Deprecating xfun::isFALSE() | /en/2023/02/xfun-isfalse/

[giscus[bot]]

Deprecating xfun::isFALSE()

Base R has had the function isTRUE() for many years, and I had wished for a few years that there were an isFALSE(), too. I had used identical(x, FALSE) many times before I added it to the first …

https://yihui.org/en/2023/02/xfun-isfalse/

[mingxin-zhang]

谢老师您好，感谢您在RMarkdorn等诸多领域的贡献！我是一名来自中国的学生，希望向您请教一个问题：我最近在写一个计量经济学领域的R package，我比较确定我所写的函数没有用到xfun::isFALSE，但是在我进行devtolls::check()时，总是会返回错误
Error(s) in re-building vignettes:
Error: processing vignette 'Introduction_to_JTZ.Rmd' failed with diagnostics:
The function xfun::isFALSE() will be deprecated in the future. Please consider using base::isFALSE(x) or identical(x, FALSE) instead.
我完全理解这个错误的含义，但是我实在搞不明白究竟是哪一个函数使用了xfun::isFALSE。
所以我想向您请教：是否有直接在所有函数中替换xfun::isFALSE的方式？

[yihui]

首先保证你所有的 R 包都已经升级至最新版本，如果已经升级所有包还是有这个错误，那么手工运行 rmarkdown::render("Introduction_to_JTZ.Rmd")，出错后看 traceback()，看是哪个函数调用了 xfun::isFALSE()。

[liuyanguu]

I got a similar issue and solved it by removing install.packages() or library() from the Rmd.

[liuyanguu]

Actually not solved. Running rmarkdown::render("my_vignette.Rmd") is fine, but the devtools::check() returns the same error.

[icasas]

Hello. I have the same problem. The worst is that I do not know what I should change of my .Rmd to not call xfun::isFALSE(). Any help?
Thanks a lot.

[yihui]

When in doubt, try to update software packages.

update.packages(ask = FALSE, checkBuilt = TRUE)

[matloff]

Same problem. Re-installed devtools, to no avail.

[matloff]

Right, but there may be an indirect call. Anyway, of course I tried traceback(), but e.g. calling it from devtools::check(), I get the message "No traceback available." Note: It MATTERS; I'm trying to submit a major package to CRAN.

[yihui]

Have you tried updating all packages as mentioned above?

[matloff]

I started doing so last night, but it wasn't done after an hour, and I needed to shut down the machine. (PC running Linux. All seems to work on my Mac.) I'm going to finish the updates today, will let you know if it helps.

[yihui]

I think the problem is from one of your outdated packages. Normally CRAN uses the latest versions of packages when checking your package, so this problem is unlikely to prevent you from passing CRAN's auto-check.

[matloff]

As always, I salute the CRAN reviewing team. They diligently perform a thankless task. But don't they have their own internal checks, beyond R CMD check -cran? As we all know, one's package can pass the later check with flying colors, only to have the CRAN team say, "We see that....," pointing to something wrong that didn't emerge in the above test?

[matloff]

Good news, Yihui! Updating the packages seems to have worked. At least the error message no longer appears, and "post hoc ergo propter hoc." :-) Thanks very much.

[yihui]

Good to know! You are welcome!

[matloff]

Unfortunately, there is a postscript. :-( With the "updated" files (reason for the quote marks below), I started getting error messages from R CMD check --as-cran "there is no package 'tufte'" regarding the vignettes. This on both Mac and Linux. But I also had a brand new Mac, and after installing everything, it ran fine, no error messages. So apparently "updating" had not been thorough, or something. Anyway, I built on the new machine, and submitted to CRAN. But sure enough, Uwe's message then informed me that the missing-tufte problem had occurred on both Windows and Debian. Were the CRAN machines also not updated?

Of course, I can remove the tufte stuff from the vignettes, but as you can understand :-) I hate to do so. Great package!

The package, BTW, is at github.com/matloff/qeML, if you are curious.

[yihui]

I don't think this is relevant to updating packages. R CMD check --as-cran is quite strict about package dependencies declared in your DESCRIPTION file. Basically any package that you use in your package (including vignettes) must be declared in DESCRIPTION. You didn't declare tufte: https://github.com/matloff/qeML/blob/133b1587/DESCRIPTION Therefore no matter which platform you use, R CME check --as-cran should signal the problem. I don't know what you ran exactly on the new Mac, but in theory, it should give you the same note.

