renv::load(other_project) fails if autoload is disabled

[klmr]

In my environment I have intentionally set options(renv.config.autoloader.enabled = FALSE). In principle this works well.

However, it also breaks renv::load(other_project), because renv:::renv_load_switch(), which is called by renv::load(), sources the other project’s renv/activate.R file, which aborts if auto-loading is disabled.

As a user, this is unexpected: explicitly calling load() is the opposite of auto-loading. I’m wondering whether that’s an oversight (I am guessing that few people disable auto-loading, so that option is probably not as battle-tested) — or whether this is working as intended: in that case, is there a way of explicitly loading/switching to another renv project while keeping auto-loading disabled?

A workaround would be to temporarily unset the renv.config.autoloader.enabled option, but that requires awareness of the semantics.

[kevinushey]

This definitely seems like an oversight; I'll take a look.

[kevinushey]

Should be addressed via 2c51854; please let me know if this still appears to be an issue.

[klmr]

(I am super confused by the version numbers… I wanted to try out this commit but when comparing version numbers I noticed that this is apparently v1.0.7.9000, whereas I currently have v1.0.11 installed from CRAN. Are y’all counting backwards? 😆)

[kevinushey]

Shouldn't be that way; the current version on main is 1.0.11.9000:

renv/DESCRIPTION

Line 4 in e7e1f4b

Version: 1.0.11.9000

How did you install the development version of renv? Is an older version somehow getting installed here?

[kevinushey]

(The associated commit fixing the issue was actually merged some time ago; perhaps that's why?)

[klmr]

(The associated commit fixing the issue was actually merged some time ago; perhaps that's why?)

Ah I hadn’t noticed that! So the relevant change 2c51854 is already included in the version 1.0.11 that’s on CRAN. — I can’t quite reconstruct what happened since the bug was not yet fixed when I tested it yesterday before applying the patch, and packageVersion('renv') told me I was running 1.0.11.

However, after some back and forth, first installing the specific rstudio/renv@2c51854, and now also when (again?) installing renv@1.0.11 from CRAN, the issue is indeed fixed. Thank you very much. 🎉