[with patch, positive review] move basic types to new coercion model

[robertwb]

This ticket depends on #4058

[robertwb]

Patches rebased against 3.1.2.

[mhansen]

Looks good and passes tests for me. Excellent work getting this patch made!

[mabshoff]

Robert,

unfortunately I cannot get the patches to apply cleanly and there are various rejects. Sage 3.1.3.alpha1 should be out in a few hours. Could you please rebase them against that release and I will make sure they go in first.

Cheers,

Michael

[robertwb]

Argh... I just rebased these against 3.1.2...

OK, I'll rebase them again (shouldn't be too hard).

[robertwb]

I got a single reject in 4111-coerce-RR.patch, which I resolved, and some fuzz in 4111-coerce-QQ.patch, which I'm reposting. These should all apply fine against 3.1.3.alpha1 now.

[robertwb]

against 3.1.3.alpha1

[robertwb]

against alpha 3.1.3.alpha1

[robertwb]

The order to apply is QQ, RR, simplify, CC.

[mabshoff]

Robert,

I merged the four patches and the only issue is the following:

sage -t -long devel/sage/sage/rings/real_mpfr.pyx           
**********************************************************************
File "/scratch/mabshoff/release-cycle/sage-3.1.3.alpha2/tmp/real_mpfr.py", line 1900:
    sage: RR(-1.234567)._pari_()
Expected:
    -1.2345670000000000000
Got:
    -1.2345670000000000001
**********************************************************************
1 items had failures:
   1 of  11 in __main__.example_58
***Test Failed*** 1 failures.
For whitespace errors, see the file /scratch/mabshoff/release-cycle/sage-3.1.3.alpha2/tmp/.doctest_real_mpfr.py
         [10.8 s]

I don't changing the doctest is the solution here since the "Real Number Inputs Improved" patch from 3.1.2 was supposed to fix this. Or am I completely wrong here? Either way: feel free to add a patch here or in case it is something more than non-trivial open another ticket.

Cheers,

Michael

[mabshoff]

Merged 4111-coerce-simplify.patch, 4111-coerce-CC.patch, 4111-coerce-QQ.2.patch and 4111-coerce-RR.2.patch in Sage 3.1.3.alpha2

[robertwb]

Oh, I think I know what the issue is (we have two real fields of 53 bits prec). I'll post a patch that addresses this.

sage: (1.2).parent() is RR
 False

[AlexGhitza]

Also, the patches at #4096 (which I will rebase against 3.1.3.alpha1 as soon as I have built it) will take care of this.

[robertwb]

Actually, after those patches I get

sage: 1.2.parent() is RR
True

Sit looks like a pari issue for sure. For example, I get

sage: (-1.23456)._pari_()
-1.2345600000000000001

which isn't using RR.__call__ at all.

[mabshoff]

I opened #4199 to make sure the issue does not escape us. I will test the pari patch and see if it fixes the problem.

Cheers,

Michael