Set factories

[nthiery]

At Sage days 30, a long brainstorm seems to have finalized the design. Here is a excerpt from the documentation:

A set factory F is device, whose goal is to construct parent P which models subsets of a big set S. Typically, the P s are constructed within families obtained by putting a bunch of constraints cons on the elements of the set S. In such a hierarchy of subsets, one needs to have a fine and easy control on the elements construction. That is, one often needs that P constructs elements in a subclass of its usual class for element. On the contrary, one also often needs P to be a facade parent, meaning that P construct element whose actual parent is not P itself.

The role of a set factory is twofold:

• manage a database of constructors for the different parents P = F(cons) depending on the various kinds of constraints cons. Note: currently there is no real support for that. We are gathering use case before fixing the interface.

• ensure that the elements e = P(...) created by the different parents follows a consistent policy concerning their class and parent.

The patch implement this idea while trying to leave as much as possible space for further improvement. In particular, I tried to specify the few possible things about constraints. I'm even not completely sure about add_constraints. Please comment and review.

Florent

[nthiery]

Patch under development on the Sage-Combinat patch server.

[davidloeffler]

This fails doctests on every version tested -- see patchbot logs

[hivert]

Replying to davidloeffler:

This fails doctests on every version tested -- see patchbot logs

I forgot a dependency to #10963. Thanks for pointing this out.

[davidloeffler]

There is no patch at #10963. I don't think this can be meaningfully tagged as "needs review" when one of its dependencies doesn't exist yet.

[hivert]

Replying to davidloeffler:

There is no patch at #10963. I don't think this can be meaningfully tagged as "needs review" when one of its dependencies doesn't exist yet.

As you can see on #10963, the patch is far for being inexistent as Simon and Christian started to review it. Of course it is on the Sage-Combinat queue. I need to ping Nicolas to have him finish it. Now, I don't see why having Nicolas's patch unfinished forbid me to have some feedback on this one. And asking for such a feedback seems to be a perfectly meaningful reason for putting "needs review"... The only reason I can see for not doing it is that it can trouble the patchbot. Do you have another one ?

[nthiery]

Salut Florent!

Why is there a dependency on 10963? In the queue, this patch is much lower than 10963.

[nthiery]

Ah, I see, it's about the glitch Sets().Facade() <-> Sets().Facades().

Since anyway the patch is up the queue, I suggest we finish commuting the two patches. Just use Sets().Facades() (we will need to keep the backward compatibility anyway).

With that, the tests pass up to to trivially failing doctests:

File "/opt/sage-5.0.beta11/devel/sage-combinat/sage/structureset_factories.py", line 644:
    sage: P2.category()
Expected:
    Join of Category of facade finite enumerated sets and Policy category with facade parent AllPairs
Got:
    Join of Category of finite enumerated sets and Policy category with facade parent AllPairs
**********************************************************************
File "/opt/sage-5.0.beta11/devel/sage-combinat/sage/structure/set_factories.py", line 972:
    sage: P.category()
Expected:
    Join of Category of facade finite enumerated sets and Policy category with facade parent AllPairs
Got:
    Join of Category of finite enumerated sets and Policy category with facade parent AllPairs

Cheers,

[chapoton]

new version, rebased on 5.12.beta0, with cleanup

[chapoton]

I have made a git branch, but it seems to be massively broken.

Some problems come from Sets().Facade() versus Sets().Facades() as explained before

Some other problems come from #14519

And some other problems I do not understand..

---

New commits:

​5dbd72b

trac_10194-factories_policy-fh.patch

[git]

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

​57f3ed3

trac #10194 minor changes

[git]

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

​28c42cb

trac #10194 typo corrections

[chapoton]

Temporary : emptied dependency field (was #10963)

[git]

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

​6fa45d8

trac #10194 other details, still not working :(

[chapoton]

This needs to be worked upon, as it is not working at all.

[chapoton]

Maybe some of the problems come from #13801 ?

[git]

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

​06beedc

Fixed passing parent to ElementWrapper

[git]

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

​d7eb4c1

Cleanup in constraints managment.

[git]

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

​b615150

Fixed printing of _repr_ for TopMostParentPolicy

[chapoton]

Maybe you should add #13580 as a dependency ?

