Pickling of Functors

[SimonKing]

We used to have:

sage: F = QQ['t'].construction()[0]
sage: G = loads(dumps(F))
sage: G == F
True
sage: G.domain() == F.domain()
False
sage: G.codomain() is None
True

In other words, the cdef'd attributes of the construction functor are lost.

I created a generic pickling method for functors, so that it works a lot better:

sage: F = QQ['t'].construction()[0]
sage: G = loads(dumps(F))
sage: G == F
True
sage: G.domain() == F.domain()
True
sage: G.codomain() is None
False

The patch contains two tests, showing that the problem is solved.

[SimonKing]

To the patchbot / release manager:

Depends on #8800

So, in fact, it also depends on #8807 and #10318.

Up to now, I only tested sage/categories/functor.pyx and sage/categories/pushout.py. However, since they use functors intensely, I am already marking the ticket as "needs review".

[SimonKing]

Since the patchbot complained, I updated the patch. It should cleanly apply to sage-4.6.2.alpha0.

It turns out that the patch does in fact not depend on #8800. It only depends on #8807 and #10318, which are already merged. So, no dependency.

[SimonKing]

I don't know why the patchbot complains. The patch applies cleanly to a fresh sage-4.6.2.alpha4.

[SimonKing]

Pickling of functors. Independent of #8800

[SimonKing]

Since the patchbot did not try to apply the patch again and since the patch cleanly applies to sage-4.6.1.alpha4, I now try to trigger another attempt of the patchbot by posting a new copy of the patch.

And, by the way, this rather simple patch still needs review...

[mraum]

I only corrected tiny typos in the documentation, and I altered one doctest to better reflect the change that you introduced.

[mraum]

Replacement for original patch.

[SimonKing]

Replying to mraum:

I only corrected tiny typos in the documentation, and I altered one doctest to better reflect the change that you introduced.

Thank you, Martin!

For the record, the patchbot and the release manager:

Apply trac-10460-pickling_functors-review.patch