catch ValueError in comparisons

[videlec]

When element construction by a parent raise a ValueError, it prevents using the coercion in comparisons.

Note: a coercion should not fail and there is no reason to catch a ValueError.

CC: @staroste @seblabbe @videlec

Component: coercion

Keywords: comparison

Reviewer: Vincent Delecroix

Issue created by migration from https://trac.sagemath.org/ticket/13957

[videlec]

Attachment: trac_13957-catch_value_error.patch.gz

[seblabbe]

comment:1

Can you add a doctest that illustrates the fix ?

[videlec]

comment:2

No... the only way I found is through a Cython class that inherits from Element. A concrete example is with finite words in #12224. I will look further.

[tscrim]

comment:4

Sebastien, did you do a review of the patch otherwise?

Best,

Travis

[seblabbe]

comment:5

Replying to @tscrim:

Sebastien, did you do a review of the patch otherwise?

Since I am not familiar with this, I would have prefer to do some test starting from a doctest, thus my question. So, no, I did not do any further review.

Sébastien

[tscrim]

comment:6

Replying to @videlec:

No... the only way I found is through a Cython class that inherits from Element. A concrete example is with finite words in #12224. I will look further.

I completely forgot about this ticket. Vincent, could you do a minimal/small example by doing the doctest something like

sage: cython("""
....: class Foo(Element):
....:     pass
....: """)
sage: Foo() == Foo()
sage: Foo() < Foo()

Thanks,

Travis

[videlec]

comment:7

Hi Travis,

I open the ticket because we made some stupid design decisions for #12224. The point was that we implement a coercion from lists to Words and we raised some ValueError if the content of the list does not match the alphabet. As I learn, it is definitely not what we want from a coercion: a coercion should always succeed.

I close it as wont fix.

[videlec]

Changed author from Vincent Delecroix to none

[jdemeyer]

Reviewer: Vincent Delecroix