Remove outdated mpmath Sage backend

[mkoeppe]

This is the first step of preparation for mpmath 1.4

• see Remove sage backend mpmath/mpmath#732
• Fixes Incompatibilities in Sage mpmath cause downstream issues in SymPy #25445
• Fixes mpmath.mp.mpf has outdated __float__ dunder #38223

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.
• I have created tests covering the changes.
• I have updated the documentation and checked the documentation preview.

⌛ Dependencies

[github-actions]

Documentation preview for this PR (built with commit 6ddabe9; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[kwankyu]

@asmeurer

[kiwifb]

I get plenty of doctest errors of the kind

mpmath_create() expects 'mpz','int'[,'int','str'] arguments

like in

sage -t --long --random-seed=116790357055389508556504894381735541723 /usr/lib/python3.12/site-packages/sage/symbolic/function.pyx
**********************************************************************
File "/usr/lib/python3.12/site-packages/sage/symbolic/function.pyx", line 960, in sage.symbolic.function.BuiltinFunction.?
Failed example:
    cos(mpmath.mpc(1,1))
Exception raised:
    Traceback (most recent call last):
      File "/usr/lib/python3.12/site-packages/sage/doctest/forker.py", line 716, in _run
        self.compile_and_execute(example, compiler, test.globs)
      File "/usr/lib/python3.12/site-packages/sage/doctest/forker.py", line 1146, in compile_and_execute
        exec(compiled, globs)
      File "<doctest sage.symbolic.function.BuiltinFunction.?[10]>", line 1, in <module>
        cos(mpmath.mpc(Integer(1),Integer(1)))
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "/usr/lib/python3.12/site-packages/mpmath/ctx_mp_python.py", line 562, in __new__
        r_real, r_imag = map(cls.context.mpf, [r_real, r_imag])
        ^^^^^^^^^^^^^^
      File "/usr/lib/python3.12/site-packages/mpmath/ctx_mp_python.py", line 69, in __new__
        val = cls.mpf_convert_arg(val, prec, rounding, base)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "/usr/lib/python3.12/site-packages/mpmath/ctx_mp_python.py", line 79, in mpf_convert_arg
        if isinstance(x, numbers.Rational): return from_rational(x.numerator,
                                                   ^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "/usr/lib/python3.12/site-packages/mpmath/libmp/libmpf.py", line 406, in from_rational
        return mpf_div(from_int(p), from_int(q), prec, rnd)
                       ^^^^^^^^^^^
      File "/usr/lib/python3.12/site-packages/mpmath/libmp/libmpf.py", line 249, in from_int
        return from_man_exp(n, 0, prec, rnd)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    TypeError: mpmath_create() expects 'mpz','int'[,'int','str'] arguments
**********************************************************************

[mkoeppe]

If you use your own version of src/sage/env.py or set MPMATH_* variables elsewhere, check if you have made the changes made here

[kiwifb]

I do not have any changes to env.py anymore. I have nothing about mpmath in sage_conf. From a sage session

fbissey@tarazed ~ $ sage
┌────────────────────────────────────────────────────────────────────┐
│ SageMath version 10.5.beta2, Release Date: 2024-08-10              │
│ Using Python 3.12.5. Type "help()" for help.                       │
└────────────────────────────────────────────────────────────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ Warning: this is a prerelease version, and it may be unstable.     ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛
sage: import sage.env
sage: sage.env.MPMATH_SAGE
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[2], line 1
----> 1 sage.env.MPMATH_SAGE

AttributeError: module 'sage.env' has no attribute 'MPMATH_SAGE'
sage: sage.env.MPMATH_NOSAGE
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[3], line 1
----> 1 sage.env.MPMATH_NOSAGE

AttributeError: module 'sage.env' has no attribute 'MPMATH_NOSAGE'
sage: import os
sage: os.getenv("MPMATH_SAGE")
sage: os.getenv("MPMATH_NOSAGE")
'1'

I will try some test runs with those variables defined that way in the running shel.

