). Up til now, four layouts were supported: “Title Slide” (used for the automatically generated metadata slide), “Section Header” (used for headings above slide level), “Two Column” (used when there’s a columns div), “Title and Content” (used for all other slides). We now support three additional layouts: “Comparison”, “Content with Caption”, and “Blank”. The manual describes the logic that determines which layout is used for a slide. Layouts may be customized in the reference doc.
• Support specifying slide background images using a background-image attribute on the slide’s heading. Only the “stretch” mode is supported, and the background image is centred around the slide in the image’s larger axis, matching the observed default behaviour of PowerPoint.
• Add support for incremental lists (through same methods as in other slide writers) (#5689).
• Copy embedded fonts from reference doc.
• Include all themes in output archive.
• Fix list level numbering (#4828, #4663). In PowerPoint, the content of a top-level list is at the same level as the content of a top-level paragraph: the only difference is that a list style has been applied. Previously, the writer incremented the paragrap h level on each list, turning what should be top-level lists into second-level lists.
• Line up list continuation paragraphs. This commit changes the marL and indent values used for plain paragraphs and numbered lists, and changes the spacing defined in the reference doc master for bulleted lists. For paragraphs, there is now a left-indent taken from the otherStyle in the master. For numbered lists, the number is positioned where the text would be if this were a plain paragraph, and the text is indented to the next level. This means that continuation paragraphs line up nicely with numbered lists. Existing reference docs may need to be modified so that otherStyle and bodyStyle indent levels match, for this feature to work with them.
• Consolidate text runs when possible (jgm). This slims down the output files by avoiding unnecessary text run elements.
• Support footers in the reference doc. There is one behaviour which may not be immediately obvious: if the reference doc specifies a fixed date (i.e. not automatically updating), and there’s a date specified in the metadata for the document, the footer date is replaced by the metadata date.
• Fix presentation rel numbering. Before now, the numbering of rIds was inconsistent when making the presentation XML and when making the presentation relationships XML.
• Don’t add relationships unnecessarily. Before now, for any layouts added to the output from the default reference doc, the relationships were unconditionally added to the output. However, if there was already a layout in slideMaster1 at the same index then that results in duplicate relationships.
• If slide level is 0, don’t insert a slide break between a heading and a following table, “columns” div, or paragraph starting with an image.
• Fix capitalisation of notesMasterId.
• Restructure tests.

Asciidoc writer:

• Translate numberLines attribute to linesnum switch (Samuel Tardieu).
• Improve escaping for -- in URLs (#7529).

LaTeX writer:

• Make babel use more idiomatic (#7604, hseg). Use babel’s bidi implementation. Import babel languages individually instead of as package options. Move header-includes to after babel setup so it can be modified.
• Use babel, not polyglossia, with xelatex. Previously polyglossia worked better with xelatex, but that is no longer the case, so we simplify the code so that babel is used with all latex engines. This involves a change to the default LaTeX template.

Markdown writer:

• Avoid bad wraps at the Doc Text level. Previously we tried to do this at the Inline list level, but it makes more sense to intervene on breaking spaces at the Doc Text level.
• Use underline class rather than ul for underline. This only affects output with bracketed_spans enabled. The markdown reader parses spans with either .ul or .underline as Underline elements, but we’re moving towards preferring the latter.

RST writer:

• Properly handle anchors to ids with spaces or leading underscore (#7593). In this cases we need the quoted form, e.g.

  .. _`foo bar`:
  
  .. _`_foo`:

  Side note: rST will “normalize” these identifiers anyway, ignoring the underscore.

HTML writer:

• Render \ref and \eqref as inline math, not display (see #7589).
• Pass through \ref and \eqref if MathJax is used (#7587).
• Pass through inline math environments with KaTeX.
• Support --reference-location for HTML output (#7461, Francesco Mazzoli).
• Set “hash” to True by default (for reveal.js) (#7574). See #6968 where the motivation for setting “hash” to True is explained.

Native writer: Use pretty-show to format native output (#7580). Previously we used our own homespun formatting. But this produces over-long lines that aren’t ideal for diffs in tests. Performance is slower by about a factor of 10, but this isn’t really a problem because native isn’t suitable as a serialization format. (For serialization you should use json, because the reader is so much faster than native.)

