AssocOp._eval_template_is_attr is wrong or misused

[ness01]

This returns True if all elements have the property, False if at least one has the property set to False and all others have either True/False, and None if at least one element has the property set to None.

This is either not right or not used correctly. e.g.

In [5]: x = Symbol('x', real=True, bounded=False)
In [7]: (x*sin(1/x**2)).is_bounded
Out[7]: False # Wrong?

In [8]: k = Symbol('k', integer=True)
In [9]: (k/3).is_integer
Out[9]: False # Wrong if k=3

On the other hand changing the function to return True if all elements have the property and None otherwise introduces lots of test failures.

Original issue for #5662: http://code.google.com/p/sympy/issues/detail?id=2563
Original author: https://code.google.com/u/104531927090589914088/
Original owner: https://code.google.com/u/104531927090589914088/

[asmeurer]

I don't completely follow how the _eval_template_is_attr effects the wrong results.

**Labels:** Assumptions  

Original comment: http://code.google.com/p/sympy/issues/detail?id=2563#c1
Original author: https://code.google.com/u/asmeurer@gmail.com/

[rlamy]

Yes, it's clearly wrong, we discussed it a bit on IRC previously: http://kottnet.net/irclogs/freenode/chn-%23sympy/2011-06-30.log (starts at 17:07).

Some of the tests failures could be due to the tests being wrong. ISTR there's one somewhere that's equivalent to assert (k/3).is_integer is False.

Original comment: http://code.google.com/p/sympy/issues/detail?id=2563#c2
Original author: https://code.google.com/u/101272611947379421629/

[ness01]

aaron: _eval_template_is_attr is called by various _eval_is_foo methods.

While some of the tests are obviously wrong, some are not. What I'm trying to say is that apparently some uses of _eval_template_is_attr where the current semantics are /correct/.

Original comment: http://code.google.com/p/sympy/issues/detail?id=2563#c3
Original author: https://code.google.com/u/104531927090589914088/

[asmeurer]

Perhaps it could benefit from a more descriptive name then.  Though I wouldn't go too far redesigning the system beyond fixing this bug, as it's all old assumptions stuff.  Did you try some of these things in the new assumptions to see if they are handled correctly there?

Original comment: http://code.google.com/p/sympy/issues/detail?id=2563#c4
Original author: https://code.google.com/u/asmeurer@gmail.com/

[ness01]

No. I fixed Mul and Add ._eval_is_rational, ronan suggested I should fix _eval_template_..., but I couldn't figure out how to. Hence this bug report.

Original comment: http://code.google.com/p/sympy/issues/detail?id=2563#c5
Original author: https://code.google.com/u/104531927090589914088/

[asmeurer]

**Status:** Valid  

Original comment: http://code.google.com/p/sympy/issues/detail?id=2563#c6
Original author: https://code.google.com/u/asmeurer@gmail.com/