sbml_math_to_sympy fails on powers with a dimensionless numeric base and a symbolic exponent #50

sbml_math_to_sympy cannot parse <power/> with a numeric base that has sbml:units and a non-numeric exponent. set_math (with the default literals_dimensionless=True) writes exactly such MathML for, e.g., 10**x, so the output of set_math cannot be read back:

import libsbml
import sympy as sp
from sbmlmath import sbml_math_to_sympy, set_math

doc = libsbml.SBMLDocument(3, 2)
model = doc.createModel()
ia = model.createInitialAssignment()
ia.setSymbol("y")
set_math(ia, sp.Integer(10) ** sp.Symbol("x"))
print(libsbml.formulaToL3String(ia.getMath()))  # 10 dimensionless^x
sbml_math_to_sympy(ia)

  File ".../sbmlmath/mathml_parser.py", line 54, in _pow
  ...
  File ".../pint/util.py", line 650, in __pow__
TypeError: Cannot power UnitsContainer by <class 'sympy.core.symbol.Symbol'>

The same happens for 2.5**x. x**2 (units on the exponent) works.

The base is converted to a pint quantity, and pint can't raise units to a symbolic power. For a dimensionless base, the result is dimensionless, so the units could be dropped (or the magnitude used) before taking the power. For a base with other units and a non-numeric exponent, the units of the result are not defined; this could raise a clear error, or drop the units with a warning.

Workaround: write the MathML with SBMLMathMLPrinter(literals_dimensionless=False).

Found in libpetab-python (PEtab-dev/libpetab-python#560), where p = 10^log10_p initial assignments are written.

sbmlmath 0.4.1, pint 0.26.1, sympy 1.14.0, python-libsbml 5.21.2, Python 3.12.10

🤖 Generated with Claude Code