Org writer:

• Don’t indent contents of code blocks. We previously indented them by two spaces, following a common convention. Since the convention is fading, and the indentation is inconvenient for copy/paste, we are discontinuing this practice.
• Update list of supported source languages in org writer (#5440).

Ipynb writer (Kolen Cheung):

• Improve round trip identity for raw cell output. See jupyter/nbformat#229. The Jupyter ecosystem, including nbconvert, lab and notebook, deviated from their own spec in nbformat, where they used the key raw_mimetype instead of format. Moreover, the mime-type of rst used in Jupyter deviated from that suggested by https://docutils.sourceforge.io/FAQ.html and is defined as text/restructuredtext when chosen from “Raw NBConvert Format” in Jupyter. The new behavior should matche the real world usage better, hence improving the round-trip “identity” in raw-cell.
• Add more formats that can be handled as “raw” cells.

EPUB writer:

• Add EPUB3 subject metadata (authority/term) (nuew). This adds the ability to specify EPUB 3 authority and term specific refinements to the subject tag. Specifying a plain subject tag in metadata will function as before.
• Treat epub:type “frontispiece” as front matter (#7600).

reveal.js template: Fix line numbers in source code (#7634). We need overflow: visible for these to work, and reveal’s default css disables this. So we re-enable it in the default template.

Text.Pandoc.Writers.Shared:

• Export splitSentences as a Doc Text transform [API change]. Use this in man/ms. We used to attempt automatic sentence splitting in man and ms output, since sentence-ending periods need to be followed by two spaces or a newline in these formats. But it’s difficult to do this reliably at the level of [Inline].

Text.Pandoc.Translations: small revisions for compatibility with aeson 2.

Don’t prepend file:// to --syntax-definition on Windows (#6374). This was a fix for a problem in skylighting, but this problem doesn’t exist now that we’ve moved from HXT to xml-conduit.

Text.Pandoc.Extensions:

• Add Ext_footnotes to default gfm etxensions. Now gfm supports footnotes.
• Alphabetize Extension constructors (also affects --list-extensions).

Text.Pandoc.Citeproc.Util: Better implementation of splitStrWhen. Previously the citeproc code had two less efficient implementations.

Update documentation for definition_list extension (#7594). In 2015, we relaxed indentation requirements for the first line of a definition (see commit d3544dc and issue #2087), but the documnentation wasn’t updated to reflect the change.

Text.Pandoc.Citeproc.BibTeX: Fix expansion of special strings in series e.g. newseries or library (#7591). Expansion should not happen when these strings are protected in braces, or when they’re capitalized.

Text.Pandoc.Logging: add NotUTF8Encoded constructor to LogMessage [API change].

Text.Pandoc.App.FormatHeuristics: remove .tei.xml extension for TEI (#7630). This never worked, because takeExtension only returns .xml. So it won’t be missed if we remove it.

Text.Pandoc.Image:

• Generalize svgToPng to MonadIO.
• svgToPng, change first parameter from WriterOptions to Int.

Text.Pandoc.Class:

• Add readStdinStrict method to PandocMonad [API change].
• Generalize type of extractMedia [API change]. It was uselessly restricted to PandocIO, instead of any instance of PandocMonad and MonadIO.

Text.Pandoc.Class.PandocIO: derive MonadCatch, MonadThrow, MonadMask. This allows us to use withTempDir [API change].

Add module Text.Pandoc.Class.Sandbox, defining sandbox. Exported via Text.Pandoc.Class. [API change]

Text.Pandoc.Filter: Generalize type of applyFilters from PandocIO to any instance of MonadIO and PandocMonad [API change].

Text.Pandoc.PDF: generalize type of makePDF: instead of PandocIO, it can be used in any instance of PandocMonad, MonadIO, and MonadMask [API change].

Lua subsystem and custom writers: generalize types from PandocIO to any instance of PandocMonad and MonadIO [API change]. The type of runLua is now

(PandocMonad m, MonadIO m) =>
LuaE PandocError a -> m (Either PandocError a)

The change from Lua to LuaE PandocError is due to the switch to hslua-2.0; see next item.

