- Add support and validation for
cautionfenced divs introduced in carpentries/varnish#149 (@froggleston #160)
- Hotfix for {tinkr} upcoming bump to 0.3.0 that changes yarn$show() arguments: ropensci/tinkr#108. (reported @froggleston #156, fixed: @froggleston, #157)
- Remove support and checks for Windows R 3.6
- Add support for tabbed content in lessons
(reported: @astroDimitrios,
implemented: @astroDimitrios, @froggleston,
#148
carpentries/sandpaper#571, carpentries/varnish#121, ).
- Hotfix for R 4.3.3 bump that results in trailing commas in glue function causing argument errors: tidyverse/glue#320 (reported: @ns-rse, #149; fixed: @Bisaloo, #150). Required for sandpaper CSS fix
- New
vignette("intro-lesson", package = "pegboard")provides information about working withpegboard::Lessonobjects. - Documentation for
EpisodeandLessonobjects have been updated to be a bit more descriptive and point to the vignettes.
Episode$unblock()method gains theforceargument which will allow the method to run even if it has previously been run.
- Snapshot update for testthat 3.2.0
validate_links()now differentiates between links and images in reporting (reported: @joelnitta, #143; fixed: @zkamvar, #144)
- Two vignettes added for XML and Episode class
- Documentation for validation reporting added
- child chunk options that would fail out of context no longer cause a failure. (reported: @trhallam, carpentries/workbench#74, #139; fixed: @zkamvar, #140)
This release introduces automated processing of {knitr} child files, which enables them to be automatically available for validation and processing.
Episodeclass objects gain thechildren,parents, andbuild_parentsfields which will contain the known relationships between files. Bothchildrenandparentsrepresent immediate relationships whilebuild_parentsrepresent the most distant ancestor of the object that would trigger it to build. Note that theparentsfields are only populated for anEpisodein the context of aLessonobject.
Lessonclass objects gain thechildrenfield which is a list that storesEpisodeobjects derived from child files.Episodeobjects gain the$has_parentsactive binding reporting if the object has a parent object. This is only used in the context of aLesson.- Both
EpisodeandLessonobjects gain the$has_childrenactive binding, reporting if there are any child episodes in the lesson or episode. Lessonobjects gain the$trace_lineage()method to return the absolute paths to all child files from the source path of a given episode if they exist.make_link_table()now has a new column calledparentswhich contains a list column of the parents (if any) of the files checked$validate_links(),$validate_divs(), and$validate_headings()will now all process child files by default. Note that$validate_links()will report missing files relative to the build parents.
- The
$load_built()method will now load all built markdown content, not just the episodes.
$validate_divs()method now recognises thespoilerclass of fenced divs which allow optional/expandable items that are not automatically shown to the learner (implemented: @tobyhodges, #134)
- Workshop overview pages are now able to be processed into
Lessonobjects. For Jekyll- based lessons, the directory name must end with-workshop(as is the standard for workshop overview lessons without episodes), for sandpaper- based lessons, theconfig.yamlfile must contain theoverview: truetag (reported: @zkamvar, #118; fixed: @zkamvar, #132, reviewed by @klbarnes20). These pages are indicated by the$overviewfield in theLessonobject. For lessons that are not workshop overview lessons, nothing will change.
- Lessons that underwent an incomplete conversion from Jekyll to sandpaper will now have a more appropriate error message provided.
$validate_links()no longer throws an error when there are HTML images embedded in comments (reported: @beastyblacksmith, #130; fixed: @zkamvar, #131, reviewed by @ErinBecker)- (transition)
$move_objectives()and$move_questions()methods no longer place these blocks as the second element in the markdown. This was originally implemented when we thought {dovetail} would be our solution to the lesson infrastructure (and thus needed a setup chunk before any blocks). - (transition) liquid-formatted links with markdown inside them are now parsed correctly. This leads lessons to be more accurately transitioned to use {sandpaper} (reported: @uschille, carpentries/lesson-transition#46; fixed: @zkamvar, #121)
- (transition) images with kramdown attributes that are on a new line are now more accurately transitioned to use {sandpaper} (reported: @uschille, carpentries/lesson-transition#46; fixed: @zkamvar, #121)
- A failing test due to the workbench transition was fixed (reported: @zkamvar, #125; fixed: @zkamvar, #127)
- The README file is no longer run through validation as it is not generally a part of the website and more often than not, creates distractions.
