Fix another "recursion depth exceeded" in memory deallocation for weak dictionaries

[nthiery]

See: #10963 and follow up for details

This is a follow up to #13394 #15069, and a blocker for #10963.

Depends on #15432

CC: @sagetrac-sage-combinat @vbraun @jdemeyer @simon-king-jena

Component: performance

Author: Nils Bruin

Branch/Commit: u/SimonKing/ticket/15506 @ 6147777

Reviewer: Simon King

Issue created by migration from https://trac.sagemath.org/ticket/15506

[nbruin]

Branch: u/nbruin/ticket/15506

[nbruin]

comment:2

This should do the trick for WeakValueDictionary. Merge #15367 to get the same benefit for MonoDict and TripleDict.

Code that verifies the problem has been solved:

sage: class A: pass
sage: from sage.misc.weak_dict import WeakValueDictionary
sage: a=A(); prev=a;
sage: M=WeakValueDictionary()
sage: for i in range(10^3+10): newA = A(); M[newA] = prev; prev = newA
sage: del a

This used to fail, but with the code on this branch it works.

---

New commits:

[5de9c7c](https://github.com/sagemath/sagetrac-mirror/commit/5de9c7c)

trac 15506: ensure that WeakValueDictionary deletes entries via a list, so that we benefit from Python's trashcan

[nbruin]

Commit: 5de9c7c

[sagetrac-git]

Changed commit from 5de9c7c to e23ec99

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

[e23ec99](https://github.com/sagemath/sagetrac-mirror/commit/e23ec99)	Merge branch 'ticket/15432' into ticket/15506
[4ac6866](https://github.com/sagemath/sagetrac-mirror/commit/4ac6866)	trac #15432: Code context manager in iterator via try/finally. This saves an entire class and some reference management.Z
[8b70d1d](https://github.com/sagemath/sagetrac-mirror/commit/8b70d1d)	trac #15432: also equip iteration context with a weak reference
[9bb9241](https://github.com/sagemath/sagetrac-mirror/commit/9bb9241)	trac #15432: use a callback class object holding a weak reference for sage.misc.weak_dict.WeakValueDictionary

[nbruin]

comment:4

Oops... while we're at it, we should merge #15432 as well. Do I need to merge #15367 into this branch as well or do we leave that separate?

[nthiery]

comment:5

That was quick :-) Thank you Nils!

Is there a reason for not putting the above code that verifies the solution as a doctest?

Other than that, what shall we do? Backport this ticket and its dependencies to mercurial, or move #10963 to git?

[nbruin]

comment:6

Replying to @nthiery:

Is there a reason for not putting the above code that verifies the solution as a doctest?

No, if you want to add a doctest, go ahead.

Backport this ticket and its dependencies to mercurial, or move #10963 to git?

I have no preference. Given that you'd have to backport #15367 as well, I think you'd better bite the bullet and go with git.

[nthiery]

comment:7

Replying to @nbruin:

Replying to @nthiery:

Is there a reason for not putting the above code that verifies the solution as a doctest?

No, if you want to add a doctest, go ahead.

Backport this ticket and its dependencies to mercurial, or move #10963 to git?

I have no preference. Given that you'd have to backport #15367 as well, I think you'd better bite the bullet and go with git.

I am fine with both, as long as it goes quickly. Volker, Jeroen, what do you think from a release management point of view?

[jdemeyer]

comment:8

I don't see a good reason to force this ticket in Sage 5.13. So it should not be a blocker and should be postponed to Sage 6.0 which means git.