/pane

A modern dataclass library for Python, focused on speed and expressiveness

Primary LanguagePythonMIT LicenseMIT

pane

pane is a modern Python library for dataclasses and data conversion, aiming for speed and expressiveness.

pane draws heavy inspiration from Pydantic (among others), but its goals are quite different. For example, pane has first-class conversion to and from all JSON datatypes, not just mappings (and no 'root' hacks necessary). In this sense pane is a library for general data conversion & validation, while Pydantic is a dataclass-first library. In addition, pane is stricter than Pydantic in several cases. For instance, pane will not attempt to coerce 3.14 or "3" to an integer.

pane is designed to be used to create complex declarative configuration languages in formats like JSON, TOML, and YAML. This requires full support for complex, nested types like t.Union[int, t.Tuple[int, str], list[int]]. It also requires useful error messages:

>>> pane.convert('fail', t.Union[int, t.Tuple[int, str], list[int]])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "pane/convert.py", line 489, in convert
    return from_data(data, ty)
           ^^^^^^^^^^^^^^^^^^^
  File "pane/convert.py", line 484, in from_data
    return converter.convert(val)
           ^^^^^^^^^^^^^^^^^^^^^^
  File "pane/convert.py", line 128, in convert
    raise ConvertError(node)
pane.convert.ConvertError: Expected one of:
- an int
- tuple of length 2
- sequence
Instead got `fail` of type `str`

Features

pane is a work in progress. The following is a roadmap of features:

Feature State
Basic type conversions Done
Numeric array types (numpy) Done
Sum & product type support Done
Tagged unions Partial
'Flattened' fields Planned
Basic dataclasses Done
Dataclass helpers Partial
Generic & inherited dataclasses Done
Parameter aliases & renaming Done
Arbitrary validation logic Done
Schema import/export Not planned