Request: knitr as pandoc filter?

[cysouw]

I love Knitr and Rmarkdown/Bookdown! However, in many situations I need knits only a few times in a long (scientific) text. In such situations I tend to use just Pandoc Markdown with filters and my text editor of choice. For such a workflow it would be great if it would be possible to reverse the workflow and use Knitr as a filter inside Pandoc. Is that already possible? And if not, would it be possible to make such a filter?

---

By filing an issue to this repo, I promise that

• [x] I have fully read the issue guide at https://yihui.org/issue/.
• [x] I have provided the necessary information about my issue.
  • If I'm asking a question, I have already asked it on Stack Overflow or RStudio Community, waited for at least 24 hours, and included a link to my question there.
  • If I'm filing a bug report, I have included a minimal, self-contained, and reproducible example, and have also included xfun::session_info('knitr'). I have upgraded all my packages to their latest versions (e.g., R, RStudio, and R packages), and also tried the development version: remotes::install_github('yihui/knitr').
  • If I have posted the same issue elsewhere, I have also mentioned it in this issue.
• [x] I have learned the Github Markdown syntax, and formatted my issue correctly.

I understand that my issue may be closed if I don't fulfill my promises.

[cderv]

That is an interesting idea. Can you share an example of what you have in mind ?

It is possible to have a Pandoc filter running any program. So it would be possible to have a pandoc filter execute R to process some part. I don't this would be included in knitr directly. It would be just a way to use knitr (or mimick knitr) in a Lua filter.

I know about this Lua filter to have "diagram chunk" and it is close from what knitr could do https://github.com/pandoc/lua-filters/tree/master/diagram-generator

Unfortunately I won't have time to look into it very soon, even if I find such challenges quite fun. However, if you are willing to try something for a Proof of Concept of what you want, we'll be happy to look into it and help !

Here are some ressources:

• What is a Lua filter : https://bookdown.org/yihui/rmarkdown-cookbook/lua-filters.html
• Documentation: https://pandoc.org/lua-filters.html
• Examples of filters: https://github.com/pandoc/lua-filters

And maybe someone already tried something. We'll see if anyone reach out here.

Thanks for the suggestion !

[cysouw]

There is also https://github.com/cdupont/R-pandoc by @cdupont to insert R plots (very nice!). However, it does not allow chunks to be evaluated and prepared as nicely as knitr does.

As an example of what I think of, take the minimal markdown examples from the manual, simply rename the .Rnw document as .md and then I would like to use something like:

pandoc knitr-minimal.md --to html --output knitr-minimal.html --filter knitr_filter

[cderv]

Thanks for sharing this project I wasn't aware of.

simply rename the .Rnw document as .md

I am not sure to follow. .Rnw is a latex file with some code chunk in it. I don't see why you would have this type of content in an .md file.

Also, if the aim is to have the exact same file in input as you would have with knitr why not use knitr directly ?

• Use knitr to process the R chunk in an initial .md document and produce another file .md as output with source code chunks, output code chunks, ... in a format Pandoc can use.
• Use this file to convert to html at the commandline using Pandoc

This is basically the workflow behind Rmarkdown (https://bookdown.org/yihui/rmarkdown-cookbook/rmarkdown-process.html) but it would allow to have very minimal impact as it just knit (even if R Markdown was designed so that this workflow goes well)

I am not sure to completely understand yet what would be the advantage of not using rmarkdown to have (knitr then Pandoc) and do the other way around Pandoc calling knitr on some part. I can however understand why R-pandoc to have so part of a .md file processed by Pandoc to be evaluated using R code.

I am surely missing something.

[cysouw]

sorry, I meant .Rmd not .Rnw ...

[cderv]

ok, this is what I thought. Thanks for confirming.

For a full existing .Rmd to .md, best is to use rmarkdown I think. It would be a full project to do all what knitr+rmarkdown are doing as a filter - but maybe possible. It need some thoughts (and time + work).

However, .md to .md can be done with knitr only if you have some code chunks you want to evaluate before running Pandoc.

Having a filter to run block identified as R code could be interesting too for users relying already on an existing Pandoc workflow.

This is an interesting discussion, Thank you!

[cysouw]

and yes, I can simply write a wrapper script around Pandoc when I use a file test.md

• rename it to .Rmd
• run knitr
• send the result to pandoc
• delete the new .md file
• change back the .Rmd file to .md

That is basically what I have done in the past.

However, Pandoc now has this nice defaults system in which to store all settings. It would be so much easier to add here a filter knitr and have all be done.

The problem is, filters work inside the processing of Pandoc, not outside as my workflow above illustrates.