Caching?

[larskotthoff]

It takes a long time to render the book -- why is caching turned off?

[mllg]

• If you run ./serve, only the Rmd with changes gets recompiled after the first compilation. Unfortunately, we have very large Rmd files because bookdown requires us to put every chapter in a single file.
• We could enable a knitr cache, I don't know how well this one works though.
• Travis should run without cache.

@pat-s Have you tried to enable the cache?

[mllg]

Actually, it might be possible to split the chapters into multiple files. We should give it a shot after all open PRs are merged.

[pat-s]

Actually, it might be possible to split the chapters into multiple files.

I tried for almost two hours around 2 weeks ago and failed.
Multiple reasons for it.
You can try again if you have free time but I would not recommend it.

Have you tried to enable the cache?

Yes, several times during the workshop. Simply setting knitr::opts_chunk$set(cache = T) does not work because you will face spurious errors that some objects were not created along the way.
I suspect that this is an R6 issue.

Currently I think that our setup with Michels ./serve is the most convenient one we can aim for. Only the Rmd with changed made is rerun in this case.

[mllg]

Maybe we can do a video call on this. I just tried to move a subsection into a separate file, and it seemed to compile just fine.

[mllg]

For the caching: Patrick is right, this is a R6 thing. knitr saves its cache with internal functions intended for lazy loading data, and these are not working together with active bindings :(

[mllg]

Here is the issue: r-lib/R6#152

[larskotthoff]

Ok, thanks for the clarification.

[berndbischl]

Fuck this issue and that this was closed in R6. This is a horrible language bug.