- Fenced divs with attributes are now properly parsed (@zkamvar, #115).
Lessonobject validators now validate non-episode files (reported: @zkamvar #110; fixed: @zkamvar #113).
$validate_links()will now respect links to anchors in spans.
- validators will no longer truncate on GitHub actions (reported: @zkamvar #111, fixed: @zkamvar: #114).
- validators will provide full context for invalid links (not just link or text, but link with text).
{tinkr}'s minimum version has been set to 0.2.0 to recognise the release to CRAN and bring in new bugfixes.
$validate_links()now checks if the URL protocol in an external link matches a known list of protocols. Those that do not match (e.g.javascript:andbitcoin:) will be flagged. (@zkamvar #109)
- A bug where attributes following an image would cause missing alt text to not be reported was fixed. (discovered: @dpshelio and @karenword; reported: @zkamvar #106; fixed: @zkamvar #108). This fix also makes the alt text parsing and validation more robust
- A bug where an unknown protocol was not recognised as invalid was fixed. (discovered: @ndporter; reported: @zkamvar, #107; fixed: @zkamvar, #109)
- New internal function
find_between_nodes()will get all nodes between two sibling nodes.
- A test that depended on an upstream resource was once again fixed.
- Tests were modified to account for a new case for image fixing.
- Internal
fix_links()function has improved documentation.
- A test that dependend on an upstream resource was fixed.
- The {tinkr} minimum version has been upgraded to 0.1.0.9000 to address the square bracket protection implemented in that version.
- Jekyll links with spaces are once again processed correctly. They were broken in the update to {tinkr} (reported: @zkamvar, #100; fixed: @zkamvar, #102)
- GitHub workflows have been updated to run weekly.
- Soft dependencty {cli} has been recommended at 0.3.4 to prevent reporting issues (see #97)
- Soft dependencies ggraph, ggplot2, and tidygraph have been removed. These dependencies were only needed for producing a now-out-of-date survey document
In preparation for {tinkr} 0.1.0, which changes the path of the default
stylesheet, we are using the tinkr::stylesheet() convenience function to
access it.
$summary()method which can summarise counts of elements in the episode.- fixes for
$errorand$outputactive bindings - new
$warningactive binding that will show code blocks with thewarningclass.
- new public field "built" that will contain XML representations of markdown files built from RMarkdown files in sandpaper lessons.
- new public field "sandpaper" is a boolean that indicates if a lesson can be built with sandpaper.
- new
$load_built()method will load the built files if they exist in a sandpaper lesson. - new
$get()method which will get any element from any Episode class object contained within. - new
$summary()method which will call the$summary()method for any Episode class object.
muffle_messages()is an internal function that will muffle any messages that originate from the {cli} or {pegboard} packages.- If the {cli} package is not available, messages will have the class of
pbMessage, which will allow end users/package authors to catch and manipulate any messages that originate from {pegboard}
make_pandoc_alt()(an internal converter function) will no longer create alt text from a caption if it contains a URL. This messes with the downstream validation of image links.fix_sandpaper_links()now also fixes links that use{{ site.baseurl }}.
fix_links()now processes links in headers and links with unescaped ampersands- internal function
text_to_links()now processes unescaped ampersands - internal function
find_lesson_links()no longer expects links to be strictly in paragraph elements.
- internal function
validate_links()will no longer flagalt=""as errors. These indicate decorative images. That being said, these should be rare in our lessons, but this is here just in case it's needed. Source: https://webaim.org/techniques/alttext/#decorative
- The {fs} package needs to be >= 1.5.0 (#83, @sstevens2)
- footnotes with no trailing newline are no longer accidentally appended with
relative link anchors when
getOption('sandpaper.links')is not NULL.
- If
getOption("sandpaper.links")is not NULL (in the context of a {sandpaper} lesson) and is a valid file, it will be appended to any file read in viaEpisode$new()
$validate_links()no longer throws warnings about short or uninformative text for link anchors (@zkamvar, #81)
- The inline messages for link validation errors are more verbose (@tobyhodges, #79)
validate_divs()will validate that the divs in an Episode are ones we expect.