[kiwifb]

No luck and it happens outside of the testing framework too

fbissey@tarazed ~ $ sage
┌────────────────────────────────────────────────────────────────────┐
│ SageMath version 10.5.beta2, Release Date: 2024-08-10              │
│ Using Python 3.12.5. Type "help()" for help.                       │
└────────────────────────────────────────────────────────────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ Warning: this is a prerelease version, and it may be unstable.     ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛
sage: import os
sage: os.getenv("MPMATH_SAGE")
sage: os.getenv("MPMATH_NOSAGE")
'1'
sage: import mpmath
sage: os.getenv("MPMATH_SAGE")
sage: os.getenv("MPMATH_NOSAGE")
'1'
sage: cos(mpmath.mpf('1.321412'))
mpf('0.24680737898640387')
sage: cos(mpmath.mpc(1,1))
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[8], line 1
----> 1 cos(mpmath.mpc(Integer(1),Integer(1)))

File /usr/lib/python3.12/site-packages/mpmath/ctx_mp_python.py:562, in _mpc.__new__(cls, real, imag)
    560 else:
    561     i_real, i_imag = imag, 0
--> 562 r_real, r_imag = map(cls.context.mpf, [r_real, r_imag])
    563 i_real, i_imag = map(cls.context.mpf, [i_real, i_imag])
    564 real = r_real - i_imag

File /usr/lib/python3.12/site-packages/mpmath/ctx_mp_python.py:69, in _mpf.__new__(cls, val, **kwargs)
     67         raise ValueError
     68 else:
---> 69     val = cls.mpf_convert_arg(val, prec, rounding, base)
     70 v._mpf_ = mpf_pos(val, prec, rounding)
     71 return v

File /usr/lib/python3.12/site-packages/mpmath/ctx_mp_python.py:79, in _mpf.mpf_convert_arg(cls, x, prec, rounding, base)
     77 if isinstance(x, str): return from_str(x, prec, rounding, base)
     78 if isinstance(x, cls.context.constant): return x.func(prec, rounding)
---> 79 if isinstance(x, numbers.Rational): return from_rational(x.numerator,
     80                                                          x.denominator,
     81                                                          prec, rounding)
     82 if hasattr(x, '_mpf_'): return x._mpf_
     83 if hasattr(x, '_mpmath_'):

File /usr/lib/python3.12/site-packages/mpmath/libmp/libmpf.py:406, in from_rational(p, q, prec, rnd)
    403 def from_rational(p, q, prec, rnd=round_fast):
    404     """Create a raw mpf from a rational number p/q, round if
    405     necessary."""
--> 406     return mpf_div(from_int(p), from_int(q), prec, rnd)

File /usr/lib/python3.12/site-packages/mpmath/libmp/libmpf.py:249, in from_int(n, prec, rnd)
    247     if n in int_cache:
    248         return int_cache[n]
--> 249 return from_man_exp(n, 0, prec, rnd)

TypeError: mpmath_create() expects 'mpz','int'[,'int','str'] arguments

[mkoeppe]

Is this with the same mpmath version as in the Sage distribution?

[kiwifb]

I was going to comment about that. I use mpmath 1.4.0a1 to get python 3.13 support. So no, but it is indicative of future behavior.
Additionally, I need the stuff here to build the documentation with that version of mpmath.

[kiwifb]

In this version of mpmath, the only thing tested is MPMATH_NOGMPY - I am now wondering if it is a mpmath bug and not something we can correct in sage.

[mkoeppe]

I have so far not tested with anything in the 1.4 development series.
For this PR, I would set an upper bound mpmath <1.4

[kiwifb]

OK, I will try to get to the bottom of this separately. It looks like coercion from sage integer to something mpmath likes is at fault.

[kiwifb]

For the record, I dug a bit deeper and the message is not quite what I thought it was about. It ultimately comes from gmpy2 (and I thought for a moment that it could be a change in gmpy 2.2.1). A closer examination of the code shows that this particular complaint is displayed when less than two arguments are passed to mpmath_create. On the other hand doing the same operation (mpmath.mpc(1, 1)) in python rather than sage does works.

It literally looks like sage integers (real numbers are fine) are reduced to "None" in mpmath 1.4a.

[mkoeppe]

Thanks for the update. I've opened #38565 to reproduce this, and as a possible target PR for an continuous integration workflow for mpmath

[skirpichev]

Maybe you could include tests? // #36447

[mkoeppe]

Maybe you could include tests? // #36447

Good idea, I've cherry-picked one of your commits there