Lua modules (Albert Krewinkel):

• Switch to hslua-2.0. The new HsLua version takes a somewhat different approach to marshalling and unmarshalling, relying less on typeclasses and more on specialized types. This allows for better performance and improved error messages. Furthermore, new abstractions allow to document the code and exposed functions.
• Marshal Version values, Inline elements, Attr elements, and Pandoc elements as userdata.
• Remove deprecated inline constructors DoubleQuoted, SingleQuoted, DisplayMath, and InlineMath.
• Attr values are no longer normalized when assigned to an Inline element property.
• It’s no longer possible to access parts of Inline elements via numerical indexes. E.g., pandoc.Span('test')[2] used to give pandoc.Str 'test', but yields nil now. This was undocumented behavior not intended to be used in user scripts. Use named properties instead.
• Accessing .c to get a JSON-like tuple of all components no longer works. This was undocumented behavior.
• Only known properties can be set on an element value. Trying to set a different property will now raise an error.
• Adds a new pandoc.AttributeList() constructor, which creates the associative attribute list that is used as the third component of Attr values. Values of this type can often be passed to constructors instead of Attr values.
• Convert IOErrors to PandocErrors in pandoc.pipe function (#7523).

Text.Pandoc.PDF: Previously we had to run runIOorExplode inside withTempDir. Now that PandocIO is an instance of MonadMask, this is no longer necessary.

Text.Pandoc.App:

• Reorganize to make it easier to limit IO in main loop. Previously we used liftIO fairly liberally. The code has been restructured to avoid this.
• Move output-file writing out of PandocMonad action.

Text.Pandoc.App.OutputSettings: Generalize some types so we can run this with any instance of PandocMonad and MonadIO, not just PandocIO.

Use simpleFigure builder in readers and SimpleFigure pattern synonym in writers (Aner Lucero).

Allow time 1.12.

Use skylighting-0.12, skylighting-core-0.12. This fixes highlighting issues with typescript, scala, and other syntaxes that include keyword lists from different syntaxes.

Use citeproc 0.6, commonmark 0.2.2.1, commonmark-extensions 0.2.2, texmath 0.12.3.2, ipynb 0.1.0.2. (These changes also allow building with aeson >= 2.)

Require doclayout >= 0.3.1.1. This fixes recognition of “real widths” of emoji characters, which is important for tabular layout.

Cut out over 100K of fat in epub test golden files.

Make test/epub/wasteland.epub valid.

Add missing % on some command tests. This prevented --accept from working properly.

Command tests: raise error if command doesn’t begin with %.

OOXML tests: use pretty-printed form to display diffs. Otherwise everything is on one line and the diff is uninformative.

Fix compareXML helper in Tests.Writers.OOXML. Given how it is used, we were getting “mine” and “good” flipped in the test results.

MANUAL.txt:

• Clarify attributes extension support (William Lupton).
• Document formats affected by --reference-location.
• Document error code 25
• Add some more info regarding --slide-level=0 (Salim B).
• Add more to security section of manual.
• Mention support of title-toc (#7171, Christophe Dervieux).

doc/lua-filters.md:

• Add missing type for Image title (Quinn).
• Improve order of Image fields (Quinn).
• Rephrase pandoc.path docs (#7548, Quinn).
• Do not leak working directory in TikZ filter (Jeroen de Haas).

Allow --slide-level=0 (#7476). When the slide level is set to 0, headings won’t be used at all in splitting the document into slides. Horizontal rules must be used to separate slides.

Add RTF reader (#3982). rtf is now supported as an input format as well as an output format. New module Text.Pandoc.Readers.RTF (exporting readRTF). [API change]

HTML reader: treat comments as blank when parsing (#7482).

Markdown reader:

• Fix raw LaTeX injection issue (#7497). Using a code block containing \end{verbatim}, one could inject raw TeX into a LaTeX document even when raw_tex is disabled. Thanks to Augustin Laville for noticing the bug.
• Multimarkdown sub- and superscripts (#5512, OCzarnecki). Added an extension short_subsuperscripts which modifies the behavior of subscript and superscript, allowing subscripts or superscripts containing only alphanumerics to end with a space character (eg. x^2 = 4 or H~2 is combustible). This improves support for multimarkdown.

