search

lanl-ansi / PowerModelsDistribution.jl

A Julia/JuMP Package for Unbalanced Power Network Optimization
https://lanl-ansi.github.io/PowerModelsDistribution.jl/stable/
Other
146 stars 43 forks source link

UPD: update to the NonlinearExpr syntax of JuMP v1.15 #454

Closed odow closed 1 month ago

odow commented 9 months ago

Closes #453

The tests haven't finished running locally, so there might be some more changes needed, but this is a start. I'll comment in-line about some formatting changes we could do here, or in a separate PR.

codecov[bot] commented 9 months ago

Codecov Report

Attention: Patch coverage is 80.61798% with 69 lines in your changes are missing coverage. Please review.

Project coverage is 73.64%. Comparing base (80d6178) to head (c968b1c). Report is 2 commits behind head on main.

:exclamation: Current head c968b1c differs from pull request most recent head a6b5224. Consider uploading reports for the commit a6b5224 to get more accurate results

Files Patch % Lines
src/form/acp.jl 64.64% 35 Missing :warning:
src/core/objective.jl 41.66% 14 Missing :warning:
src/form/en_acr.jl 84.61% 8 Missing :warning:
src/form/acr.jl 93.33% 4 Missing :warning:
src/form/bf_fbs.jl 0.00% 2 Missing :warning:
src/form/fotp.jl 0.00% 2 Missing :warning:
src/form/fotr.jl 0.00% 2 Missing :warning:
src/form/ivr.jl 96.42% 2 Missing :warning:
Additional details and impacted files ```diff @@ Coverage Diff @@ ## main #454 +/- ## ========================================== - Coverage 75.01% 73.64% -1.37% ========================================== Files 73 71 -2 Lines 16541 16264 -277 ========================================== - Hits 12408 11978 -430 - Misses 4133 4286 +153 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

odow commented 9 months ago

@ccoffrin (or anyone): are the failing tests global solutions (and there's a bug somewhere)? Or are these local minima, and by changing the order of constraints etc we now find new solutions?

ccoffrin commented 9 months ago

Good question. I am not that familiar with how stable the tests are in terms of local minima. The failing tests do appear to be consolidated in the non-convex formulations, so it is a reasonable hypothesis. We will need one of the primary developers to look into it to assess.

odow commented 6 months ago

Hey @pseudocubic any chance you could take a look at this? Or point me in the right direction?

pseudocubic commented 6 months ago

@odow we have been looking at this but still haven't identified an obvious pattern in the failing tests, and based on the changes in PMD and JuMP 1.15 don't understand what would have caused all the failures we're seeing.

odow commented 1 month ago

Now we're getting somewhere. These two failures just look flakey? One looks like a tolerance issue, and the other is Ipopt not succeeding.

ccoffrin commented 1 month ago

Given all tests are passing, is there any reason not to merge and tag?

odow commented 1 month ago

From my side, I'm good.

odow commented 1 month ago

I've added the CHANGELOG entry.

For the compat bounds: future versions of Julia are going to decouple stdlib versions from the Julia version, so best practice is to add compat bounds. But there was a bug in Julia 1.6 that prevented them from enforcing this rule all packages at registration time.

I've removed for now, but in some future version, we'll likely need to add them back. (Probably when/if you set julia = "1.10".)

pseudocubic commented 1 month ago

Looks good! Thanks for the info about stdlib versions getting decoupled, I didn't know that, so I'll keep an eye out in the future.

Thanks again for the contribution!