Pickling morphisms is broken

[caruso]

Here is an exemple:

sage: R.<x,y> = QQ[]
sage: theta = R.hom([y,x])
sage: hash(theta)
-1982612945128833994
sage: hash(loads(dumps(theta)))
Traceback (most recent call last):
...
ValueError: mutable sequences are unhashable

[roed]

New commits:

​94f6985

Set immutable in unpickling ring morphisms

[tscrim]

This is definitely a fix, but is it the correct fix? The PolynomialSequence does not seem to pickle immutability like I think it should:

sage: theta.__reduce__()[1][3]['__im_gens']
[y, x]
sage: S = theta.__reduce__()[1][3]['__im_gens']
sage: type(S)
<class 'sage.rings.polynomial.multi_polynomial_sequence.PolynomialSequence_generic'>
sage: S.is_immutable()
True
sage: loads(dumps(S)).is_immutable()
False

Contrast:

sage: S.__reduce__()
(<function PolynomialSequence at 0x7f62c7ef2c80>,
 (Multivariate Polynomial Ring in x, y over Rational Field, ((y, x),)))
sage: Sequence([1,2,3,4]).__reduce__()
(<function _reconstructor at 0x7f64198a5aa0>,
 (<class 'sage.structure.sequence.Sequence_generic'>,
  <type 'sage.structure.sage_object.SageObject'>,
  <sage.structure.sage_object.SageObject object at 0x7f62b02dc0b0>),
 {'_Sequence_generic__cr': False,
  '_Sequence_generic__cr_str': False,
  '_Sequence_generic__hash': None,
  '_Sequence_generic__universe': Integer Ring,
  '_is_immutable': False})

What are your thoughts about instead adjusting the pickling of PolynomialSequence?

[caruso]

I agree with you, Travis.

I pushed a fix for the pickling of PolynomialSequence. I'm not sure I've done it correctly (in particular, this class redefines __getattr__, I don't know why); please review :-)

---

New commits:

​dee878e	Revert "Set immutable in unpickling ring morphisms"
​e2931b6	Fix pickling for PolynomialSequence_generic

[git]

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

​1ea91f7

Add a doctest

[tscrim]

Yep, LGTM. Thanks.