RST reader: Fix :literal: includes (#7513). These should create code blocks, not insert raw RST.

LaTeX reader:

• Proper implicit grouping around environment macros.
• Support \global before \def, \let, etc. (#7494).
• Fix scope for LaTeX macros (#7494). They should by default scope over the group in which they are defined (except \gdef and \xdef, which are global). In addition, environments must be treated as groups.
• Improve handling of plain TeX macro primitives (#7474). Fixed semantics for \let.
• Implement \edef, \gdef, and \xdef.

Docx reader: Improve docx reader’s robustness in extracting images (#7511). The docx reader made some assumptions about how docx containers were laid out that were not always true, with the result that some images in documents did not get extracted.

LaTeX writer: Increase table column width precision (#7466, Peter Fabinski). In some cases, the rounding performed by the LaTeX table writer would introduce visible overrun outside the text area. This adds two more decimal places to the width values.

Powerpoint writer:

• Include image title in description (#7352, Emily Bourke). The image title (i.e. ![alt text](link "title")) was previously ignored when writing to pptx. This commit includes it in PowerPoint’s description of the image, along with the link.
• Select layouts from reference doc by name (Emily Bourke). Until now, users had to make sure that their reference doc contains layouts in a specific order: the first four layouts in the file had to have a specific structure. Now the layout selection uses the layout names rather than order: users must make sure their reference doc contains four layouts with specific names, and if a layout with the right name isn’t found pandoc will emit a warning and use the corresponding layout from the default reference doc as a fallback.

Docx writer: be sensitive to the native_numbering extension (#7499). Figure and table numbers are now only included if native_numbering is enabled. (By default it is disabled.) This is a behavior change with respect to 2.14.1, but the default behavior is now that of previous versions. The change was necessary to avoid incompatibilities between pandoc’s native numbering and third-party cross reference filters like pandoc-crossref.

RTF writer:

• Omit \bin in \pict. According to the spec, this is not needed or wanted when the data is in hexadecimal format, as here.
• Emit ``` for section headings.

RTF template: specify font family for fixed-width font f1. According to the spec, this is mandatory.

LaTeX writer: Use ulem for underline (#7351). ulem is conditionally included already when the strikeout variable is set, so we set this when there is underlined text, and use \uline instead of \underline. This fixes wrapping for underlined text.

Text.Pandoc.Citeproc:

• Revise citeproc code to fit new citeproc 0.5 API (thanks to Benjamin Bray). Linkification of URLs in the bibliography is now done in the citeproc library, depending on the setting of an option. We set that option depending on the value of the metadata field link-bibliography (defaulting to true, for consistency with earlier behavior). If a DOI, PMID, PMCID, or URL field is present but not explicitly rendered, the title (or if no title, the whole entry) is hyperlinked. These changes implement the recommendations from the draft CSL v1.0.2 spec (Appendix VI): https://github.com/citation-style-language/documentation/blob/master/specification.rst#appendix-vi-links
• Avoid odd handling of quotes. Recent citeproc changes allow us to ignore Quoted elements; citeproc now uses its own method for represented quoted things, and only localizes and flipflops quotes it adds itself. Convert Quoted in bib entries to special Spans before passing them off to citeproc. This ensures that we get proper localization and flipflopping if, e.g., quotes are used in titles (jgm/citeproc#87).
• Removed quote localization from citeproc processing. This is now done in citeproc itself.

Text.Pandoc.Logging: Add PowerpointTemplateWarning log message type [API change] (Emily Bourke).

Text.Pandoc.Extension: Add Ext_short_subsuperscripts constructor to Extension [API change] (OCzarnecki).

Various sample.lua editorial fixes (#7493, #7487, William Lupton).

Bump base-compat version so we get compatibility with base 4.12.

Use Prelude from base-compat for ghc 8.4 too.

Add haskell-language-server to shell.nix (#7496, Emily Bourke).

Tests.Helpers: export testGolden and use it in RTF reader. This gives a diff output on failure.

Remove obsolete and incorrect sentence in --slide-level docs.