- Correct mis-attribution for LICENSE
This is a soft release of {pegboard} to coincide with the first announcement of The Carpentries Workbench.
- {lifecycle} is no longer an imported package. We were not using it, so it makes more sense to leave it out in production.
get_list_block()will now select the last block if there are multiple "keypoints" blocks (@zkamvar, #75)get_list_block()will now throw a warning if a block does not contain any list elements (@zkamvar, #74)
get_stylesheet()now escapes spaces and normalizes the windows path to the tinkr stylesheet before embedding it (@zkamvar, #72)
- the
$move_*()methods no longer create a redundant heading in the fenced div (@zkamvar, #26)
- The pkgdown site has been updated to be more complete.
- the div family of functions uses the roxygen family tag to be more complete.
- The
$use_sandpaper()method for Episode objects will now remove "root" and "layout" yaml directives (@zkamvar, #68)
Images that had attributes added are now post-processed in use_sandpaper() and
will retain their original attributes.
- replace
relative_root_pathwith nothing instead on., which fixes a bug introduced with 0.0.0.9028.
- fix for jekyll-based lessons using
base_path.htmlto definerelative_root_pathare now corrected to no longer include those directives for links.
- fix in jekyll-based lesson auto-detection of RMD lessons will not error for pure Rmd lessons
- Jekyll-based lessons will now auto-detect and read in R Markdown content if
it exists. The
rmdflag for the Lesson initializer now does nothing.
- We now use the
_config.ymlfile to parse site-specific liquid template links to fix #60
- RMarkdown episodes that had a setup chunk without specifying
includesare now considered to have valid setup chunks and can have those chunks converted.
Episode$handout()will create trimmed-down R Markdown document with only challenge blocks and code chunks withpurl = TRUE, that can be passed toknitr::purl()for processing into an R code handout.Lesson$handout()will create a concatenated version ofEpisode$handout().
- validation messages have been revamped to be more consistent across messages.
- All validation methods in
EpisodeandLessonnow return data frames that contain detailed information for each element and what tests were passed and what were failed for downstream analysis. Importantly, they all will contain a column called "node", which points to the exact XML node containing the link/image/heading for inspection/manipulation. - validation reporting is no longer grouped by error
- link validation information has been switched to show the error message and then the problematic aspect/fix
- heading validation now works on continuous integration
- these functions no longer rely on dplyr being installed
- The
Episodeclass now gains the$confirm_sandpaper()method to bypass the assumption that all Episodes start as kramdown-formatted documents and will attempt to label divs in the episode (with a warning if there is no success). - The
Lessonclass will now run the$confirm_sandpaper()method for all markdown files ifjekyll = FALSE. Lesson$new()will now default to the current working directory.
- The internal
get_list_block()will no longer auto-label divs.
- Link validation now checks for more general uninformative link text and empty links (@zkamvar, #49)
make_link_table()will treat linebreaks in link text as a space character.
- The
Lessonclass now has the$validate_links()and$validate_headings()methods. (@zkamvar, #48)
NOTE: All of these are from (@zkamvar, #44)
- The Episode class now has the
$linksand$imagesactive bindings that extracts the links and images (markdown and HTML) from the document. make_link_table()creates a table of links parsed viaxml2::url_parse()with additional information about caption and alternative text (for images).- The Episode class now has the
$validate_links()method, which will validate links and images for common errors such as not using https and unresolved relative links. Episode$use_sandpaper()now converts images to use alt text over captions. Images that hadare converted to{alt='alt'}because pandoc uses everything in square brackets to be caption text. NOTE: this now makes a copy of the XML document.
Episode$new()gains the argumentfix_liquid, which fixes liquid variables in relative links before they are passed to {tinkr} (#46)- Post-processed links via
fix_links()retains theirsourceposattribute (fixed in e5508cc9c9a3821381293bdac12647edfbc0608e). Episode$lessonno longer assumes that the episode is inside a sub-folder (fixed in 63432ef83ecc41a6aab53fe768e8eaec107278d5).
