[Rd] Top level \Sexpr and R CMD check

Duncan Murdoch murdoch@dunc@n @ending from gm@il@com
Thu Jul 12 16:35:40 CEST 2018

On 12/07/2018 9:46 AM, Gábor Csárdi wrote:
> On Thu, Jul 12, 2018 at 2:30 PM Gábor Csárdi <csardi.gabor using gmail.com> wrote:
>> On Thu, Jul 12, 2018 at 2:21 PM Duncan Murdoch <murdoch.duncan using gmail.com> wrote:
>>> I think I found the bug.  The tools::checkRd function only processes
>>> \Sexpr's with "stage=render".  I think the author (who might have been
>>> me, I forget) assumed that would imply all the earlier stages as well,
>>> but apparently it doesn't.
>>> So you could use that as a workaround.
>>> I'll do some more checking, then submit a bug report and patch to Bugzilla.
>>> Duncan Murdoch
>> Thanks much! I tried using stage=render, but then I get an error at
>> install time:
>> Warning: /private/var/folders/59/0gkmw1yj2w7bf2dfc3jznv5w0000gn/T/RtmpCG4Qz9/R.INSTALLec4743ba8cf4/ps/man/ps_handle.Rd:45-48:
>> Section \Sexpr is unrecognized and will be dropped
>> And indeed the whole section is dropped.
>> Seems like there is no clean workaround here.
>> Thanks again,
>> G.
> Btw. would it make sense to just allow \Sexpr as a top level section?
> Maybe here:
> https://github.com/wch/r-source/blob/98e9999eb0e8616550632a1675e4d2dbe630d5e4/src/library/tools/R/RdConv2.R#L500-L503

I think the mental model is that \Sexpr is essentially a preprocessor 
directive, and should disappear before rendering.  It's the final result 
that needs to be checked.  So \Sexpr should be able to produce top level 
sections, but it shouldn't be considered to be one.

The reason checkRd has a stages argument is because it might be working 
on partially processed Rd objects, rather than the original .Rd file. 
Each of the stages should be processed exactly once, in order.

> At least if stage=render, there is no way to check if the returned
> value is always a valid top level section, anyway.

But you can check if it is sometimes, which is better than nothing. 
Most \Sexpr content is really simple, so this is probably good enough.


> If it is not a valid section (or \Sexpr returns some bad markup in
> general), then the user gets a render-time error,
> but with stage=render I guess one cannot do better.
> G.

More information about the R-devel mailing list