Added minimal book and website project templates that can be discovered by RStudio from File > New Project > New Directory > Project using litedown. You can also create them programmatically with litedown:::proj_skeleton(path, type), where type is "book" or "site" (thanks, @jangorecki, #97).
Added fuse_exit(), which can be called inside a code chunk to exit fuse() early and discard the rest of the document. It takes an optional append argument to add text after the current chunk (thanks, @TimTaylor, #117).
Added support for optimizing PNG plot files via the global option litedown.tinypng. When set to TRUE and the tinyimg package is installed, PNG images generated from code chunks will be optimized by tinyimg::tinypng() to reduce file size. For other PNG files (e.g., images manually included in the document), users can call tinyimg::tinypng('.') to optimize all PNG files under the current directory (thanks, @bastistician, #116).
Added new engines exec, sh, bash, zsh, and powershell to run arbitrary commands via system2() and capture the output. The exec engine requires the chunk option command; the sh/bash/zsh/powershell engines use the engine name as the command.
When the output argument of mark() is a .pdf file, Markdown will be converted to a full .tex file instead of a LaTeX fragment before it is compiled to PDF.
Fixed a bug that sections in the appendix could not be cross-referenced.
Added a meta variable lang for HTML output, which is the language of the document (e.g., en-US for US English). This variable is used in the <html lang="..."> tag of the HTML output file. By default, the language is detected from the system locale, but you can also set it via the lang field under meta in YAML metadata, e.g., lang: en-GB for British English (thanks, @TimTaylor, #121).
Fixed a bug when embedding JS resources: previously </ was escaped to <\/ to avoid </script> from being present in the JS source, but the escaping was too general (e.g., </ may appear in a regex /</g). Now we only escape </script> to <\/script>.
For fuse_site(), the generated site menu now automatically includes landing pages for one-level subdirectories containing index.html (e.g., playground/index.html), so subdirectory index pages appear in the navigation.
Fixed a re-entrancy bug that chunk options fig.path and cache.path in nested fuse() calls caused overridden figure/cache files (thanks, @nanxstats, #127).
Fixed a bug that detect_pkg() doesn’t work when R CMD check checks package vignettes.
Provided a new chunk option filter to filter the output elements via a custom function. This makes it possible to re-order output elements. As a result, text output and plots from a for-loop can be interleaved (thanks, @reedacartwright, #106).
The chunk option attr.source will default to .lang (where lang is the engine name) only when it is not provided, i.e., NULL. Previously the value .lang would still be used when attr.source has been provided. Now it’s possible to completely override it, e.g., attr.source = 'language-r' as in #107 (thanks, @ThomasSoeiro).
The chunk option collapse = TRUE also applies to message blocks, including warnings, messages, and errors (thanks, @ThomasSoeiro, #108).
pkg_desc() gained a new argument type and the package description can be generated to either a table or a definition list now.
The embed_resources option was buggy for deferred JS resources. Previously they were moved to <head>, but deferred scripts should be executed after the full DOM is ready, so they are moved before </body> instead if they are to be embedded.
Fixed a bug in pkg_manual() that may lead to omission of certain items when building TOC.
Added a new chunk option fig.keep to select plots to be kept in a code chunk (thanks, @Gabrielforest, #99). See https://pkg.yihui.org/litedown/book/#sec:option-fig for documentation.
Improved support for LaTeX footnotes. The footnote identifier no longer has to be a number, and the footnote can include arbitrary elements (not necessarily a single paragraph).
The chunk option results = 'hide' will imply collapse = TRUE, i.e., when text output is hidden, the source blocks will be merged (thanks, @jangorecki, #87).
fuse(text = '# text') will be treated as Markdown input instead of R code input (thanks, @chuxinyuan, #102).
Fixed the bug that the chunk option fig.path does not work when it does not contain / (thanks, @J-Moravec, #88).
get_context('full_input') gives the full path to the input file of fuse() (thanks, @rikivillalba, #104).
pkg_manual() will exclude help pages with the keyword internal (thanks, @TimTaylor, #78).
pkg_news() will add links to Github users and issue numbers if the package is hosted on Github.
Allow , and . in superscripts and subscripts (thanks, @janlisec, #81).
Special characters in bibliography entries such as {\"u} can be correctly read now (thanks, @bastistician).
Fixed a bug that inline code expressions (`{lang} expr`) cannot be correctly located when the line has leading spaces that are not meaningful.
Deleted vignettes markdown-examples and markdown-output. They are rendered on the package site now: https://git.yihui.org/litedown/examples/test-options.html
Added a Markdown rendering option offline to download web resources when this option is set to true, so that the HTML output can be viewed offline (thanks, @TimTaylor, #73). See https://pkg.yihui.org/litedown/book/#sec:offline for more info.