Episode$validate_headings()now properly displays duplicated headings. (@zkamvar, #45)
- The Lesson class now respects the order of the contents in
config.yaml(#42 via #43)
- the Episode class now has the
$headingsactive binding and the$validate_headings()method to validate headings (#23 via #41) - We now use {testthat} 3e for our tests
- Cloning an Episode object is now more reliable
- the Episode class is now a sub-class of the tinkr::yarn class and gains the
show(),head(),tail(), andprotect_math()methods.
- Innocent block quotes that have not been sullied by the ruthless kramdown
postfix operators are kept as block quotes instead of failing on
$unblock()conversion.
- The omnipresent
{% include links.md %}is now removed on sandpaper conversion.
- Lesson class will now work with {sandpaper} (#24) with a new parameter
jekyll - Episode class gains new slot called
extrasto handle the sandpaper non-episodic things - A better error message is thrown with
Episode$label_divs().
- Swap {git2r} dependency for {gert}, which has a smoother interface and matches with the dependencies of {sandpaper}.
- Missing questions, objectives, or keypoints will no longer fail with a cryptic error. An informative warning will be thrown and an empty character vector will be returned. This addresses an issue in {sandpaper}: carpentries/sandpaper#79
- The URI for pegboard tags is now "http://carpentries.org/pegboard/", which fixes #18
- Several Bug fixes, see #21 for details.
- Travis hopefully banished
$label_divs()no longer modifies the fenced divs.@dtaglabels attached tohtml_blockandparagraphelements are now replaced by<dtag>elements that live within a custom namespace called "pegboard". This allows us to avoid manipulating the document paragraph structure in the case of fenced divs.$get_divs()now includes the div tags/fences in the output.- Internally, namespace handling has gotten marginally better where the
default namespace prefix is modified to
md:. - fenced divs are no longer manipulated on labelling.
- A new test expectation,
expect_moved_yaml()tests that a yaml element was successfully moved to the body of the document.
$keypointsand$objectivesnow are available and act like$questions- The
$move_*()methods will now add an h2 header to the block - The
$move_*()methods will use pandoc syntax instead of html div blocks. - Internal changes to creation of elements no longer relies on complex namespace juggling
$label_divs()method now will label any div tags in the episode$move_*methods will now auto-label div tags$questionsfield now returns the questions block or yaml header as a character vector.
- pandoc-style fenced divs are now processed the same as native div tags
@dtaglabels are now in the formatdiv-{n}-{class}where {n} is the sequential number in the document and {class} is the type of div.- labelling of
@dtagsis more straightforward and will label the tags sequentially.
The changes in this version largely are enhancements for handling div tags and conversion. See #9 for details
$unblock()defaults to converting to div tags unless$use_sandpaper()has been called.$unblock()will auto-name ALL the divs with {class}-div-{number}$move_*functions will now name the html_blocks$get_divs()returns div tags in a named list$challengeswill now find either blocks, divs or code, depending on the mutations$soltuionssame as challenges (see above)$use_dovetail()will warn the user if the body is empty$use_sandpaper()will do the same
Jekyll-specific and relative links are now converted as part of use_sandpaper().
fix_sandpaper_links()will fix relative paths and jekyll-specific links inside of lessons that have not yet otherwise been converted.
This version introduces conversions that work together and can be chained to convert an episode from the old Jekyll style to the new {sandpaper} style. Things are still very much in development though.
- conversions for sandpaper and dovetail are clearly separated
- the Episode class understands what conversions have been performed with a private logical vector that tracks changes.
- pkgdown site is now built automatically from github actions
use_dovetail()inserts a setup chunk at the top of the fileuse_sandpaper()converts chunks from their liquid/kramdown syntax to the commonmark or RMD syntaxmove_*()will generate a dovetail block or just a plain div block depending on whether or notuse_dovetail()has been called.remove_output()does what it says on the tinremove_error()removes error code blocks$outputand$errorcan now grab output and error chunks that were converted viause_sandpaper()
- Added a
NEWS.mdfile to track changes to the package. - Dependencies on {yaml} and {commonmark} now explicit
- Added methods to Episode class to move questions, objectives, and keypoints out of the YAML and into the lesson body.