Skip to content

Conversation

@gdementen
Copy link
Contributor

No description provided.

purposefully not including this in the changelog because I might deprecate that syntax in 0.36 (see larray-project#1142)
should also be faster for object dtypes and when rtol=0 and atol=0 and nans_equal is True
…ct#895)

considered a fix because it sort of worked but silently ignored the index and name of the series
the fix I will commit next will make it more likely users see them, so we might as well make it more useful
it slowed things down in the usual case and messed some error messages
…sage (fixes larray-project#1134)

* the code to compute the error message was itself raising an exception when the array key had ndim > 1
* mention the bad key parts if not too large
* avoid useless copy of the main array when display_progress is False
* avoid computing the sum on the last axis at last iteration
* also added a special case for ndim == 1, but that is just to
  avoid breaking that corner case by the other two changes
  rather than making it more efficient given that it will
  probably never be used

Overall, this makes a ~7% speed improvement for tiny test arrays.
No idea for larger arrays.
instead of NotImplementedError
since I think (hope), nobody uses this, go directly to the exception step instead of a warning
…#501)

in the process had to implement Axis.align(*axes), AxisCollection.align(*axis_collections) and Array.align(*arrays)

also deprecated passing non (arrays or scalar) to Array.align and check that each aligned axis is either anonymous or have the same name than others
@gdementen gdementen merged commit 635b948 into larray-project:master Nov 28, 2025
5 checks passed
@gdementen gdementen deleted the 035issues branch November 28, 2025 17:02
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant