Upgrade to Pynac-0.6.5

[rwst]

That Pynac version has:

• fix for: assume(x>0) sets integer flag (assume(x>0) sets integer flag #20456)
• fix Maxima crashes by implementing fast Hermite (ECL crash with Hermite polynomials #20297) and Gegenbauer (crash with ultraspherical polynomials #20428) polynomials in Pynac
• some fixes to handling of elements of rings with positive characteristics (properties of converted finite field elements are wrong #20162)
• Python 3 build fix
• automatic trigonometric and log simplifications

https://github.com/pynac/pynac/releases/download/pynac-0.6.5/pynac-0.6.5.tar.bz2

Depends on #20312

Component: packages: standard

Author: Ralf Stephan, Benjamin Hackl, Aaditya Thakkar

Branch/Commit: 72d96a5

Reviewer: Eric Gourgoulhon, Volker Braun

Issue created by migration from https://trac.sagemath.org/ticket/20475

[rwst]

Branch: u/rws/upgrade_to_pynac_0_6_5

[rwst]

Author: Ralf Stephan, Benjamin Hackl, Aaditya Thakkar

[rwst]

New commits:

39b85b0	new version/chksum
11e2b78	doctest changes and a few fixes in support of pynac-0.6.5

[rwst]

Commit: 11e2b78

[jdemeyer]

comment:3

I consider this a bug:

         sage: S.<y> = PolynomialRing(RR)
         sage: hermite(3,y)
         8*y^3 - 12*y

[jdemeyer]

comment:4

Does Sage use Ginac for other polynomial functions?

[rwst]

Dependencies: #20312

[rwst]

comment:5

Replying to @jdemeyer:

I consider this a bug:

         sage: S.<y> = PolynomialRing(RR)
         sage: hermite(3,y)
         8*y^3 - 12*y

Indeed, see #20312.

Does Sage use Ginac for other polynomial functions?

No. So?

[jdemeyer]

comment:6

Replying to @rwst:

No. So?

Because we should look for a solution which would work generally for all polynomial symbolic functions.

For this ticket, I suggest to remove the changes to the hermite function since that's probably not easily fixed.

[rwst]

comment:7

Replying to @jdemeyer:

Replying to @rwst:

No. So?

Because we should look for a solution which would work generally for all polynomial symbolic functions.

The problem has nothing to do with just polynomial symbolic functions. See the example in #20312 or #20060.

For this ticket, I suggest to remove the changes to the hermite function since that's probably not easily fixed.

I intend to try to fix this (EDIT: I mean #20312), so any hints from you would be welcome.

[sagetrac-git]

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

e06a26c	Merge branch 'develop' into t/20475/upgrade_to_pynac_0_6_5
eb5bf93	20312: preserving function arg parent, first version
8f91703	Merge branch 'u/rws/parent_of_argument_lost_with_functions' of git://trac.sagemath.org/sage into t/20475/upgrade_to_pynac_0_6_5
1986ca5	fix to make fixed doctest pass
5c377cc	20312: fix for constant results
024924e	Merge branch 'u/rws/parent_of_argument_lost_with_functions' of git://trac.sagemath.org/sage into t/20475/upgrade_to_pynac_0_6_5

[sagetrac-git]

Changed commit from 11e2b78 to 024924e

[sagetrac-git]

Changed commit from 024924e to 72d96a5

[sagetrac-git]

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

72d96a5

Merge branch 'develop' into t/20475/upgrade_to_pynac_0_6_5

[egourgoulhon]

comment:11

I gave a try and it looks good to me. It would be nice if this is merged in Sage 7.2 before the final release, to correct #20456 (a rather severe bug IMHO). @jdemeyer: do you agree to set the ticket to positive review?

[egourgoulhon]

Reviewer: Eric Gourgoulhon

[egourgoulhon]

comment:13

sage -coverage says:

SCORE src/sage/symbolic/pynac.pyx: 97.4% (37 of 38)

Missing doctests:
     * line 2271: def init_function_table()

[rwst]

comment:14

Come on, we didn't have a doctest for this since it was added 6 years ago and I don't think we need one. Most of these functions aren't used anyway because the Function code already bypasses Pynac with FP evaluation, and Pynac itself does not need it.

[egourgoulhon]

comment:15

Replying to @rwst:

Come on, we didn't have a doctest for this since it was added 6 years ago and I don't think we need one. Most of these functions aren't used anyway because the Function code already bypasses Pynac with FP evaluation, and Pynac itself does not need it.

OK, I see.

[vbraun]

Changed reviewer from Eric Gourgoulhon to Eric Gourgoulhon, Volker Braun

[vbraun]

Changed branch from u/rws/upgrade_to_pynac_0_6_5 to 72d96a5