Hi Pushkar,
I can't say whether the details of your proposed fix are reasonable as
I have not looked through the relevant code in solveset.
What I can say is that if things were properly designed and
implemented then I don't think that there would be any circumstances
in which solveset_real ends up calling solveset_complex. I suspect
that the real fix here is not catching any exception but rather
working through carefully to see how this happens and ensuring that
real solvers are kept separate from the complex ones.
Oscar
On Mon, 7 Jul 2025 at 14:46, pushkar sharma <pushkarsharma....@gmail.com> wrote:
>
> Hi everyone,
>
> I'm Pushkar, and this is my first time contributing to SymPy. I'm still
> learning my way around the codebase and contribution process, so please let
> me know if anything I’m doing is not quite right.
>
> I came across SymPy issue #28213, where calling solveset_real on an
> expression involving Abs and trigonometric functions raises a ValueError,
> even though all variables are declared real.
>
> Here’s a minimal reproducible example that fails:
>
> ```
>
> from sympy import Abs, sin, cos, symbols
> from sympy.solvers.solveset import solveset_real
>
> t1, t2, t3, u1, u2, u3 = symbols('t1 t2 t3 u1 u2 u3', real=True)
> expr = sin(t1)*cos(t3) + sin(t2)*sin(t3)*cos(t1) - (
> sin(t2)*sin(u3)*cos(t1)*cos(t2)/Abs(cos(t2)) + cos(u3)*Abs(sin(t1))
> )
> solveset_real(expr, u3)
>
> ```
>
> This raises:
>
> ```
>
> ValueError: Absolute values cannot be inverted in the complex domain.
>
> ```
>
> I looked into it and noticed that _solve_trig1 in solveset_real is falling
> back to solveset_complex in cases involving Abs, even though the domain is
> real. To fix this, I modified _solve_abs so that instead of raising a
> ValueError when inequalities fail, it now returns a ConditionSet fallback.
>
>
> My Fix:
>
> I modified _solve_abs to handle NotImplementedError during inequality solving
> by wrapping the fallback in a ConditionSet, like this:
>
>
> ```
>
> def _solve_abs(f, symbol, domain):
> ...
> if not (f_p.is_zero or f_q.is_zero):
> ...
> try:
> q_pos_cond = solve_univariate_inequality(...)
> ...
> return Union(sols_q_pos, sols_q_neg)
> except NotImplementedError:
> return ConditionSet(symbol, Eq(f, 0), domain)
> else:
> return ConditionSet(symbol, Eq(f, 0), domain)
> ```
>
>
> Test Results:
>
> After applying the patch, I ran:
>
> ```
> python -m pytest -s sympy/solvers/tests/test_solveset.py
> ```
>
> and got the following:
>
> 180 tests passed
>
> 2 failed (but they are unrelated to this patch — test_solve_trig and
> test_issue_19050)
>
> 2 skipped, 1 deselected, 13 xfailed (as expected)
>
> I checked the failures, and they seem to stem from existing limitations in
> solving sin(x) + cos(x) exactly, not related to my changes.
>
> I'm happy to open a PR if this approach seems reasonable. But since I’m new,
> I wanted to get feedback here first. Would love any suggestions or
> corrections.
>
> Thanks,
> Pushkar
>
>
>
>
>
>
>
>
>
> --
> You received this message because you are subscribed to the Google Groups
> "sympy" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sympy+unsubscr...@googlegroups.com.
> To view this discussion visit
> https://groups.google.com/d/msgid/sympy/1cea363f-e0fe-4196-89e3-9c9bd954616dn%40googlegroups.com.
--
You received this message because you are subscribed to the Google Groups
"sympy" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to sympy+unsubscr...@googlegroups.com.
To view this discussion visit
https://groups.google.com/d/msgid/sympy/CAHVvXxTN3NcWMT5R5a7kE043nMG-d720o%3DihqzR7S%2B%3DDXa6yGQ%40mail.gmail.com.
