style R Markdown (code chunks)

[lorenzwalthert]

It would be nice if styler could not only format plain .R files, but also .Rmd files.

Code Chunks

This could probably by implemented by changing transform_files, i.e. adding a function transform_lines_rmd_chunk. Main steps are:

• first read read_lines_enc
• look for regex ^´´´\\{r.*\\} *$ and ^``` *$
• transforming lines between these expressions while leaving everything else unchanged
• write back with write_lines_enc

@krlmlr maybe this functionality fits even in the utf8 package?

Plain Text

• Line breaks if text is longer than 80 characters
• punctuation (i.e. space after . , ? ! ect.)

The second point might be out of the scope of the package though...

[krlmlr]

Could we use knitr to parse the .Rmd file? The transform_lines_enc() function in utf8 seems generic enough.

Not sure about styling plain text, realistically we'd need to parse and write back Markdown.

[lorenzwalthert]

As far as I know, there is only the knitr code chunk option tidy that tidies up the output based on formatR, but I don't know how knitr could tidy the source.

[krlmlr]

knitr is not supposed to test it, just to do the extraction work for us (determine what is code, what language is that code in, what is text).

[jimhester]

We do this in lintr, https://github.com/jimhester/lintr/blob/aa70430282852afa15a61cc69771e71dd32e1aec/R/extract.R

[lorenzwalthert]

Thanks @jimhester for that hint. I just tried out the function and it indeed seems to me that we could use it to achieve our goal.
I know its a minor thing but I just thought about inline R code in Rmd documents. It would be nice if that could also be covered, but I think this is not the case with extract_r_code, right?
Also, what was the reason against using (internal) knitr functionality in the lintr package to do the extraction, as @krlmlr suggests?

@krlmlr I found the exported function knitr:pulr(text = character_vec) that returns all code lines (not only R) as a character vector, but I think this does not really help us. I just had a quick look at internal functions of knitr, and probably there are tools that could help us, but I did not really have time yet to explore that further.

[jimhester]

Inline R code is typically a single statement long, so likely does not have anything worth linting / styling. Supporting it would complicate the implementation for limited gain.

We do use internal knitr functionality https://github.com/jimhester/lintr/blob/aa70430282852afa15a61cc69771e71dd32e1aec/R/extract.R#L40-L41, to detect the code boundries. The reason you can't use purl et all directly is they transform the source code lines into a new script, so the line numbers do not match the original version. For the linting case we simply blank out the non-R lines and lint that file, you would have to do something more complicated for styler.

[lorenzwalthert]

I'd say closed with #233 for now. Styling markdown text is out of scope I think. The only thing one might consider is formatting to 80 characters (for non-code). We could integrate stylermd into styler (i.e. copying relevant code over). If you @jonmcalder are interested, you can have a look, otherwise we leave it for now.

[tvatter]

Question: how is this supposed to work?

I tried to style the file at https://gist.github.com/tvatter/25dbb789b64e7ba9484c65b5d4c4f828
using styler::style_file("sample.Rmd") and I got:

Status    Count   Legend 
✔         1       File unchanged. 
ℹ         0        File changed. 
✖         0       Styling threw an error. 

(and obviously the ugly code chunk was left untouched)

[tvatter]

Oh gosh, my bad.... the code is in text... Forget it :)