Fix complex handling of libs/pynac/py_float()

[rwst]

The goal is to always return some float, either real or complex but:

sage: from sage.libs.pynac.pynac import py_float_for_doctests as py_float
sage: py_float(I, {'parent':RealField(10)})
...
TypeError: unable to convert '1.0*I' to a real number

The code shows it's only done if there is no parent given:

    if kwds is not NULL:
        return (<object>kwds)['parent'](n)
    else:
        try:
            return RR(n)
        except TypeError:
            return CC(n)

The ticket should make it possible to fall back on the associated complex field of any parent.

Depends on #23145

Component: interfaces

Author: Ralf Stephan

Branch/Commit: 856f8dd

Reviewer: Travis Scrimshaw

Issue created by migration from https://trac.sagemath.org/ticket/23077

[rwst]

Branch: u/rws/fix_complex_handling_of_libs_pynac_py_float__

[rwst]

comment:2

Still one doctest to fix. #18386 depends on this.

---

New commits:

cb88513

23077: Fix complex handling of libs/pynac/py_float()

[rwst]

Commit: cb88513

[rwst]

comment:3

BTW, the numeric doctest changes have better accuracy, so consider it a bug fix for inaccurate results as well.

[sagetrac-git]

Changed commit from cb88513 to b5c0864

[sagetrac-git]

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

b5c0864

23077: fix jacobi_P numerics

[rwst]

Author: Ralf Stephan

[tscrim]

comment:6

Is it ever (natually) possible for the parent to not have a complex_field method?

[rwst]

comment:7

Replying to @tscrim:

Is it ever (natually) possible for the parent to not have a complex_field method?

Try RLF.

[tscrim]

comment:8

So could we get into a situation where this raises an AttributeError:

diff --git a/src/sage/functions/orthogonal_polys.py b/src/sage/functions/orthogonal_polys.py
index 931b247..51da378 100644
--- a/src/sage/functions/orthogonal_polys.py
+++ b/src/sage/functions/orthogonal_polys.py
@@ -1942,7 +1942,10 @@ class Func_jacobi_P(OrthogonalFunction):
         prec = the_parent.precision()
         BF = CBF(prec+5)
         ret = BF(x).jacobi_P(BF(n), BF(a), BF(b))
-        return the_parent(ret)
+        try:
+            return the_parent(ret)
+        except (TypeError, ValueError):
+            return the_parent.complex_field()(ret)
 
 jacobi_P = Func_jacobi_P()
 

Should we catch the error?