highlight Rd code blocks according to their language #1700
Conversation
|
|
||
| non_r_blocks <- xml2::xml_find_all( | ||
| html, | ||
| "//div[contains(@class, 'sourceCode') and not(contains(@class, 'sourceCode r'))]" |
There was a problem hiding this comment.
the XPath query is different because we need to keep the div until we extract the language.
| highlight_other_block <- function(block) { | ||
| lang <- sub("sourceCode ", "", xml2::xml_attr(block, "class")) | ||
| code <- xml2::xml_text(xml2::xml_find_first(block, "pre/code")) | ||
| highlighted <- markdown_text( |
There was a problem hiding this comment.
a new use case for markdown_text() 😁 In that function pkg is needed for finding external links but in code we do not need to try and identify them I think (and passing pkg all the way to here would warrant more refactoring).
| pkg = NULL | ||
| ) | ||
| code_node <- xml2::xml_find_first(block, "pre/code") | ||
| xml2::xml_text(code_node) <- paste( |
There was a problem hiding this comment.
passing stuff via the text is clunky because it gets escaped. But as the Rd transformation creates HTML as text, I couldn't find a better idea, and unescaped HTML at the very end of the code I added.
| xml2::xml_text(block), | ||
| classes = downlit::classes_pandoc() | ||
| ) | ||
| if (!is.na(out)) { |
There was a problem hiding this comment.
The case where out is NA probably only happens when the code is not valid R code (e.g. a YAML block with no language information) but I suppose it could be some sort of pseudo-code, so hard to warn?
There was a problem hiding this comment.
Or there could be a warning, and the workaround for non-valid pseudo-code would be to use
```pseudor
bla
```
| @@ -74,4 +74,5 @@ Roxygen: list(markdown = TRUE) | |||
| RoxygenNote: 7.1.1.9001 | |||
| SystemRequirements: pandoc | |||
| Remotes: | |||
| r-lib/downlit | |||
| r-lib/downlit, | |||
| r-lib/roxygen2 | |||
There was a problem hiding this comment.
Depends on r-lib/roxygen2#1229
There was a problem hiding this comment.
Maybe I should make it more general. 🤔
|
Why is this pkgdown's responsibility and not downlit's? |
|
You mean pkgdown could pass the code, and code language information, to downlit? And downlit would use Pandoc via rmarkdown for language other than R/undefined? |
|
I think this would be cleaner if the Rd parsing code simply left preformatted code blocks alone, and then we did this syntax highlighting when we tweak the generated html. I don't think we want xml munging code in the Rd parsing code. |
|
This is an excellent idea! I opened a new PR that feels much less clunky! |
Fix #1690
Fix #1692
It feels a bit clunky for now, but the result looks much better. 😅