Add lua_filters to pandoc_options to expose them in the output_format #1899
Conversation
|
If we export |
Yes this was idea, but this means we impose the directory structure. I am more inclined to export But, indeed we could start suggesting some standard. |
|
I see. I am fine with not exporting |
|
This is ready I think. I'll test it with bookdown now. |
2.3.9, the one after merging rstudio/rmarkdown#1899
* rstudio_origin/master: fix rstudio#1905: a chunk with the class "fold-hide" should only hide itself (rstudio#1906) bump version after merging rstudio#1899 lua filter -> Lua filter Add an argument `lua_filters` to pandoc_options() to store them in the output format object (rstudio#1899) recursve into the parent dir at the end also suppor dir input in proj_root() list.files() with full.names = TRUE (so the files could be read with corret paths), and pass down the `file`/`pattern` arguments in the recursion use our own proj_root() instead of introducing a dependency rename package_root() to proj_root() generalize the package_root() function to work with more types of projects Discover site generators in index.Rmd files in parent directories (rstudio#1898) only install pkgdown for website build (rstudio#1900) Generalize number sections (rstudio#1879) enable figure cropping in pdf_document() only when both pdfcrop and ghostscript are found (yihui/knitr#954), and do not limit this feature to non-Windows platforms add the citation entry for the R Markdown Cookbook add lua filter vignette to pkgdown website (rstudio#1894) close rstudio#1889: evaluate the output argument before changing working directory in pandoc_convert() (rstudio#1890)
Merge remote-tracking branch 'rstudio_origin/master' into css-slash-fix # By Yihui Xie (12) and others # Via GitHub * rstudio_origin/master: (22 commits) fix rstudio#1905: a chunk with the class "fold-hide" should only hide itself (rstudio#1906) bump version after merging rstudio#1899 lua filter -> Lua filter Add an argument `lua_filters` to pandoc_options() to store them in the output format object (rstudio#1899) recursve into the parent dir at the end also suppor dir input in proj_root() list.files() with full.names = TRUE (so the files could be read with corret paths), and pass down the `file`/`pattern` arguments in the recursion use our own proj_root() instead of introducing a dependency rename package_root() to proj_root() generalize the package_root() function to work with more types of projects Discover site generators in index.Rmd files in parent directories (rstudio#1898) only install pkgdown for website build (rstudio#1900) Generalize number sections (rstudio#1879) enable figure cropping in pdf_document() only when both pdfcrop and ghostscript are found (yihui/knitr#954), and do not limit this feature to non-Windows platforms add the citation entry for the R Markdown Cookbook add lua filter vignette to pkgdown website (rstudio#1894) close rstudio#1889: evaluate the output argument before changing working directory in pandoc_convert() (rstudio#1890) tweak news require Pandoc >= 1.14 xfun 0.16 is on CRAN now ... # Conflicts: # NEWS.md
2.3.9, the one after merging rstudio/rmarkdown#1899
This will partly resolves #1897 and this is needed for rstudio/bookdown#940 where we add a lua filter in bookdown.
Adding
lua_filterstopandoc_optionsallows exposes the filters in the output format and that will allow to manipulate them to change the order or even remove some if a format use another as its base.We need this in bookdown because we want to be sure to run the bookdown lua filter first.
Currenlty, the PR assumes lua filters are appended by default, i.e if a format depend on an other, the lua filter from the former will execute after the latter. We could consider to inverse this maybe: always run lua filter of a format before the filters of its base format.
With this change, the addition of the filters at the pandoc command line append in
renderdirectly and will be always put at the end of the command line.I have changed the helpers function for lua but they where not exposed, only accessible by
:::. I wonder if that is fine. I can do otherwise but it seems better this way.