Add internal module Text.Pandoc.Network.HTTP, exporting urlEncode.

Text.Pandoc.Parsing: parseFromString: preserve at least the source directory (#7464). Previously we just set the source name to “chunk” when parsing from strings, to avoid misleading source positions. This had the side effect that rebase_relative_paths would break inside sections that were parsed as strings. So, now we use “ORIGINAL_SOURCE_PATH_chunk” instead of just “chunk”.

Text.Pandoc.MIME: use image/x-xcf instead of application/x-xcf (#7454).

Don’t compare cdLine in OOXML golden tests (Emily Bourke). The cdLine field gives the line of the file some CData was found on, which reflects irrelevant formatting differences.

Provide more detailed XML diff in tests (Emily Bourke).

OOXML tests: silence warnings. These can make the test output confusing, making people think tests are failing when they’re passing.

INSTALL.md: Add GitLab CI/CD example (#7448, Veratyr).

MANUAL.txt

• Clarifications (William Lupton).
• Add a note on security risks of include directives.

Document use of the ‘underline’ class (#7492, #7484, William Lupton).

Add a FAQ about the “Cannot allocate memory” error on M1 macs.

Use texmath 0.12.3.1.

Use released citeproc 0.5.

Remove dependency on HTTP package (#7456, mt_caret).

Text.Pandoc.ImageSize: Add Tiff constructor for ImageType (#7405) [Minor API change]. This allows pandoc to get size information from tiff images.

Markdown reader: don’t try to read contents in self-closing HTML tag. Previously we had problems parsing raw HTML with self-closing tags like <col/>. The problem was that pandoc would look for a closing tag to close the markdown contents, but the closing tag had, in effect, already been parsed by htmlTag.

LaTeX reader:

• Avoid trailing hyphen in translating languages (#7447). Previously \foreignlanguage{english} turned into <span lang="en-">. The same issue affected Arabic.
• Support \cline in LaTeX tables (#7442).
• Improved parsing of raw LaTeX from Text streams (rawLaTeXParser, used to read LaTeX in Markdown files, #7434). We now use source positions from the token stream to tell us how much of the text stream to consume. Getting this to work required a few other changes to make token source positions accurate.

DocBook reader:

• Handle images with imageobjectco elements (#7440).
• Add support for citerefentry (#7437, Jan Tojnar).

RST reader: fix regression with code includes (#7436). With the recent changes to include infrastructure, included code blocks were getting an extra newline.

HTML reader:

• Recognize data-external when reading HTML img tags (#7429, Michael Hoffmann). Preserve all attributes in img tags. If attributes have a data- prefix, it will be stripped. In particular, this preserves a data-external attribute as an external attribute in the pandoc AST.
• Add col, colgroup to ‘closes’ definitions

HTML writer:

• Remove duplicated alt text in HTML output (Aner Lucero).
• Remove aria-hidden when explicit alt text is provided (Aner Lucero).
• Set boolean values for reveal.js variables.

Docx writer:

• Add table numbering for captioned tables. The numbers are added using fields, so that Word can create a list of tables that will update automatically.
• Support figure numbers. These are set up in such a way that they will work with Word’s automatic table of figures (#7392).

Markdown writer: put space between Plain and following fenced Div (#4465).

EPUB writer: Don’t incorporate externally linked images in EPUB documents (#7430, Michael Hoffmann). Just as it is possible to avoid incorporating an image in EPUB by passing data-external="1" to a raw HTML snippet, this makes the same possible for native Images, by looking for an associated external attribute.

Text.Pandoc.PDF:

• Fix svgIn path error (#7431). We were duplicating the temp directory; this didn’t cause problems on macOS or linux because there we use absolute paths for the temp directory. But on Windows it caused errors converting SVG files.
• convertImage: normalize paths (#7431). This will avoid paths on Windows with mixed path separators.

Text.Pandoc.Class: Always use / when adding directory to image destination with extractMedia, even on Windows.

