Deprecate set.is_real property, use set.is_subset(Reals) instead

[skirpichev]

fixes #6212

(on top of #7861, see also #7972)

[smichr]

+1 to this commit

[skirpichev]

But please note, that we can't merge this, unless we merge first #7861 or #7972.

I really like quantum stuff in the sympy tree, but this shouldn't block development for month!

[smichr]

Let's rebase on #7972 and commit.

[skirpichev]

Let's rebase on #7972 and commit.

Why on this? Yes, less "broken" tests here, but I did here a lot of changes in the quantum module.

[smichr]

You made changes in the quantum module by adding assumptions to symbols which are consistent with the existing test results (which suggests that this is what the quantum folks meant).

[skirpichev]

On Mon, Sep 08, 2014 at 10:50:54AM -0700, Christopher Smith wrote:

You made changes in the quantum module by adding assumptions to symbols
which are consistent with the existing test results

Yes. And I did very questionable change in the represent.py for this.

(In my view, the usage of Symbol's is totally broken in the quantum
module and I don't see an easy way to fix this.)

(which suggests that this is what the quantum folks meant).

Or that they have very bad tests or(and) bad test coverage.

[smichr]

Is there any way to make XKet and Xbra be real and bounded? Otherwise this result could be Piecewise((x_DiracDelta(x - x_3)_DiracDelta(x_1 - y), Ne(x, -oo)), (S.NaN, True)).

[smichr]

ditto below

[skirpichev]

Is there any way to make XKet and Xbra be real and bounded?

I don't know what you mean by this. But if you do care about this test - we can preserve it, see #7972

But lets discuss #7972 and #7861 not in this pr.

[skirpichev]

rebased

[skirpichev]

it seems others are ok with this deprecation in #6212

we can merge this pr after either #7972 or #7861

[skirpichev]

@asmeurer, can we consider this for 0.7.6?

[asmeurer]

Absolutely. Is it ready to go?

[skirpichev]

Absolutely. Is it ready to go?

The only problem - it depends on either #7861 or #7972.

[asmeurer]

Can you summarize the differences? It says the only difference is the fixes in the quantum module. The one uses real=True, bounded=True, and I can't tell what the other does.

[asmeurer]

Oh I guess the other one changes the test output.

[asmeurer]

I actually have issues with both of those. They both introduce this strange new Heaviside(-x - oo) term, leading to

In [1]: integrate(-x*DiracDelta(x - y), (x, -oo, oo))
Out[1]: y⋅Heaviside(-y - ∞) - y

If y is an infinity that is a nan and otherwise it is -y (the result you get in master for this integral).

[skirpichev]

On Sat, Sep 13, 2014 at 02:32:07PM -0700, Aaron Meurer wrote:

I actually have issues with both of those. They both introduce this
strange new Heaviside(-x - oo) term, leading to

In [1]: integrate(-x*DiracDelta(x - y), (x, -oo, oo))
Out[1]: y⋅Heaviside(-y - ∞) - y

If y is an infinity that is a nan

Yes. That's why we can't simplify this term. -y can be oo.

[skirpichev]

rebase on top of #7861

[skirpichev]

will merge in 24hr