You can look for _R_CHECK_SUGGESTS_ONLY_ in the R-ints manual for an official explanation.

[matloff]

Ah, so even vignettes' use of libraries must be stated in the DESCRIPTION file. I said I had "tried everything," but this never occurred to me. Thanks!

[kenahoo]

Hi @yihui, we hit this issue in our company's pipelines because we run R CMD check in our build processes for our locally-developed code. Those pipelines started failing with the error you mention.

This doesn't seem like a "deprecation", it's a backward incompatibility, because things that used to work no longer work, right? A deprecation would give people a period of time to switch off the implementation, emitting a warning (or not) in the meantime.

Aside from that - is there any real pressing need to deprecate/remove this? Instead of removing it, could it just be turned into a pass-through to base::isFALSE()? That would probably be a lot gentler to people's existing pipelines.

[kenahoo]

Thanks for your reply! It's not totally clear to me which code had the errant isFALSE() in it, but it wasn't the package being R CMD checked. The user whose pipeline failed reported to me that when he upgraded various packages the problem went away, and I think the culprit was probably either rmarkdown or knitr. Those are two of the packages he upgraded, at least. I haven't found the actual smoking gun, and the user's pipeline is now passing so he's not troubleshooting anymore (but there will likely be other users whose pipelines break in the same way).

Specifically, I think the fix I need to make now is to identify the package that was causing the call to isFALSE(), and rebuild the Docker image that my users use in their pipelines so it starts out with a version of that package that doesn't cause the problem. It might most likely be knitr, that was installed at version 1.39 in the base image. Does that seem like it might be correct?

[yihui]

when he upgraded various packages the problem went away

That's what I suggested in the beginning of this post---do a full update of packages. The most likely way to run into this problem is a partial update, e.g., updating xfun but not knitr. I have removed the use of xfun::isFALSE() in all packages that I maintain last year (e.g., yihui/knitr@ccb07c4), and no CRAN packages should be affected by the deprecation of xfun::isFALSE (this is guaranteed by checking reverse dependencies).

In short, if you are using the latest (or relatively recent) versions of CRAN packages, you shouldn't be affected by the deprecation of xfun::isFALSE.

[kenahoo]

In our setting, which is not just a bunch of individual workstations, but a set of deployed production applications, we generally need to control the deploy environment much more carefully than that, so we can't generally just "do a full update of packages". That's why I need to try to understand which specific pieces need to be updated, and do those in a controlled fashion.

[yihui]

If you desire the controlled fashion, I think using packages like renv would be a much better solution, otherwise it would be too difficult to know which versions of packages work or don't work with which versions of other packages.

[kenahoo]

Yes, we do use renv when feasible.

[laceto]

Hi, I cannot understand what is going in my package. if I run the cmd check from workbench's console I did not get any error while when running the check into the pipeline to create and pull the package to package manager I get the error. can you tell me why? thank you

[yihui]

Have you read the note in the beginning of this post? https://yihui.org/en/2023/02/xfun-isfalse/

[laceto]

yes, I've updated already all packages and checked that this funs is not used anywhere but the error persists.

[yihui]

I'm reasonably sure that certain packages are still outdated. Here is how you can check:

pkgs = tools::dependsOnPkgs('xfun', dependencies = 'all', recursive = FALSE)
vers = sapply(pkgs, function(p) as.character(packageVersion(p)))
vers

The results depend on the packages you installed in the system. Here are some of my package versions:

blogdown   bookdown      highr      knitr   markdown Rd2roxygen  rmarkdown    rticles      servr    tinytex 
  "1.18"     "0.36"     "0.10"     "1.45"     "1.11"     "1.14"     "2.25"     "0.25"     "0.27"     "0.48" 

You can run the above code in the system where you got the error, and see if the packages are truly up-to-date.

I don't know what your "pipeline" means or is, but please note that R can have multiple library paths (.libPaths()) and you could install multiple copies and different versions of a package into libraries with different paths. R CMD check will ignore library paths you set in .Rprofile but respect those set in .Renviron.

[yihui]

I just made a change in xfun so that you will know better which packages are outdated in the error message if you install the dev version:

remotes::install_github('yihui/xfun')