Text.Pandoc.Citeproc:

• Allow $ characters in bibtex keys (#7409).
• Set proper initial source name in parsing BibTeX (for better error messages.)
• Revamp note citation handling (#7394). Use latest citeproc, which uses a Span with a class rather than a Note for notes. This helps us distinguish between user notes and citation notes. Don’t put citations at the beginning of a note in parentheses. Fix small bug in handling of citations in notes, which led to commas at the end of sentences in some cases.
• Cleanup and efficiency improvement in deNote.
• Improve punctuation moving with --citeproc. Previously, using --citeproc could cause punctuation to move in quotes even when there aer no citations. This has been changed; punctuation moving is now limited to citations. In addition, we only move footnotes around punctuation if the style is a note style, even if notes-after-punctuation is true.

Use citeproc 0.10. This helps improve note citations (see above) and eliminates double hyperlinks in author-in-text citations. Author-only citations are no longer hyperlinked. See jgm/citeproc#77. It also fixes moving of punctuation inside quotes to conform to the CSL spec: only comma and period are moved, not question mark or exclamation point.

Text.Pandoc.Error: fix line calculations in reporting parsec errors. Also remove a spurious initial newline in the error report.

Use doctemplates 0.4.1, which gives us better support for boolean variable values. Previously $if(foo)$ would evaluate to true for variables with boolean false values, because it cared only about the string rendering (#7402).

Require commonmark-pandoc >= 0.2.2.1. This fixes task lists with multiple paragraphs.

Use skylighting 0.11.

CSS in HTML template: reset overflow-wrap on code blocks (Mauro Bieg, #7423).

LaTeX template: Revert change in PR #7295: “move title, author, date up to top of preamble.” The change caused problem for people who used LaTeX commands defined defined later in the preamble in the title or author fields (#7422).

Add doc/faqs.md. This is imported from the website; in the future the website version will be drawn from here. Added a FAQ on the use of \AtEndPreamble for cases when the contents of header-includes need to refer to definitions that come later in the preamble. See #7422.

Upgrade Debian 10 AMI for build-arm.sh.

CircleCI: change to using xcode 11.1.0 (macOS 10.14.4). We previously built on 10.13, but 10.13 no longer gets security updates and CirclCI is deprecating.

• Add strut at end of minipage if it contains line breaks. Without them, the last line is not as tall as it should be in some cases.
• Always use a minipage for cells with line breaks, when width information is available (#7393). Otherwise the way we treat them can lead to content that overflows a cell.
• Use \strut instead of ~ before \\ in empty line.

Fix MediaBag regressions (#7345). iIn the 2.14 release --extract-media stopped working as before; there could be mismatches between the paths in the rendered document and the extracted media. This patch makes several changes that restore the earlier behavior (while keeping the same API). The mediaPath in 2.14 was always constructed from the SHA1 hash of the media contents. Now, we preserve the original path unless it’s an absolute path or contains .. segments (in that case we use a path based on the SHA1 hash of the contents).

In Text.Pandoc.MediaBag, mediaDirectory and mediaItems now use the mediaPath, rather than the mediabag key, for the first component of the tuple. This makes more sense, I think, and fits with the documentation of these functions; eventually, though, we should rework the API so that mediaItems returns both the keys and the MediaItems.

In Text.Pandoc.Class.IO, rewriting of source paths in extractMedia has been fixed.

In Text.Pandoc.Class.PandocMonad, fillMediaBag has been modified so that it doesn’t modify image paths (that was part of the problem in #7345).

We now do path normalization (e.g. \ separators on Windows) in writing the media.

Text.Pandoc.PDF:

• Text.Pandoc.PDF: Fix regression in 2.14 for generation of PDFs with SVGs (#7344).
• Only print relevant part of environment on --verbose. Since --verbose output might be put in an issue, we want to avoid spilling out secrets in environment variables.

Markdown reader: fix pipe table regression in 2.11.4 (#7343). Previously pipe tables with empty headers (that is, a header line with all empty cells) would be rendered as headerless tables. This broke in 2.11.4. The fix here is to produce an AST with an empty table head when a pipe table has all empty header cells.

LaTeX reader: don’t allow optional * on symbol control sequences (#7340). Generally we allow optional starred variants of LaTeX commands (since many allow them, and if we don’t accept these explicitly, ignoring the star usually gives acceptable results). But we don’t want to do this for \(*\) and similar cases.

