sage.rings.real_double: Move methods using GSL to a separate extension module

[mkoeppe]

Discussion: https://groups.google.com/g/sage-devel/c/HTN6KoqZVJQ/m/2BzAx5erBAAJ

We first replace use of some gsl functions like gsl_isnan by their C standard library counterparts, as provided by Cython's libc.math (
https://github.com/cython/cython/blob/master/Cython/Includes/libc/math.pxd)

Likewise for some uses of these functions in src/sage/plot/plot3d/implicit_surface.pyx

We move the remaining methods that use gsl functions to a new subclass RealDoubleElement_gsl, in the new module sage.rings.real_double_element_gsl.

RealDoubleElement already uses a nonstandard allocator for speed. We adjust it so that it actually creates an instance of RealDoubleElement_gsl when the module sage.rings.real_double_element_gsl can be imported.

(Making similar changes to ComplexDoubleElement is beyond the scope of this ticket. This is more complicated because gsl_complex may be a type separate from the C99 double complex type, see https://www.gnu.org/software/gsl/doc/html/complex.html)

Depends on #32682
Depends on #32675

CC: @dimpase @kiwifb @williamstein @tscrim

Component: refactoring

Author: Matthias Koeppe

Branch/Commit: 520da8f

Reviewer: François Bissey

Issue created by migration from https://trac.sagemath.org/ticket/32677

[mkoeppe]

Description changed:

--- 
+++ 
@@ -1,3 +1,4 @@
 Discussion: https://groups.google.com/g/sage-devel/c/HTN6KoqZVJQ/m/2BzAx5erBAAJ
 
-
+We first replace use of gsl functions like `gsl_isnan` by their C standard library counterparts, as provided by Cython's `libc.math` (
+https://github.com/cython/cython/blob/master/Cython/Includes/libc/math.pxd)

[mkoeppe]

Branch: u/mkoeppe/sage_rings_real_double__move_methods_using_gsl_to_a_separate_extension_module

[mkoeppe]

Commit: f04ad5a

[mkoeppe]

comment:3

This simple change already breaks stuff.... what am I doing wrong?

---

New commits:

f04ad5a

src/sage/rings/real_double.pyx: Replace gsl_{isnan,isinf,finite} by libc.math functions

[kiwifb]

comment:4

Are you doing an incremental build? I pointed sage-on-gentoo to your commit and so far it has compiled sage for python3.8 and python3.9 is on its way. I won't know about doc building breakages without being a bit more involved.

[kiwifb]

comment:5

Installing "sagemath-standard" works here as far as I can tell.

[mkoeppe]

comment:6

Yes, it compiles but doctests fail:

sage -t --random-seed=0 src/sage/rings/infinity.py  # 4 doctests failed
sage -t --random-seed=0 src/sage/rings/complex_double.pyx  # 4 doctests failed
sage -t --random-seed=0 src/sage/rings/real_double.pyx  # 10 doctests failed

[mkoeppe]

Description changed:

--- 
+++ 
@@ -1,4 +1,6 @@
 Discussion: https://groups.google.com/g/sage-devel/c/HTN6KoqZVJQ/m/2BzAx5erBAAJ
 
-We first replace use of gsl functions like `gsl_isnan` by their C standard library counterparts, as provided by Cython's `libc.math` (
+We first replace use of some gsl functions like `gsl_isnan` by their C standard library counterparts, as provided by Cython's `libc.math` (
 https://github.com/cython/cython/blob/master/Cython/Includes/libc/math.pxd)
+
+

[kiwifb]

comment:9

I only get those

sage -t --long --random-seed=0 /usr/lib/python3.9/site-packages/sage/rings/complex_double.pyx  # 2 doctests failed
sage -t --long --random-seed=0 /usr/lib/python3.9/site-packages/sage/rings/real_double.pyx  # 2 doctests failed

