Upgrade to pynac-0.7.15

[rwst]

In 0.7.15:

• fix mpz conversion (Doctest: bug numerical_approx(2^(450232897/4888643760)) #24418)
• fix infinity.info(), function.info(), power.info(negative)
• fix symbol.real/imag_part()
• fix bug showing with nonnumeric rationals
• fix Python import from sage.rings
• rational symbol domain (Rational assumption / variable domain / query #24368)
• implement constant.is_nonzero() (Doctest: Expression.is_positive/negative fixed #18630)
• fully implement info(even) (Even assumption / variable domain / query #24351)
• fully implement info(nonzero)
• fully implement info(rational)

https://github.com/pynac/pynac/releases/download/pynac-0.7.15/pynac-0.7.15.tar.bz2

CC: @kiwifb

Component: packages: standard

Author: Ralf Stephan

Branch/Commit: d085587

Reviewer: Jeroen Demeyer

Issue created by migration from https://trac.sagemath.org/ticket/24497

[rwst]

Branch: u/rws/upgrade_to_pynac_0_7_15

[rwst]

Commit: 880ebf3

[rwst]

New commits:

907aff6	24497: pkg version/chksum
880ebf3	24497: doctest fixes

[rwst]

Author: Ralf Stephan

[videlec]

comment:5

All tests pass on quasar.

[tscrim]

comment:6

The only thing I'm not sure about is the change in modular/modform_hecketriangle/graded_ring_element.py. Can you explain the rationale?

[rwst]

comment:7

The former result changes quite a bit, I think because small changes are enlarged by substitution into a polynomial. I tried to preserve the assumed spirit of the test, i.e. a nonzero real part of the result, but to make it more robust. Of course I may be wrong with my assumptions.

[jdemeyer]

comment:8

Passes tests on sardonis (ppc64le) too.

[jdemeyer]

Changed branch from u/rws/upgrade_to_pynac_0_7_15 to u/jdemeyer/upgrade_to_pynac_0_7_15

[jdemeyer]

Changed commit from 880ebf3 to 70275ab

[jdemeyer]

comment:10

Replying to @rwst:

The former result changes quite a bit, I think because small changes are enlarged by substitution into a polynomial. I tried to preserve the assumed spirit of the test, i.e. a nonzero real part of the result, but to make it more robust. Of course I may be wrong with my assumptions.

I kept the test but adjusted the answer. Does that work for you?

---

New commits:

70275ab

24497: doctest fixes

[rwst]

comment:11

Replying to @jdemeyer:

I kept the test but adjusted the answer. Does that work for you?

Yes.

[jdemeyer]

Reviewer: Jeroen Demeyer

[rwst]

comment:13

Thanks for the review.

[vbraun]

comment:14

On Ubuntu 16 32-bit:

**********************************************************************
File "src/sage/modular/modform_hecketriangle/graded_ring_element.py", line 1947, in sage.modular.modform_hecketriangle.graded_ring_element.?.evaluate
Failed example:
    f_i(i + 1e-1000)  # rel tol 1e-3
Expected:
    -5.96189764223709e-14 - 4.10147008296517e-1000*I
Got:
    -6.08402217494586e-14 - 4.10147008296517e-1000*I
Tolerance exceeded in 1 of 2:
    -5.96189764223709e-14 vs -6.08402217494586e-14, tolerance 2e-02 > 1e-03
**********************************************************************
1 item had failures:
   1 of 132 in sage.modular.modform_hecketriangle.graded_ring_element.?.evaluate
    [671 tests, 1 failure, 40.96 s]

[rwst]

comment:15

So you see how useless the test is in this form. I'm replacing it with my version.

[rwst]

Changed branch from u/jdemeyer/upgrade_to_pynac_0_7_15 to u/rws/24497

[rwst]

Changed commit from 70275ab to 7754ab0

[rwst]

comment:17

It's just a test.

---

New commits:

7754ab0

24497: doctest fixes

[jdemeyer]

comment:18

Replying to @rwst:

So you see how useless the test is in this form. I'm replacing it with my version.

Not useless. We just need to increase the tolerance.

[rwst]

Changed branch from u/rws/24497 to u/rws/24497-1

[rwst]

Changed commit from 7754ab0 to d085587

[rwst]

New commits:

d085587

24497: doctest fixes

[vbraun]

Changed branch from u/rws/24497-1 to d085587