Docx reader: handle absolute URIs in Relationship Target (#7374).

Docx writer: fix handling of empty table headers (Albert Krewinkel, #7369). A table header which does not contain any cells is now treated as an empty header.

LaTeX writer: Fix regression in table header position (#7347). In recent versions the table headers were no longer bottom-aligned (if more than one line). This patch fixes that by using minipages for table headers in non-simple tables.

CommonMark writer:

• Do not use simple class for fenced-divs (Jan Tojnar, amends #7242.)
• Do not throw away attributes when Ext_attributes is enabled. Ext_attributes covers at least the following: Ext_fenced_code_attributes, Ext_header_attributes, Ext_inline_code_attributes, Ext_link_attributes.

Markdown writer:

• Allow pipe_tables to be disabled for commonmark formats (commonmark_x, gfm) (#7375).
• Re-use functions from Text.Pandoc.Markdown.Inline (Jan Tojnar).

DocBook writer: Remove non-existent admonitions (Jan Tojnar). attention, error and hint are reStructuredText specific.

HTML writer: Don’t omit width attribute on div (#7342).

Text.Pandoc.MIME, extensionFromMimeType: add a few special cases. When we do a reverse lookup in the MIME table, we just get the last match, so when the same mime type is associated with several different extensions, we sometimes got weird results, e.g. .vs for text/plain. These special cases help us get the most standard extensions for mime types like text/plain.

Lua utils: fix handling of table headers in from_simple_table (Albert Krewinkel, #7369). Passing an empty list of header cells now results in an empty table header.

Text.Pandoc.Citeproc:

• Avoid duplicate classes and attributes on references div.
• Fix regression in citeproc processing (#7376). If inline references are used (in the metadata references field), we should still only include in the bibliography items that are actually cited (unless nocite is used).

Require citeproc 0.4.0.1. This fixes a bug which led to doubled “et al.” in some (rare) circumstances.

MANUAL.txt:

• Mention GladTeX for EPUB export (Sebastian Humenda). This updates the manual and the web site about the GladTeX usage.
• More details and a useful link for YAML syntax.

CONTRIBUTING.md: update modules overview (Albert Krewinkel).

using-the-pandoc-api.md: switch from String to Text (Albert Krewinkel).

Commonmark reader: Fix regression in 2.14 with YAML metdata block parsing, which could cause the document body to be omitted after metadata (#7339).

HTML reader: fix column width regression in 2.14 (#7334). Column widths specified with a style attribute were off by a factor of 100.

Markdown reader: in rebasePaths, check for both Windows and Posix absolute paths. Previously Windows pandoc was treating /foo/bar.jpg as non-absolute.

Text.Pandoc.Logging: In rendering LoadedResource, use relative paths.

Docx writer: fix regression on captions (#7328). The “Table Caption” style was no longer getting applied. (It was overwritten by “Compact.”)

Use commonmark-extensions 0.2.1.2

Change reader types, allowing better tracking of source positions [API change]. Previously, when multiple file arguments were provided, pandoc simply concatenated them and passed the contents to the readers, which took a Text argument. As a result, the readers had no way of knowing which file was the source of any particular bit of text. This meant that we couldn’t report accurate source positions on errors or include accurate source positions as attributes in the AST. More seriously, it meant that we couldn’t resolve resource paths relative to the files containing them (see e.g. #5501, #6632, #6384, #3752).

Add rebase_relative_paths extension (#3752). When enabled, this extension rewrites relative image and link paths by prepending the (relative) directory of the containing file. This behavior is useful when your input sources are split into multiple files, across several directories, with files referring to images stored in the same directory. The extension can be enabled for all markdown and commonmark-based formats.

Add Text.Pandoc.Sources (exported module), with a Sources type and a ToSources class. A Sources wraps a list of (SourcePos, Text) pairs [API change]. A parsec Stream instance is provided for Sources. The module also exports versions of parsec’s satisfy and other Char parsers that track source positions accurately from a Sources stream (or any instance of the new UpdateSourcePos class).

