Conversation
43a2fab to
e125570
Compare
lib/iris/fileformats/netcdf/saver.py
Outdated
| # NOTE: the goal is to process the to-be-saved blockwise via a delayed | ||
| # operation that will return (is_masked, contains_value) for each block. | ||
| # These block-wise results are then OR-ed for each source. | ||
| # N.B. `da.map_blocks` apparently will not do this, at least not without | ||
| # `drop_axis` which seems to trigger loading of the whole source array. | ||
| # TODO: It probably *can* be done with `da.blockwise`, but I have so far | ||
| # failed to grasp how you would do that. |
There was a problem hiding this comment.
This looks like a reduction, maybe that could be useful?
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #5807 +/- ##
=======================================
Coverage 89.74% 89.75%
=======================================
Files 92 92
Lines 22941 22959 +18
Branches 5463 5471 +8
=======================================
+ Hits 20589 20607 +18
Misses 1620 1620
Partials 732 732 ☔ View full report in Codecov by Sentry. |
Status update
|
⏱️ Performance Benchmark Report: 67861b8Performance shiftsFull benchmark resultsGenerated by GHA run |
|
The team are spending the next 2 weeks giving our testing infrastructure some TLC. If we haven't made further progress on this PR by the beginning of April we should look at merging as-is and getting a bugfix release out. In the meantime we'll have to encourage sufferers to use Iris 3.5. |
I also raised dask/dask#10991 |
|
Replaced by #5833 |
Addresses #5753
Another attempt !