Should we remove old assumptions?

[mrocklin]

rlamy says

I think it would be a serious mistake to throw away a working system (the old assumptions) for the sake of an untested one with known deficiencies (the new assumptions). It would probably be more productive to view the new assumptions as complementary to the old ones, and treat it as a given that they are intrinsically more powerful, but slower and more cumbersome to use than the old ones.

Original issue for #6738: http://code.google.com/p/sympy/issues/detail?id=3639
Original author: https://code.google.com/u/109882876523836932473/

[mrocklin]

**Cc:** Ronan.L...@gmail.com  
**Labels:** Assumptions  

Original comment: http://code.google.com/p/sympy/issues/detail?id=3639#c1
Original author: https://code.google.com/u/109882876523836932473/

[mrocklin]

**Blocking:** 6730  

Referenced issues: #6730
Original comment: http://code.google.com/p/sympy/issues/detail?id=3639#c2
Original author: https://code.google.com/u/109882876523836932473/

[asmeurer]

The issue is, can we refactor the core (or really do anything with it) if it still has the old assumptions sitting around.

Now, a phase out period could be argued.

Original comment: http://code.google.com/p/sympy/issues/detail?id=3639#c3
Original author: https://code.google.com/u/asmeurer@gmail.com/

[rlamy]

It depends what sort of refactoring you have in mind. Having the old assumptions around is not a problem in itself, IMO.

Original comment: http://code.google.com/p/sympy/issues/detail?id=3639#c4
Original author: https://code.google.com/u/101272611947379421629/

[mrocklin]

It would be nice to remove some of the magic around Symbol.

What will sympy use in general code? What will is_positive call? Old assumptions or new assumptions? There might be some interference if different parts of sympy use different systems. Assertions made in one system are unlikely to carry over to the other.

E.g.

x = Symbol('x', positive=True)
print ask(Q.positive(x))
None

Original comment: http://code.google.com/p/sympy/issues/detail?id=3639#c5
Original author: https://code.google.com/u/109882876523836932473/

[rlamy]

Yes, this a real problem that could be fixed quite easily today (by having the new assumptions call the old ones), were it not for the fact that there are incompatible definitions (cf. issue 5976 ).

Referenced issues: #5976
Original comment: http://code.google.com/p/sympy/issues/detail?id=3639#c6
Original author: https://code.google.com/u/101272611947379421629/