Softly deprecate PSD by removing it from the docs #3959
Conversation
| @@ -158,7 +158,7 @@ function example_k_means_clustering() | |||
| end | |||
| model = Model(Clarabel.Optimizer) | |||
| set_silent(model) | |||
| @variable(model, Z[1:m, 1:m] >= 0, PSD) | |||
| @variable(model, Z[1:m, 1:m] >= 0, set = PSDCone()) | |||
There was a problem hiding this comment.
This one is a reason not to stop using PSD. It's clunky, and I can see people getting tripped up over it.
There was a problem hiding this comment.
I like it with the "set", the PSD one does not look better to me
There was a problem hiding this comment.
Would you think its clearer if we replace ">=0" by "lower_bound=0".
I would not mind.
There was a problem hiding this comment.
No I want to keep this as it was... We have nice syntax. There's no need to use lower_bound=0 like we're a Python library.
There was a problem hiding this comment.
Was the main reason to stop using PSD because VariableInSet is confusing?
There was a problem hiding this comment.
I dont agree that
x >= 0, PSD
Is better than
x >= 0, set = ...
We should decide if its legacy or not.
For me it felt like it was.
For @blegat the same.
We can keep it, but if we do, we should have PSDRef.
There was a problem hiding this comment.
We can keep it, but if we do, we should have PSDRef.
We have VariableInSetRef though?
What about #3960
There was a problem hiding this comment.
I agree that x >= 0, set = PSDCone() is nice and I'm not missing x >= 0, PSD.
With , PSD, you're saving a few keystrokes but then it's unclear how to adapt it to hermitian, SOC, etc...
At least, with set = PSDCone() the user learns something that is transferrable
There was a problem hiding this comment.
I'm missing x >= 0, PSD). I vote we leave as-is and wait for more people to ask how to get the dual of a variable constraint.
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #3959 +/- ##
=========================================
Coverage 100.00% 100.00%
=========================================
Files 43 43
Lines 6054 6085 +31
=========================================
+ Hits 6054 6085 +31 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
|
I'm not convinced that this is a good idea. |
|
I would not mind adding a note to the variable macro docstring asking to use "set =..." |
|
It's kinda weird to be completely removing syntax from the docs. What if someone is trying to read existing code? |
I think this is especially true because it is magic syntax in the macro. You can't look it up as |
|
Closing because I don't particularly want to do this. Let's see how |
|
My suggestion was to remove PSD everywhere, except from the |
|
We could at least use |
| @@ -1082,13 +1082,13 @@ julia> model[:x] | |||
|
|
|||
| ## Semidefinite variables | |||
|
|
|||
| Declare a square matrix of JuMP variables to be positive semidefinite by passing | |||
There was a problem hiding this comment.
We could just say here that there are the two possibilities at least, we should reopen and merge part of the changes of this PR.
Is it though? |
x-ref #3955 (comment)
Thoughts?