Error in ...names() : could not find function "...names"

[MikkoVihtakari]

While trying to compile a model on Daniel's Windows machine (R 4.0) we encountered following issue:

g3_stock("test", lengthgroups = 1:5) %>% g3s_livesonareas(areas[c('1')])
Error in ...names() : could not find function "...names"

...names() was apparently introduced by R 4.1. I don't see that this function would have been used in g3s_livesonareas(), right?

Is it perhaps used by the most recent version of rlang? Or some other tidyverse packages? Can we get rid of those dependencies somehow? Or am I lost with my reasoning?

[lentinj]

It's in gadget3 itself:

https://github.com/gadget-framework/gadget3/blob/4420ddca38f342f247a8a2b87f494b335aae113d/R/formula_utils.R#L41-L43

https://github.com/gadget-framework/gadget3/blob/4420ddca38f342f247a8a2b87f494b335aae113d/R/run.R#L65

...in both cases, rewriting wouldn't be too much of a bother. This is being picked up by "oldrel-1", yes?

[MikkoVihtakari]

...in both cases, rewriting wouldn't be too much of a bother. This is being picked up by "oldrel-1", yes?

No, on a Windows machine with R 4.0.

[lentinj]

oldrel-2 would be 4.0, and pick this up. Seems like a reasonable compromise in what we support.

Tidyverse policy is oldrel-4. This seems a bit extreme for our purposes though. I'd like to get to 4.1 being the minimum version in the not too distant future, so we can use the native pipe in examples.

Regardless, should update the CI to check this, and add something into the rversions API and auto-update the DESCRIPTION, both here and MFDB.

[lentinj]

We have tests now with oldrel-1 and oldrel-2, which pass. As well as the ..names() problems there were scattered instances of native-pipe, which was predictable.

The release process will update the DESCRIPTION with a new minimum version of R if required, so we should be good here.