Added a function get_context() to query the fuse() context such as the input file path or the output format (thanks, @MichaelChirico #67, @vincentarelbundock #70).
Added a function raw_text() to output raw text content in a code chunk (thanks, @vincentarelbundock, #69).
Dropped the chunk option ref.label and added a new chunk option fill, which is more general (ref.label = "LABEL" can be achieved by `<LABEL>` inside a chunk). See https://pkg.yihui.org/litedown/book/#sec:option-fill for more information.
Fixed a bug that fuse() fails to print the error location when the whole input document consists of a single chunk that throws an error (thanks, @kevinushey, yihui/knitr#2387).
fuse_book() will ignore YAML headers in book chapters except for the index chapter.
Added a wizard in roam() to create new .Rmd/.md/.R files with selected HTML features.
Added a new engine embed to embed text files via a code chunk.
Changed the meaning of the chunk option order: previously, higher values indicate earlier execution; now higher values indicate later execution. This is a breaking change, but the new meaning should feel more natural. For example, order = i means to execute the chunk in the i-th step, and order = i - 1.5 means to move the chunk back 1.5 step in the queue so it will be executed earlier than its previous chunk. See https://pkg.yihui.org/litedown/book/#sec:option-order for details.
Shortened the output format names litedown::html_format to html, and litedown::latex_format to latex. The names litedown::* can still be used if you like.
Added options dollar, signif, and power to format numbers from inline code. See https://pkg.yihui.org/litedown/book/#sec:inline-code for details.
When embedding SVG images in HTML output, embed their raw XML content instead of base64 encoding them.
Empty table headers are removed in HTML output (they may be generated from data frames or matrices without column names).
Added support for the chunk option collapse = TRUE (thanks, @J-Moravec, #40).
Added support for the chunk option fig.dim, which is a shortcut for fig.width and fig.height.
Added a new function vest() as another way to add CSS/JS assets to HTML output.
Provided templates and a Github action yihui/litedown/site to build package websites. See https://pkg.yihui.org/litedown/book/#sec:pkg-site for details.
Added an argument examples to pkg_manual() to run examples and show their output (thanks, @TimTaylor, #54).
Fixed a bug that the default CSS wouldn’t be added when a math expression exists on the page (thanks, @calvinw, #61).
Fixed a bug that cross-references to other chapters of a book could not be resolved when previewing a single chapter.
Fixed a bug that the file navigation by line numbers on code blocks stopped working in litedown::roam() due to yihui/lite.js@5e06d19.
Fixed a bug that R code blocks could not be embedded when using prism.js for syntax highlighting (thanks, @TimTaylor, #53).
pkg_manual() will point out the name of the problematic Rd file when the Rd file fails to convert to HTML (thanks, @BSchamberger).
Dropped knitr and rmarkdown from the Suggests field in DESCRIPTION. Previously, litedown allowed rmarkdown::render() to use the output formats litedown::html_format and litedown::latex_format. Now rmarkdown::render() is no longer supported, and litedown::fuse() must be used instead.
Provided an option options(litedown.roam.cleanup = TRUE) to clean up the *__files/ directory after previewing .Rmd or .R files via litedown::roam() (thanks, @TimTaylor, #36).
Added the keyboard shortcut Ctrl + K (or Command + K on macOS) for rendering a file to disk in the litedown::roam() preview.
Cross-references also work for LaTeX output now.
Fixed an error in the internal function detect_pkg() during R CMD check on CRAN.
Set options(bitmapType = 'cairo') on macOS only when xquartz is available. Previously only capabilities('cairo') was checked, which was not enough. This option can also be manually set via options(bitmapType) in a code chunk if the automatic switch to cairo is not desired.
Fixed the bug that indented or quoted code blocks are not correctly indented or quoted when a code expression contains multiple lines.
Fixed the bug that the span syntax [text](){...} doesn’t work when text contains markup (e.g., bold or italic).
Added a new engine md to output Markdown text both verbatim and as-is, which can be useful for showing Markdown examples, e.g.,
```{md}
You can see both the _source_ and _output_ of
this `md` chunk.
```
You can also use `{md} the engine **inline**`.
Added a new engine mermaid to generate Mermaid diagrams, e.g.,
```{mermaid, fig.cap='A nice flowchart.'}
graph TD;
A-->B;
A-->C;
B-->D;
C-->D;
```
Added helper functions pkg_desc(), pkg_news(), pkg_citation(), pkg_code(), and pkg_manual() to get various package information for building the full package documentation as a single-file book (thanks, @jangorecki @llrs #24, @TimTaylor #22).