[hivert]

Replying to chapoton:

Maybe you should add #13580 as a dependency ?

Why ? Those are completely different things.

[chapoton]

Then why do you have that in the commits of this ticket ?

[hivert]

Replying to chapoton:

Then why do you have that in the commits of this ticket ?

I probably screwed things up with git ! I'm trying to fix the problem.

[nborie]

For information, this features is largely tested in #16094

[hivert]

Should be fixed now ! But I had to rebase the branch. So Nicolas, you will have to rebase yours at some point. Sorry !

---

New commits:

​dc5dfc8	# Mon Apr 07 00:47:29 2014 +0200
​e595280	Fixed passing parent to ElementWrapper
​80b08c3	Cleanup in constraints managment.
​fc6981b	Fixed printing of _repr_ for TopMostParentPolicy

[chapoton]

I have made one more commit. Now all tests pass, and code is in the pep8 standard.

---

New commits:

​77279f9	# Mon Apr 07 00:47:29 2014 +0200
​272d17f	Fixed passing parent to ElementWrapper
​1491bff	Cleanup in constraints managment.
​41c137e	Fixed printing of _repr_ for TopMostParentPolicy
​f76e0e6	Review : fonctionality, documentations, tests (very small things remains...)
​1f3e02b	Merge branch 'u/nborie/10194-rebased-beta7' of trac.sagemath.org:sage into 10194
​0eeecd1	trac #10194 make sure that all test pass (and pyflakes and pep8 compliant)

[chapoton]

There remains a pyflakes warning thta needs to be taken care of :

src/sage/structure/set_factories.py:722: undefined name 'FacadeParentPolicyCategory'

[git]

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

​f823667	Minor doc improvements
​16ca519	Merge branch 'public/ticket/10194' of trac.sagemath.org:sage into t/10194/factories_policy

[chapoton]

what about the pyflakes warning ? could you take care of that please ?

[git]

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

​364a0c0	Merge branch 'public/ticket/10194' of ssh://trac.sagemath.org:22/sage into 10194
​014d842	trac #10194 details and formatting

[git]

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

​4895320	Merge branch 'public/ticket/10194' of ssh://trac.sagemath.org:22/sage into 10194
​e061484	trac #10194 removed the category method (was useless and wrong ?)

[chapoton]

At last, got a green report from the bot. Florent, do you think that this is ready ?

[git]

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

​87a475c

Merge branch 'public/ticket/10194' of ssh://trac.sagemath.org:22/sage into 10194

[git]

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

​641c28a

trac #10194 minor doc changes

[git]

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

​3c12f5f	Merge with 6.3.beta4
​ed7f340	trac #10194 found a few typo again

[chapoton]

@Florent : What is the meaning of the sentences

Put a nice warning _single_pair

and

Comment that and put link to documentation caveat....

in the example file ?

[git]

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

​20f17b3

Merge branch 'public/ticket/10194' of trac.sagemath.org:sage into 6.4.b6

[chapoton]

@Florent : PING, cf comment 53

[git]

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

​52550e0

Merge branch 'develop' into t/10194/public/ticket/10194

[chapoton]

9 failing doctests, see patchbot report

[git]

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

​ab1507c

Re-reading set factories

[hivert]

Replying to chapoton:

9 failing doctests, see patchbot report

Hi Frederic,

I should have fixed most of the problems (including comment 53). I still need a last pass of re-reading before putting it as needs review.

Florent

[git]

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

​586248e

Various cleanup in the documentation

[hivert]

This should be ready for review

Florent

[git]

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

​763f93c

Merge tag '6.7' into t/10194/factories_policy

[chapoton]

Was there any need for a merge ? By the way, I usually do the merge in the other direction. This gives much smaller commits, as you can see in the commit list.

[hivert]

Replying to chapoton:

Was there any need for a merge ? By the way, I usually do the merge in the other direction. This gives much smaller commits, as you can see in the commit list.

No there wasn't, but my goal is to update trees to use factories and ultimately to do computation in operads. As you merged 6.7 in operads I also merged this one to avoid going back and forth.

There is not diff before and after the merge so that you can review on the unmerged version.

[chapoton]

ok, let this get in. There are several combinatorial tickets on top of it.