asteval 1.0.8

Safe, minimalistic evaluator of python expression using ast module

Links

• Documentation: https://lmfit.github.io/asteval/

• PyPI installation: https://pypi.org/project/asteval/

• Development Code: https://github.com/lmfit/asteval

• Issue Tracker: https://github.com/lmfit/asteval/issues

Installation

Use pip install asteval to install the asteval library.

Asteval supports Python 3.10 or higher. No modules outside of the standard library are required, though if NumPy is installed, many functions from it will be used by default.

About Asteval

Asteval is a safe(ish) evaluator of Python expressions and statements, using Python’s ast module. It provides an easy-to-use restricted Python interpreter that supports a pretty complete subset of the Python language and can handle user input more safely than Python’s eval(). Asteval emphasizes mathematical expressions so that many functions from NumPy are imported and used if available, but also provides a pretty complete subset of the Python language. It can be used as an embedded macro language within a large application or as a simple calculator for mathematical calculations.

Asteval supports many Python language constructs by default, including conditionals (if-elif-else blocks and if expressions), flow control (for loops, while loops, with blocks, and try-except-finally blocks), list comprehension, slicing, subscripting, and f-strings. All data are Python objects and the standard built-in data structures (dictionaries, tuples, lists, sets, strings, functions, and Numpy nd-arrays) are well supported, and most public attributes and methods of these objects are available. Asteval does place some limitations on “looking under the hood” to get private and potentially unsafe methods.

Many of the standard built-in Python functions are available, as are the functions from the math module. Some built-in operators and functions, such as getattr, and setattr are not allowed, and some including open and ** are replaced with versions intended to make them safer for user input. If the NumPy is installed, many of its functions will also be available. Programmers can add custom functions and data of their own into each Asteval session. Users can define and run their own functions within the confines of the limitations of the Asteval language.

Asteval converts user input into Python’s own abstract syntax tree (AST) representation and determines the result by walking through that tree. This approach guarantees the parsing of input will be identical to that of Python, eliminating many lexing and parsing challenges and generating a result that is straightforward to interpret. This makes “correctness” easy to test and verify with high confidence, so that the emphasis can be placed on balancing functionality with safety.

There are several absences and differences with Python, and Asteval is by no means an attempt to reproduce Python with its own ast module. While, it does support a large subset of Python, the following features found in Python are not supported in Asteval:

1. creating classes is not supported

2. many internal methods and classes of Python objects, especially __dunder__ methods cannot be accessed.

3. eval, exec, yield, async, match/case, function decorators, generators, and type annotations are not supported.

4. f-strings are supported, but t-strings are not supported.

5. importing modules is not supported by default, though it can be enabled.

Most of these omissions and limitations are intentional, and aimed to strengthen Asteval against dangerous user code. A few of these (say, match/case and t-strings) omissions may simply be viewed as not particularly compelling need for an embedded interpreter exposed to user input.

Even with these restrictions, Asteval provides a pretty complete and usable scripting or “macro” language that can easily be embedded into a larger GUI or Web application that can handle user input without many of the risks associated with using Python’s eval().

Matt Newville <[e-mail]> Last Update: 17-Dec-2025