Relational: bool(Relational) should raise

[jrioux]

Continuation of #1646. See issue 2832.

[asmeurer]

Nice. There are test failures, but once we get them all ironed out, let's get this in.

[jrioux]

SymPy Bot Summary: 🔴 Failed after merging jrioux/relational (b6ea43b) into master (e5dbc80).
@jrioux: Please fix the test failures.
✅ PyPy 2.0.0-beta-1; 2.7.3-final-42: pass
✅ Python 2.7.2-final-0: pass
🔴 Python 3.2.1-final-0: fail
✅ Sphinx 1.1.3: pass
Doc Coverage: unchanged
36.103% of functions have doctests (compared to 36.103% in master)
41.509% of functions are imported into Sphinx (compared to 41.509% in master)

[jrioux]

SymPy Bot Summary: ✅ Passed after merging jrioux/relational (abf77e5) into master (c18f063).
✅ PyPy 2.0.0-beta-1; 2.7.3-final-42: pass
✅ Python 2.7.2-final-0: pass
✅ Python 3.2.1-final-0: pass
✅ Sphinx 1.1.3: pass
Doc Coverage: unchanged
36.097% of functions have doctests (compared to 36.097% in master)
41.502% of functions are imported into Sphinx (compared to 41.502% in master)

[asmeurer]

I guess this behavior should be explicitly tested somewhere as well.

[jrioux]

SymPy Bot Summary: ✅ Passed after merging jrioux/relational (4c85b3f) into master (7291221).
✅ PyPy 2.0.0-beta-1; 2.7.3-final-42: pass
✅ Python 2.7.2-final-0: pass
✅ Python 3.2.1-final-0: pass
✅ Sphinx 1.1.3: pass
Doc Coverage: unchanged
36.065% of functions have doctests (compared to 36.065% in master)
41.552% of functions are imported into Sphinx (compared to 41.552% in master)

[asmeurer]

+1.

[asmeurer]

There is some stuff in the Greater docstring about chained inequalities that needs to be updated.

[jrioux]

What would you like to see changed? Everything in that docstring still applies, doesn't it?

[asmeurer]

I saw at least one part that indicated that relational evaluate to booleans (the numbered list).

[certik]

+1 from me.

[asmeurer]

This is just waiting on that fix to the docstring.

[jrioux]

SymPy Bot Summary: ✅ Passed after merging jrioux/relational (9138eb7) into master (3a92446).
✅ PyPy 2.0.0-beta-1; 2.7.3-final-42: pass
✅ Python 2.7.2-final-0: pass
✅ Python 3.2.1-final-0: pass
✅ Sphinx 1.1.3: pass
Doc Coverage: unchanged
36.059% of functions have doctests (compared to 36.059% in master)
41.562% of functions are imported into Sphinx (compared to 41.562% in master)

[asmeurer]

Awesome. Thanks for doing this. Can you update the release notes for the next release?

[asmeurer]

I should have reviewed this more closely. Several of the if a < b should have been left alone instead of changed to if (a < b) is True, because if they are symbolic, then it is a bug. This is definitely the case for the example in rde.py.

[jrioux]

Do you mean this change?

[asmeurer]

Yes. n should always be an integer here. I've no doubt this same issue is the case in many other places as well. I would say to not use is True unless the tests specifically demonstrate the symbolic arguments are possible.

[jrioux]

That's exactly what I did. I did not automatically change (a > b) into (a > b) is True, but I did so when it would fix a failing test, and the logic of the code seemed to work this way.

[jrioux]

This also means that lines that were not covered by tests (or only covered by slow tests) could still raise the symbolic error.

[jrioux]

If you do mean the above change, changing it back will get a test failure in test_rde.py. Since you added the test, you're probably the best person to decide if the test was wrong.

[asmeurer]

Ah OK, I stand corrected then. That test can probably be removed. I was trying to be clever and show that it could handle some general case mentioned in Bronstein's book, but it obviously isn't designed to work that way.

[asmeurer]

SymPy Bot Summary: 🔴 Failed after merging jrioux/relational (0ef7411) into master (1b1f247).
@jrioux: Please fix the test failures.
🔴 Python 2.6.8-final-0: fail
🔴 Python 2.7.3-final-0: fail
🔴 PyPy 1.8.0-final-0; 2.7.2-final-42: fail
✅ Python 3.2.3-final-0: pass
🔴 Sphinx 1.1.3: fail