Broadcast #780
Broadcast #780
Conversation
Codecov Report
@@ Coverage Diff @@
## master #780 +/- ##
==========================================
+ Coverage 78.74% 78.79% +0.05%
==========================================
Files 35 35
Lines 3044 3061 +17
==========================================
+ Hits 2397 2412 +15
- Misses 647 649 +2
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
|
@ckingdon95 I worked on this today and removed a lot of the control flow and the try/catch. It should work properly now, could you take a look at the changes and see if you think there is a better way to simplify or move common cases out of loops etc.? It would be good to get some fresh eyes on it :) |
|
@ckingdon95 the doctests are being dumb right now so I removed them, but I added an Issue to add them back as soon as I can! If this looks good we can merge it and tag a new version soon. I'll keep a close eye on performance going forward as well. |
docs/src/tutorials/tutorial_3.md
Outdated
| The next step is to run DICE using the provided API for the package: | ||
|
|
||
| ```jldoctest tutorial2; output = false, filter = r".*"s | ||
| ```julai |
There was a problem hiding this comment.
oh shoot yes thanks, let me make sure the tests all pass then I'll ping you again
This PR handles the issue shown by #779 in the underlying function
_perturb_param!called within@defsim. Broadly the issue arises when there is a need to use broadcasting. This is problematic because (1) broadcast assignment ofTimestepIndextoTimestepArraydoes not have the proper underlying methods and (2) it is not always clear whether there is a need for.=or=even when there is no time index.The current solution control-flow can be described as follows. It is more complex/deep than I'd like, but it's the simplest I have for now. NOTE FLIP Y AND N ON HAS TIME INDEX FORK