Do not use pynac's poly_mul_expand function until it has been debugged

[DaveWitteMorris]

Pynac's poly_mul_expand function silently produces wrong results when expanding certain sums. This ticket eliminates its use by pynac's ex::expand method, and therefore provides a (temporary) fix for the problems reported in #31077 and #31411.

This is part of metaticket #31478.

Use of the poly_mul_expand function can be re-enabled after the bugs have been fixed.

CC: @egourgoulhon @mjungmath

Component: symbolics

Keywords: pynac, expand

Author: Dave Morris

Branch/Commit: eeb6cc2

Reviewer: Volker Braun, Matthias Koeppe

Issue created by migration from https://trac.sagemath.org/ticket/31479

[DaveWitteMorris]

Branch: public/31479

[DaveWitteMorris]

Commit: 3dfe9a6

[DaveWitteMorris]

New commits:

371e10d	trac 31479: pynac's poly_mul_expand
3dfe9a6	add doctests

[mkoeppe]

comment:3

Patchbot indicates doctest failures:

sage -t --long --warn-long 68.0 --random-seed=0 src/sage/symbolic/expression.pyx
**********************************************************************
File "src/sage/symbolic/expression.pyx", line 4846, in sage.symbolic.expression.Expression.expand
Failed example:
    ((a + b + c)^30 * (3*b + d - 5/d)^3).expand().subs(a=0,b=2,c=-1)
Expected:
    d^3 + 18*d^2 + 93*d - 465/d + 450/d^2 - 125/d^3 + 36
Got:
    -124*d^3 + 468*d^2 - 372*d + 36
**********************************************************************
File "src/sage/symbolic/expression.pyx", line 4854, in sage.symbolic.expression.Expression.expand
Failed example:
    bool((A * B).expand() == (A * B.expand()).expand())
Expected:
    True
Got:
    False
**********************************************************************

[DaveWitteMorris]

comment:4

The doctests that failed are the new ones that are added by this ticket. They will fail without the pynac patch that is on this ticket, but will pass if the patch is applied.

I am pretty sure that the patchbot did not apply the patch to pynac. Is there something I should do to make the patchbot apply the patch?

[DaveWitteMorris]

comment:5

Doctesting this ticket on MacOS 10.15.7 with

sage -f pynac
make
sage -t -p 4 --all --long

resulted in All tests passed!. I also tested on Ubuntu 20.04 (CoCalc) and got no failures that I would attribute to this ticket. So I think we can consider the badge to be green, even though the patchbots aren't testing the pynac patch.

[vbraun]

comment:8

The patch level in build/pkgs/pynac/package-version.txt needs to be incremented so the build system knows that pynac must be rebuilt...

[DaveWitteMorris]

Changed branch from public/31479 to public/31479v2

[DaveWitteMorris]

Changed commit from 3dfe9a6 to b6f9170

[DaveWitteMorris]

comment:10

Thanks! I incremented the patch number (and rebased on beta9), so I hope the patchbots will start chiming in.

---

New commits:

270a2f9	trac 31479: pynac's poly_mul_expand
a2ebbeb	add doctests
b6f9170	increment package version

[vbraun]

Reviewer: Volker Braun

[vbraun]

comment:12

On 32-bit

**********************************************************************
File "src/sage/symbolic/expression.pyx", line 4882, in sage.symbolic.expression.Expression.expand
Failed example:
    ((a + b + c)^30 * (3*b + d - 5/d)^3).expand().subs(a=0,b=2,c=-1)
Expected:
    d^3 + 18*d^2 + 93*d - 465/d + 450/d^2 - 125/d^3 + 36
Got:
    d^3 + 18*d^2 + 1739461754973*d - 8697308774865/d + 450/d^2 - 125/d^3 + 36
**********************************************************************
1 item had failures:
   1 of  43 in sage.symbolic.expression.Expression.expand
    [2937 tests, 1 failure, 38.78 s]
----------------------------------------------------------------------
sage -t --long --random-seed=0 src/sage/symbolic/expression.pyx  # 1 doctest failed
----------------------------------------------------------------------

[DaveWitteMorris]

comment:13

The answer is correct modulo 2^32, so it seems clear that this is an integer overflow error. But pynac is not supposed to make that type of error, so I don't know what is going on. It may not be difficult to locate the bug on a machine that reproduces the error, but I don't have one right now.

I propose to mark the 32-bit version of this doctest # known bug and open a new (critical) ticket to fix the overflow error.

[sagetrac-git]

Changed commit from 96b0f8b to 2364860

[sagetrac-git]

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

4bacd76	trac 31554 do not use handle_factor
00b44eb	add doctest
707560a	increment pynac patch number
82843fe	remove debugging code
ed6cac2	Merge branch 'public/31554' of git://trac.sagemath.org/sage into t/31479/public/31479v2
2364860	build/pkgs/pynac/package-version.txt: Bump patch level

[mkoeppe]

Dependencies: #31554

[sagetrac-git]

Changed commit from 2364860 to b0b074e

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1 and set ticket back to needs_review. New commits:

0f6dcc9	Merge tag '9.3.rc2' into t/31554/public/31554
b0b074e	Merge #31479

[vbraun]

comment:23

Circular dependency

[sagetrac-git]

Changed commit from b0b074e to 0f6dcc9

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

[mkoeppe]

Changed dependencies from #31554 to none

[sagetrac-git]

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

eeb6cc2

build/pkgs/pynac/package-version.txt: Bump patch level

[sagetrac-git]

Changed commit from 0f6dcc9 to eeb6cc2

[mkoeppe]

comment:27

Sorry for messing this up.

[vbraun]

Changed branch from public/31479v2 to eeb6cc2