issue_comments: 1297143372

This data as json

html_url issue_url id node_id user created_at updated_at author_association body reactions performed_via_github_app issue
https://github.com/pydata/xarray/pull/7204#issuecomment-1297143372 https://api.github.com/repos/pydata/xarray/issues/7204 1297143372 IC_kwDOAMm_X85NUNZM 14808389 2022-10-31T14:05:04Z 2022-10-31T14:05:04Z MEMBER

I'm a bit split on this.

As far as I understand it, there were two main reasons for the switch of dask / distributed: the fact that they had imports like python from .....scheduler import something (which I agree is utter madness), and because they had modules named core, so the import was a bit ambiguous to read. I don't think either apply to us: our directory structure is sufficiently flat that we get at most two levels (from .. import something), and pretty unique file names.

We're also not in conflict with PEP8: if I read that correctly, it does say they recommend absolute imports, but then it also says that explicit relative imports are fine as well (and not just in the case where a complex structure makes explicit relative imports much more readable).

This means that unlike for dask one style does not have a clear advantage over the other, and this becomes a matter of preference.

As such, my vote would be +0.5 for automated consistency, but -0.5 for absolute imports everywhere because for me explicit relative imports are actually a bit easier to read. However, I very much agree with using absolute imports of modules from the main package in tests (I think we have something like from ..core.duck_array_ops import something in one of the tests, which seems a bit weird?)

 
{
    "total_count": 0,
    "+1": 0,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 0
}
   1420242462