Create an assuming() context manager

[charpent]

Inspiration : Ralf Stephan in #22322. Thanks to him for nudging me in the right direction.

Goal : create a tool allowing to create reusable, stackable, assumption elements, thus allowing to get the functionality of Mathematica's Assuming[...,???]

What I mean :

sage: solve(x^2==4, x)
[x == -2, x == 2]
sage: with assuming(x>0): solve(x^2==4, x)
[x == 2]
sage: assumptions()
[]

[charpent]

Initial proposition.

On 8.1beta9 + #24026+ #24006 + #23990 + #23923 + #24088 + #10035, passes ptestlong with no error whatsoever (1 timeout once on sage -t --long src/sage/modular/hecke/submodule.py, unreproductible).

==>needs_review

[charpent]

Aaarghh... I've found a subtle bug: after with assuming(): do_something(), the original assumption database is destroyed.

That's due to the fact that forget() is equivalent to forget(assumptions()), i. e. forgets all facts in the database. Hoisted by a "lazy" facility...

I'll fix that (and add a couple doctests for exceptions.

Should we reconsider forget() <==> forget(assumptions()) ?

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​c50defe

24119 : Debugging, better doctests, a couple of typos

[charpent]

Replying to git:

Branch pushed to git repo; I updated commit sha1. New commits:

​c50defe

24119 : Debugging, better doctests, a couple of typos

Passes ptestlong with no errors whatsoever, an one non-reproducible timeout on sage -t --long src/sage/symbolic/expression.pyx.

==>needs_review again.

[charpent]

Replying to charpent:

Aaarghh... I've found a subtle bug: after with assuming(): do_something(), the original assumption database is destroyed.

[ Snip ... ]

To be absolutely clear : the problem existed when one used assuming with an empty argument list : you wouldn't do that from the console, but this argument list might result from another computation, which might return an empty list.

Anyway, problem fixed.

[rws]

Could you please add direct doctests to the member functions?

[rws]

Also add a doctest with the class that checks correctness with at least doubly stacked assumptions.

[charpent]

Replying to rws:

Could you please add direct doctests to the member functions?

What do you mean ? I added exactly ONE function (assuming()), which I documented and doctested, and added a small blurb at the end of the module documentation. Since assuming() is designed to interact exclusively with assume(), I don't see what you mean.

[charpent]

Replying to rws:

Also add a doctest with the class that checks correctness with at least doubly stacked assumptions.

This I understand and will do. Stay tuned...

[rws]

The class assuming has three member functions, no? Check class hold_class in #10035 where __enter__ and __exit__ is doctested.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​a43a32a

24119 : doctest for stacks of assumptions.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​127d507

24119 : doctesting individual assuming's member functions.

[charpent]

Replying to git:

Branch pushed to git repo; I updated commit sha1. New commits:

​127d507

24119 : doctesting individual assuming's member functions.

Builds and doctests OK. make ptestlong underway (don't hold your breath...).

[rws]

I like it. Here some minor typos: double space after declaration; :func:assume()) should have backticks; typo in useable in a in a with statement; small caps begin new sentence in ...can be stacked. we can use this functionality can be used to... also use is used twice; please correct (and by yhe way,; finally please add an empty line before def __enter__ and before def __exit__.

[charpent]

Passes ptestlong with no error whatsoever.

I'll fix the typos and re-upload, but won't re-ptestlong before late tonight (I *have* to go now...).

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​017f67c

24119 : small typos fixed.

[charpent]

Replying to git:

Branch pushed to git repo; I updated commit sha1. New commits:

​017f67c

24119 : small typos fixed.

Typos fixed, except a double space after declaration, which I didn't find. Builds and doctests OK. ptestlong later tonight. Leaving NOW.

[charpent]

Replying to charpent:

Replying to git:

Branch pushed to git repo; I updated commit sha1. New commits:

​017f67c

24119 : small typos fixed.

Typos fixed, except a double space after declaration, which I didn't find. Builds and doctests OK. ptestlong later tonight. Leaving NOW.

[ Later... ]

• Passes ptestlong with no error whatsoever.
• Most of the 17 patchbots seem to like it, but all but one of them complain that "a plugin failed" ; it seems that this is the "coverage" plugin. Not my watch, I'm afraid...
• Four patchbots got a failure or a timeout :
  • timeout in src/sage/interfaces/giac.py : doesn't seem related to my ticket ;
  • one failure on src/sage/doctest/external.py : ditto ;
  • timeout in src/sage/graphs/graph_generators.py : ditto ;
  • one failure in src/sage/schemes/elliptic_curves/heegner.py and one timeout in src/sage/coding/linear_code.py : ditto.

These failures seems "too random" to have a systematic cause, and may result from a race condition somewhere in the test infrastructure. We have been plagued with such "transient" failures for some time now.

Seems good to review...

[rws]

Replying to charpent:

• Most of the 17 patchbots seem to like it, but all but one of them complain that "a plugin failed" ; it seems that this is the "coverage" plugin. Not my watch, I'm afraid...

The coverage plugin (of patchbot) looks if you added code without doctest. It uses the call sage --coverage which you can do too. You fixed it however when I asked you so the recent patchbot runs no longer show it. It's all fine now.