gh-145117: fix pprint.isreadable() for non-finite float and complex values

[zetzschest]

Issue

isreadable is documented as returning True if the object's repr is readable by the interpreter — implying eval(pformat(obj)) == obj. For non-finite float and complex values, isreadable returns True, but the roundtrip fails — eval raises NameError since inf and nan aren't valid Python expressions.

Reproducer

import pprint

print(pprint.isreadable(float('inf')))        # True — but eval('inf') raises NameError
print(pprint.isreadable(float('nan')))        # True
print(pprint.isreadable(complex(float('inf'), 0)))  # True

Fix

Two options: update the documentation to weaken the guarantee so isreadable no longer implies the eval roundtrip, or fix the implementation so non-finite float and complex values are correctly marked as not readable. We think the latter is the right call — the documented contract is reasonable and worth preserving. In _safe_repr, use cmath.isfinite() to catch both float and complex non-finite values:

if (typ is float or typ is complex) and not _cmath.isfinite(object):
    return rep, False, False

• Issue: pprint.isreadable returns True for non-eval-able float and complex specials #145117

[python-cla-bot]

All commit authors signed the Contributor License Agreement.