LaTeX math environments such as equations can be numbered and cross-referenced now (thanks, @hturner, #32).
Section headings containing the class name “unlisted” will be excluded in the table of contents.
Provided a way to write <span> with attributes based on empty links, i.e., [text](){.class #id ...}. The empty URL here tells mark() to treat the link as a <span> instead of <a>.
Added back/forward/refresh/print buttons to the toolbar in the litedown::roam() preview interface.
Changed the behavior of .Rmd and .R file links in the litedown::roam() interface: previously, clicking on an .Rmd or .R filename will execute them; now it will only show their content, because fully executing the code may be expensive or even dangerous (especially when the files were not authored by you). A new “Run” button has been provided in the interface, on which you can click on to run a file in memory and preview it (i.e., the old behavior of clicking on filenames). You should use this button only if you trust the file.
Added the JS asset @mathjax-config to enable equation numbering by default when the JS math library is set to MathJax (thanks, @hturner, #32).
Set options(bitmapType = 'cairo') in fuse() if capabilities('cairo') is TRUE, which will generate smaller bitmap plot files (e.g., .png) than using quartz or Xlib, and is also a safer option for fuse() to be executed in parallel (rstudio/rmarkdown#2561).
Added a new vignette engine litedown::book to make it possible to build multiple vignettes into a book. To use this engine, declare \VignetteEngine{litedown::book} only in the book index file (e.g., index.Rmd) but not in other book chapter files.
Added support for an array of multiple authors in the YAML metadata (thanks, @AlbertLei, #28). If the author field in YAML is an array of length > 1, each author will be written to a separate <h2> in HTML output, or concatenated by \and in LaTeX output. Note that you can also write multiple authors in a single string (e.g., author: "Jane X and John Y") instead of using an array (author: ["Jane X", "John Y"]), in which case the string will be treated as a single author (they will be put inside a single <h2> in HTML output).
Fixed the bug that the leading -, +, or * in a LaTeX math expression was recognized as the bullet list marker, which would invalidate the math expression (thanks, @hturner, #33).
Changed the first - to : in automatically generated element IDs, including section, figure, and table IDs, e.g., the ID sec-intro-methods is changed to sec:intro-methods, and fig-nice-plot is changed to fig:nice-plot. You can still use - when manually assigning IDs to elements, e.g., # Intro Methods {#sec-intro-methods}. For backward compatibility, cross-references using - will be resolved if the : version of the ID can be found, e.g., @sec-intro-methods will be resolved to @sec:intro-methods if the former cannot be found but the latter can.
Fixed a bug that when LaTeX math environments are written in raw LaTeX blocks (i.e., ```{=latex}), mark() will not load the math JS library such as MathJax or KaTeX unless $ $ or $$ $$ expressions are present in the document.
As-is output accepts attributes via the chunk option attr.asis now. If provided, as-is output will be wrapped in a fenced Div with these attributes.
Numeric output from inline code will no longer be formatted if the value is wrapped in I().
The prefix for the automatic IDs of h1 headings has been changed from sec: to chp:. For other levels of headings, the prefix is still sec:.
Provided a new option embed_cleanup to clean up plot files that have been embedded in HTML output (thanks, @TimTaylor, #16).
fuse() supports the output format litedown::markdown_format now, which generates the intermediate Markdown from R Markdown without further rendering Markdown to other formats. Using this output format is equivalent to fuse(..., output = '.md') or fuse(..., output = 'markdown') (thanks, @mikmart, #35).
A data frame (or matrix/tibble) wrapped in I() is fully printed to a table now by default. Without I(), data objects are truncated to 10 rows by default when printing to tables.
When options(litedown.fig.alt = TRUE) and the chunk option fig.alt is unset, fuse() will emit reminders about the missing alt text for code chunks containing plots (thanks, @TimTaylor, #23). Providing alt text can improve the accessibility of images in HTML output. To avoid omitting the alt text inadvertently, you can set the option litedown.fig.alt in your .Rprofile.
Added the meta variable plain-title for HTML output, which is the plain version of the document title (i.e., without HTML tags), and used in the <title> tag.
Check boxes from - [ ] ... are no longer disabled in HTML output.
The implicit latest version of jsdelivr resources will be resolved to an explicit version, e.g., https://cdn.jsdelivr.net/npm/@xiee/utils/css/default.css will be resolved to https://cdn.jsdelivr.net/npm/@xiee/utils@X.Y.Z/css/default.css, where X.Y.Z is the current latest version. This will make sure the HTML output containing jsdelivr resources is stable.