Slash-normalize PrefixRecord.files for conda list
#14523
Conversation
conda-bot
added
the
cla-signed
CodSpeed Performance ReportMerging #14523 will not alter performanceComparing Summary
|
| anchor_paths = [] | ||
| for fpath in prefix_record.files: | ||
| if on_win: | ||
| fpath = fpath.replace("\\", "/") | ||
| if matcher(fpath): | ||
| anchor_paths.append(fpath) |
There was a problem hiding this comment.
We could fix this earlier in the PrefixRecord instantiation, but then we would pay for this conversion every time we load any record. However we only need this fix for conda list, which is the only one that uses PrefixData with pip_interop_enabled=True. This function in only called in that code path.
|
@conda/conda-maintainers, ready for review. |
There was a problem hiding this comment.
This doesn't have an issue assigned, please open one for record keeping (and user discovery).
On the specific change, this looks a little weird as a fix to work around a behavior that only happens when used with micromamba. I do think that in case another conda-like system uses the same files internally, it should follow conda's pattern of encoding data, like file paths.
I haven't done the research why this uses forward slashes, and why micromamba doesn't. But either way, this smells like something to fix in the path actions, instead of just superficially when running conda list.
|
Oh, also, this kind of needs tests it seems? |
Description
Fixes conda-forge/conda-forge-ci-setup-feedstock#362
Checklist - did you ...
newsdirectory (using the template) for the next release's release notes?