Issue 2696 in sympy: Assumptions fail for non-commutative symbols
sy...@googlecode.com
Owner: MRock...@gmail.com
Labels: Type-Defect Priority-Medium WrongResult
New issue 2696 by MRock...@gmail.com: Assumptions fail for non-commutative
symbols
http://code.google.com/p/sympy/issues/detail?id=2696
In [1]: x = Symbol('x', commutative=False)
In [2]: x.is_irrational
In [3]: (-x).is_irrational
Out[3]: False
The Mul._eval_is_irrational method has correct behavior. I suspect this is
an issue with the new assumptions
sy...@googlecode.com
Labels: Assumptions
Comment #1 on issue 2696 by asmeurer: Assumptions fail for non-commutative
symbols
http://code.google.com/p/sympy/issues/detail?id=2696
The new assumptions seem to handle it correctly:
In [1]: x = Symbol('x', commutative=False)
In [2]: print ask(Q.irrational(-x))
None
sy...@googlecode.com
Comment #2 on issue 2696 by smi...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
in facts.py it says
rules: {} of a -> [b, c, ...]
return: {} of c <- [a, b, ...]
Note however, that this prerequisites may be *not* enough to prove a
fact. An example is 'a -> b' rule, where prereq(a) is b, and
prereq(b)
is a. That's because a=T -> b=T, and b=F -> a=F, but a=F -> b=?
The prereqs traversed for is_irrational (after obtianing a correct None
from Mul.is_irrational) are integer, even, odd, all of which are False. The
conclusion returned is False. So perhaps the above applies?
sy...@googlecode.com
Comment #3 on issue 2696 by asmeurer: Assumptions fail for non-commutative
symbols
http://code.google.com/p/sympy/issues/detail?id=2696
The result should be None then. To put it in the perspective of the
comment from facts.py, ? should equal None.
sy...@googlecode.com
Comment #4 on issue 2696 by Ronan.L...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
irrational implies real, which implies complex, which implies commutative,
so x.is_irrational should return False.
The new assumptions aren't involved here.
sy...@googlecode.com
Comment #5 on issue 2696 by Ronan.L...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
The problem is that 'commutative' is special-cased in Symbol, which
prevents it from being handled correctly by the assumption system (this was
briefly discussed in issue 2446). I think that the solution is to remove
this special-casing.
sy...@googlecode.com
Comment #6 on issue 2696 by asmeurer: Assumptions fail for non-commutative
symbols
http://code.google.com/p/sympy/issues/detail?id=2696
Ah, I see. noncommutative implies not complex. So this is correct.
Should this be closed then?
By the way, I seem to remember discussing this before, but is it a good
idea to make irrational imply real? This way, irrational != not rational,
which is somewhat confusing. I suppose it might seem strange to say that a
number like I is irrational, but I think it technically fits with the
definition (cannot be written as the quotient of integers).
sy...@googlecode.com
Comment #7 on issue 2696 by Ronan.L...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
There's still the issue of x.is_irrational returning None instead of False,
and also the fact that inconsistent assumptions are allowed, like
Symbol('x', real=True, commutative=False).
Concerning is_irrational, I think that this kind of negative assumption is
more confusing than helpful. Testing for not rational is easy: 'if not
x.is_rational: ...' so rather than making irrational == not rational, it
would be simpler to remove it altogether.
sy...@googlecode.com
Comment #8 on issue 2696 by Ronan.L...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
I sent a PR implementing the idea from comment 5.
https://github.com/sympy/sympy/pull/618
sy...@googlecode.com
Comment #10 on issue 2696 by julien.r...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
About Symbol('x', real=True, commutative=False): I though that would be a
symbol representing e.g. a real-valued matrix.
sy...@googlecode.com
Comment #11 on issue 2696 by asme...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
I would use MatrixSymbol for that. The way our assumptions deductions
work, real implies things that are only true for real numbers
(commutativity being one of them).
sy...@googlecode.com
Comment #12 on issue 2696 by julien.r...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
Well, that's one example. It could also represent a differential operator.
Anyway, I always took this real=True assumption to mean conjugate(x)==x and
nothing else. I would be interested to know, where in the code does real
imply commutative.
MatrixSymbol looks like a relatively recent addition. Maybe I ought to use
that, as you say, but up to now I never needed anything else then Symbol. I
can't seem to assume conjugate(x)==x when I use MatrixSymbol.
sy...@googlecode.com
Comment #13 on issue 2696 by asme...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
> I would be interested to know, where in the code does real imply
> commutative.
Well, where this bug manifests itself for one thing :)
This sort of thing is tricky, because a lot of it is just implicit
assumptions about how things work. That's why we should really fix issue
2196.
Another thing that is assumed by real=True is comparable=True. In other
words, it is assumed that you have a < b, a > b or a == b for any two real
a and b. This is only true if a and b are real numbers (not matrices or
operators or anything).
I suppose could consider Hermitian operators to be "real", since they act
very similarly to real numbers in most contexts. In many cases, you could
probably get sympy to do what you want for such an operator by using a
Symbol that is assumed to be real.
MatrixSymbol is still relatively new, so there are many things that aren't
implemented yet. Feel free to open issues as you see fit.
sy...@googlecode.com
Comment #14 on issue 2696 by julien.r...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
MatrixSymbol has many problems at first glance (issue 3150). Since I can no
longer have Symbols assumed real and noncommutative, I've opened issue 3151
to have an Hermitian assumption.
sy...@googlecode.com
Status: Fixed
Comment #16 on issue 2696 by julien.r...@gmail.com: Assumptions fail for
non-commutative symbols
http://code.google.com/p/sympy/issues/detail?id=2696
The pull request has been committed but the issue tracker wasn't updated:
commit 09409374fb9d5e468a0dd771f145c884f86ba472
Merge: d26aadc 6a357cf
Author: Christopher Smith <smi...@gmail.com>
Date: Wed Mar 14 19:20:39 2012 -0700
Merge pull request #1116 from smichr/commut
Symbol commutativity patches
And indeed this is fixed:
In [1]: from sympy import *
In [2]: x=Symbol('x', commutative=False)
In [3]: x.is_irrational
Out[3]: False
In [4]: (-x).is_irrational
Out[4]: False