Should purl() wrap try({}) blocks around error=true chunks ?

[jeroen]

Running R CMD check --no-build-vignettes on a package which has a vignette with a '''{r error=TRUE} block results in a check error because R script that was purl'ed from the vignette, does not account for the error=TRUE.

For example, if we check the V8 package with --no-build-vignettes I get:

R CMD check --no-build-vignettes V8_5.0.1.tar.gz
#  ....
#  * checking package vignettes ... OK
#  * checking running R code from vignettes ...
#    ‘npm.Rmd’ using ‘UTF-8’... OK
#    ‘v8_intro.Rmd’ using ‘UTF-8’... failed
#   ERROR

Because CRAN now published the purled script on the site, we can easily see the problem:

• https://cran.r-project.org/web/packages/V8/vignettes/npm.Rmd
• https://cran.r-project.org/web/packages/V8/vignettes/v8_intro.R <-- purled script by cran

Note that the Rmd contains several eval=FALSE and error=TRUE chunks. The eval=FALSE chunks are commented out in the output script, which makes sense. However the error=TRUE chunk is run as-is, and hence script errors.

I think the chunk that contains error=TRUE should be wrapped in try({....}) in the purled script. The default behavior of try is silent=FALSE so it results in the identical behavior as when building the vignette: display the error and proceed.

[yihui]

Yes, I think that's a good idea. Will do.