Closed eddelbuettel closed 2 years ago
I'm strongly in favor of this. Over the weekend I got the package up and running on cygwin at home and have some fresh wounds from knitr-related dependency hell.
Yep. As they "when you know, you know".
If you don't mind, let me try running it (later this evening) before approving.
By all means! I have done the same (doing the conversion here in code first, and then one package (to CRAN) at a time).
I also noticed our Travis setup is dead (as it is everywhere else). I have a working (and again, simpler / smaller / more minimal than other folks) setup in r-ci that I can bring easily to GitHub Actions. I'll file a separate ticket for that.
And if and when you approve (not jumping the gun, just thinking out loud) it may make sense to roll this to CRAN too as we did in fact get real bug fixes in in this last cycle.
Approved/merged. And yes to deploying a new version to CRAN.
This PR suggests to follow a practice I just introduced in six of my own packages to replace use of packages
knitr
,rmarkdown
and (for the CSS formatting)minidown
(which taken together have a bit of footprint) with the much simpler new packagesimplermarkdown
which only needs one other package (for JSON) (plus a tiny CSS file) and wrapspandoc
directly. See the reverse suggests of simplermarkdown for examples.Our vignette only used
pandoc
for code formatting and actually ran no R code (which is not uncommon) so the switch was trivial. I put a locally rendered html version of the vignette here if you want to peruse it.As for motivation, I find this (done via package
deepdep
) quite illustrative:All that said, this is of course not required as a change but merely reflects my usage preference and experience. We can easily leave things as they are for no harm. So if the majority vote is "no change" we can close this PR.