Text.Pandoc.Parsing

• Export the modified Char parsers defined in Text.Pandoc.Sources instead of the ones parsec provides. Modified parsers to use a Sources as stream [API change].
• Improve include file functions [API change]. Remove old insertIncludedFileF. Give insertIncludedFile a more general type, allowing it to be used where insertIncludedFileF was.
• Add parameter to the citeKey parser from Text.Pandoc.Parsing, which controls whether the @{..} syntax is allowed [API change].

Text.Pandoc.Error: Modified the constructor PandocParsecError to take a Sources rather than a Text as first argument, so parse error locations can be accurately reported.

Fix source position reporting for YAML bibliographies (#7273).

Issue error message when reader or writer format is malformed (#7231). Previously we exited with an error status but (due to a bug) no message.

Smarter smart quotes (#7216, #2103). Treat a leading " with no closing " as a left curly quote. This supports the practice, in fiction, of continuing paragraphs quoting the same speaker without an end quote. It also helps with quotes that break over lines in line blocks.

Markdown reader:

• Use MetaInlines not MetaBlocks for multimarkdown metadata fields. This gives better results in converting to e.g. pandoc markdown.
• Implement curly-brace syntax for Markdown citation keys (#6026). The change provides a way to use citation keys that contain special characters not usable with the standard citation key syntax. Example: @{foo_bar{x}'} for the key foo_bar{x}. It also allows separating citation keys from immediately following text, e.g. @{foo}A.

RST reader:

• Seek include files in the directory of the file containing the include directive, as RST requires (#6632).
• Use insertIncludedFile from Text.Pandoc.Parsing instead of reproducing much of its code.

Org reader: Resolve org includes relative to the directory containing the file containing the INCLUDE directive (#5501).

ODT reader: Treat tabs as spaces (#7185, niszet).

Docx reader:

• Add handling of vml image objects (#7257, mbrackeantidot).
• Support new table features (Emily Bourke, #6316): column spans, row spans, multiple header rows, table description (parsed as a simple caption), captions, column widths.

LaTeX reader:

• Improved siunitx support (#6658, #6620).
• Better support for \xspace (#7299).
• Improve parsing of \def macros. We previously set “verbatim mode” even for parsing the initial \def; this caused problems for \def nested inside another \def.
• Implement \newif.

ConTeXt writer: improve ordered lists (#5016, Denis Maier). Change ordered list from itemize to enumerate. Add new itemgroup for ordered lists. Remove manual insertion of width attributes. Use tabular figures in ordered list enumerators.

HTML reader:

• Don’t fail on unmatched closing “script” tag (Albert Krenkel, #7282).
• Keep h1 tags as normal headers (#2293, Albert Krewinkel). The tags <title> and <h1 class="title"> often contain the same information, so the latter was dropped from the document. However, as this can lead to loss of information, the heading is now always retained. Use --shift-heading-level-by=-1 to turn the <h1> into the document title, or a filter to restore the previous behavior.
• Handle relative lengths (e.g. 2*) in HTML column widths (#4063). See https://www.w3.org/TR/html4/types.html#h-6.6.

DocBook/JATS readers:

• Fix mathml regression caused by the switch in XML libraries (#7173).
• Fix “phrase” in DocBook: take classes from “role” not “class” (#7195).

DocBook reader: ensure that first and last names are separated (#6541).

Jira reader (Albert Krewinkel, #7218):

• Support “smart” links: [alias|https://example.com|smart-card] syntax.
• Allow spaces and most unicode characters in attachment links.
• No longer require a newline character after {noformat}.
• Only allow URI path segment characters in bare links.
• The file: schema is no longer allowed in bare links; these rarely make sense.

Plain writer: handle superscript unicode minus (#7276).

LaTeX writer:

• Better handling of line breaks in simple tables (#7272). Now we also handle the case where they’re embedded in other elements, e.g. spans.
• For beamer output, support exampleblock and alertblock (#7278). A block will be rendered as an exampleblock if the heading has class example and an alertblock if it has class alert.
• Separate successive quote chars with thin space (#6958, Albert Krewinkel). Successive quote characters are separated with a thin space to improve readability and to prevent unwanted ligatures. Detection of these quotes sometimes had failed if the second quote was nested in a span element.
• Separate successive quote chars with thin space (#6958, Albert Krewinkel).