Make array equal work with masked Dask arrays and add tests#6325
Make array equal work with masked Dask arrays and add tests#6325pp-mo merged 17 commits intoSciTools:mainfrom
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #6325 +/- ##
=======================================
Coverage 89.79% 89.80%
=======================================
Files 90 90
Lines 23554 23576 +22
Branches 4391 4398 +7
=======================================
+ Hits 21150 21172 +22
Misses 1662 1662
Partials 742 742 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
⏱️ Performance Benchmark Report: 7408a83Performance shiftsFull benchmark resultsGenerated by GHA run |
⏱️ Performance Benchmark Report: f30aca9Performance shiftsFull benchmark resultsGenerated by GHA run |
|
I suspect the benchmark results are worse because the masks of Dask arrays are now compared and previously were not compared. |
⏱️ Performance Benchmark Report: b341799Performance shiftsFull benchmark resultsGenerated by GHA run |
|
Hmm looks like something subtle has gone wrong here. Assuming we can fix, I am still minded to get this into Iris 3.12, Performance-wise, this looks great, but there is a problem .. See #6349 for my crude assessment of the "combination" with #6339. |
⏱️ Performance Benchmark Report: 09805c5Performance shiftsFull benchmark resultsGenerated by GHA run |
⏱️ Performance Benchmark Report: 52b3f8cPerformance shiftsFull benchmark resultsGenerated by GHA run |
|
@pp-mo Thanks for the extensive analysis. I think that with the latest commits, the implementation here is now both correct and slightly faster than the original. See the latest benchmark run here #6349 (comment). I've also done a test comparing two ~50GB arrays on my laptop and that works fine too. |
pp-mo
left a comment
There was a problem hiding this comment.
One tiny tweak to make, otherwise all good now I think
Thanks for your diligence on this @bouweandela Annoying I just found a small thing to change, |
Also consider non-floating point arrays equal with withnans=False
⏱️ Performance Benchmark Report: 1cacdd6Performance shiftsFull benchmark resultsGenerated by GHA run |
pp-mo
left a comment
There was a problem hiding this comment.
Now all OK
Thanks again for all your patience @bouweandela !
🚀 Pull Request
Description
Make array equal work with masked Dask arrays and add tests.
Closes #6188
Consult Iris pull request check list
Add any of the below labels to trigger actions on this PR: