Explicitly set MPMATH_SAGE env variable

[antonio-rojas]

Since mpmath 1.2.1, specifically since commit [1], mpmath checks for the SAGE_ROOT or MPMATH_SAGE env variables in order to enable sage types instead of relying on whether sage.all can be imported.

This breaks distro packages where SAGE_ROOT is undefined. I reported this upstream a while ago [2] with not much luck. Here we explicitly define MPMATH_SAGE in sage-env so mpmath uses Sage types in a Sage session even if SAGE_ROOT is not defined.

This is still broken when using sage as a python library (in which case sage-env is not read at all), but fixing that use case is beyond the scope of this ticket.

[1] mpmath/mpmath@84ca137
[2] mpmath/mpmath#574

CC: @mkoeppe @kiwifb @dimpase @isuruf

Component: packages: standard

Author: Antonio Rojas

Branch/Commit: d4b1b60

Reviewer: François Bissey

Issue created by migration from https://trac.sagemath.org/ticket/32799

[antonio-rojas]

Branch: u/arojas/explicitly_set_mpmath_sage_env_variable

[antonio-rojas]

Author: Antonio Rojas

[antonio-rojas]

New commits:

c2cde67

Explicitly set MPMATH_SAGE env variable

[antonio-rojas]

Description changed:

--- 
+++ 
@@ -1 +1,8 @@
+Since mpmath 1.2.1, specifically since commit [1], mpmath checks for the SAGE_ROOT or MPMATH_SAGE env variables in order to enable sage types instead of relying on whether sage.all can be imported.
 
+This breaks distro packages where SAGE_ROOT is undefined. I reported this upstream a while ago [2] with not much luck. Here we explicitly define MPMATH_SAGE in sage-env so mpmath uses Sage types in a Sage session even if SAGE_ROOT is not defined.
+
+This is still broken when using sage as a python library (in which case sage-env is not read at all), but fixing that use case is beyond the scope of this ticket.
+
+[1] https://github.com/fredrik-johansson/mpmath/commit/84ca137ff84f7b3d22cabcf8bfe260146db4defe
+[2] https://github.com/fredrik-johansson/mpmath/issues/574

[antonio-rojas]

Commit: c2cde67

[slel]

comment:3

See also

• Incompatibilities in Sage mpmath cause downstream issues in SymPy #25445: Incompatibilities in Sage mpmath cause downstream issues in SymPy
• sage-devel, 2021-10-17:
  BUG: Si(pi).n() throws error in sage 9.2-2
• sage-devel, 2021-03-18:
  Pointy Python question related to Sage

[kiwifb]

comment:4

This is the minimum to do vanilla sage. I have already endorsed the approach in principle months ago. I am not sure I would check MPMATH_SAGE first since I'd always want it to be 1, unless I explicitly want to break things.

I would like more opinion on that one because not providing a way to override is contentious.

[mkoeppe]

comment:5

We should probably set this environment variable in the Sage Python library, not in the sage-env script.

[kiwifb]

comment:6

Replying to @mkoeppe:

We should probably set this environment variable in the Sage Python library, not in the sage-env script.

I agree with that. If I knew how to do it, there would already be a ticket for it ready to review. The stuff I tried didn't work, the only things I can think of require cooperation from mpmath devs.

[sagetrac-git]

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

d4b1b60

Set the env variable in the Sage python library