No interpolation on dayofyear

[coxipi]

Pull Request Checklist:

• This PR addresses an already opened issue (for bug fixes / features)
  • This PR fixes #xyz
• Tests for the changes have been added (for bug fixes / features)
  • (If applicable) Documentation has been added / updated (for bug fixes / features)
• CHANGELOG.rst has been updated (with summary of main changes)
  • Link to issue (:issue:number) and pull request (:pull:number) has been added

What kind of change does this PR introduce?

• During the adjusting part of quantile methods, a interpolation on the grouping dimension is done if the interpolation "interp" is chosen. This should not be the case when `group="time.dayofyear", so this option is now blocked

Does this PR introduce a breaking change?

No? Even if interpolation was applied before, it should have done nothing, just a more complicated algorithm for nothing

Other information:

See the discussion here #2019 (comment)

[aulemahal]

That makes sense! Have you seen a change in performance ?

[coxipi]

That makes sense! Have you seen a change in performance ?

good question, let me see ... darn, wish I had made that benchmark suite...

[coxipi]

This would not work the way I implemented in the end. One way to handle would be to do a similar implementation SPI/SPEI indices for the case of dayofyear? broadcast the dayofyear params, etc.

[github-actions]

Note

This Pull Request modifies the SDBA module.
The changes made here have been approved. Be sure to port any relevant changes to xsdba.
[!NOTE]
This Pull Request modifies the SDBA module.
The changes made here have been approved. Be sure to port any relevant changes to xsdba.

[coxipi]

Don't know how to transfer this to xsdba? I can transfer an issue but not a PR? I will simply close it and open a new issue in xsdba. Maybe. It's not really a priority. It would be nice to avoid useless 2d interpolation when possible though, probably it might save some time on DQM computations. This would not affect MBCn as it is currently, and that's the big bottleneck computation-wise.