And it looks like conversion from boolean to integer issues

fbissey@tarazed ~ $ sage -t --long --random-seed=0 /usr/lib/python3.9/site-packages/sage/rings/complex_double.pyx
too many failed tests, not using stored timings
Running doctests with ID 2021-10-13-13-12-48-70eba795.
Using --optional=dochtml,pip,sage
Doctesting 1 file.
sage -t --long --random-seed=0 /usr/lib/python3.9/site-packages/sage/rings/complex_double.pyx
**********************************************************************
File "/usr/lib/python3.9/site-packages/sage/rings/complex_double.pyx", line 1626, in sage.rings.complex_double.ComplexDoubleElement.is_infinity
Failed example:
    CDF(1, 2).is_infinity()
Expected:
    False
Got:
    0
**********************************************************************
File "/usr/lib/python3.9/site-packages/sage/rings/complex_double.pyx", line 1628, in sage.rings.complex_double.ComplexDoubleElement.is_infinity
Failed example:
    CDF(0, oo).is_infinity()
Expected:
    True
Got:
    1
**********************************************************************
1 item had failures:
   2 of   3 in sage.rings.complex_double.ComplexDoubleElement.is_infinity
    [345 tests, 2 failures, 0.18 s]
----------------------------------------------------------------------
sage -t --long --random-seed=0 /usr/lib/python3.9/site-packages/sage/rings/complex_double.pyx  # 2 doctests failed
----------------------------------------------------------------------
Total time for all tests: 0.3 seconds
    cpu time: 0.2 seconds
    cumulative wall time: 0.2 seconds
Pytest is not installed, skip checking tests that rely on it.
fbissey@tarazed ~ $ sage -t --long --random-seed=0 /usr/lib/python3.9/site-packages/sage/rings/real_double.pyx
too many failed tests, not using stored timings
Running doctests with ID 2021-10-13-13-13-13-b89854b8.
Using --optional=dochtml,pip,sage
Doctesting 1 file.
sage -t --long --random-seed=0 /usr/lib/python3.9/site-packages/sage/rings/real_double.pyx
**********************************************************************
File "/usr/lib/python3.9/site-packages/sage/rings/real_double.pyx", line 1730, in sage.rings.real_double.RealDoubleElement.is_infinity
Failed example:
    (a/b).is_infinity()
Expected:
    True
Got:
    1
**********************************************************************
File "/usr/lib/python3.9/site-packages/sage/rings/real_double.pyx", line 1732, in sage.rings.real_double.RealDoubleElement.is_infinity
Failed example:
    (b/a).is_infinity()
Expected:
    False
Got:
    0
**********************************************************************
1 item had failures:
   2 of   4 in sage.rings.real_double.RealDoubleElement.is_infinity
    [454 tests, 2 failures, 0.15 s]
----------------------------------------------------------------------
sage -t --long --random-seed=0 /usr/lib/python3.9/site-packages/sage/rings/real_double.pyx  # 2 doctests failed
----------------------------------------------------------------------
Total time for all tests: 0.2 seconds
    cpu time: 0.1 seconds
    cumulative wall time: 0.2 seconds

[sagetrac-git]

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

d3caeb4

src/sage/rings/real_double.pyx: Fixup: Replace gsl_{isnan,isinf,finite} by libc.math functions

[sagetrac-git]

Changed commit from f04ad5a to d3caeb4

[mkoeppe]

Description changed:

--- 
+++ 
@@ -3,4 +3,6 @@
 We first replace use of some gsl functions like `gsl_isnan` by their C standard library counterparts, as provided by Cython's `libc.math` (
 https://github.com/cython/cython/blob/master/Cython/Includes/libc/math.pxd)
 
+Likewise for some uses of these functions in `src/sage/plot/plot3d/implicit_surface.pyx`
 
+

[sagetrac-git]

Changed commit from d3caeb4 to 31d188f

[sagetrac-git]

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

e340b2c	sage.misc.allocator: Add cdef hook_tp_functions_type
31d188f	sage.rings.real_double_element_gsl: Split out from sage.rings.real_double

[mkoeppe]

Author: Matthias Koeppe