Expreval, A Minimal, robust Python library for any math expressions

Changelog

Minimal. Single function. No dependencies. You give a string with a numeric expression; it returns a float.

pip install expreval

from expreval import evaluate
evaluate("sin(pi/2) + log(e)")  # 2.0

CLI:

expreval "sin(pi/4)**2"

• Numbers: int / float literals
• Operators: + - * / % ** and unary + -
• Grouping: ( )
• Names: pi, e, any function from the standard math module (sin, cos, sqrt, log, ...)
• Simple function calls with positional arguments only

That is all - for now.

Only a hand‑written walk over Python's ast for a very small subset. No attribute access, no imports, no keywords, no assignments, no lambdas, no comprehensions. If it's not listed above it should raise an error.

evaluate(expression: str) -> float

Returns a float (even if you pass an int literal). Raises standard Python exceptions (NameError, TypeError, etc.) for invalid input.

evaluate("2*(3+4)-5/2")        # 11.5
evaluate("sin(pi/6)**2")       # 0.249999... (floating point)
evaluate("sqrt(2)**2")         # 2.000000...

Original itch: certain very large numeric expressions (or results) caused numexpr to raise errors in our workflow. For simple scalar math that shouldn't fail, we just needed a tiny, predictable evaluator that:

• Has zero heavy dependencies
• Doesn't optimize or chunk arrays (so no surprise shape / size limits)
• Always returns a plain Python float for valid math
• Is easy to audit (a short AST walk) and extend later if truly needed

So expreval exists to reliably handle those "big result" cases where bringing in numexpr (and hitting its internal limits) was overkill. If you only need quick scalar math, this keeps it boring and dependable.

MIT. See LICENSE.