[R-pkg-devel] winUCRT failures

[Duncan Murdoch]

On 26/04/2021 9:20 a.m., Ott Toomet wrote:
> I think part of the problem is that the corresponding section in 
> "Writing R extensions" is not very clear for newcomers.  I admit I have 
> read it several times as I have been confused about the difference 
> between suggest, imports and depends, and so combed section 1.1.3 
> repeatedly.  But from this email thread I realize that I have overlooked 
> some of the basics.
> 
> I would like to have an authoritative document (preferably WRE) that 
> provides some sort of clear distinction between these concepts, 
> including a few examples.  In particular, the fact that one should wrap 
> the suggested packages in "requireNamespace" has escaped me.  I guess it 
> is not just me as a large chunk of package authors are applied 
> researchers and not professional software developers.
> 
> Not sure if one should rewrite the docs here, or maybe alternatively 
> create some sort of simplified version of the document.

There are tons of simplified versions of the document (in blogs, books, 
mailing list postings, etc.), but many of them also contain errors or 
make up different versions of the advice, so a rewrite or simplified 
version that had the support of R Core would be valuable.

Unfortunately, even if someone else wrote it, it would take a lot of R 
Core work to check it.  And if it was a separate simplified version, 
there'd be likely divergence over time from the full technical version.

"R CMD check" raising issues which beginners can ask about seems like a 
reasonable solution to this problem, as long as there are people willing 
to help them.  It offloads the writing work to the people who are giving 
the responses.  It's still hard for beginners to choose between 
conflicting versions of the advice, but that will always be true, even 
if one version is "endorsed by R Core".

Duncan Murdoch