API Reference¶

attributedict exposes a single public type:

from attributedict import AttributeDict

__all__ == ["AttributeDict"]. The C extension module attributedict._attributedict is an implementation detail and isn't part of the public API.

Reference oracle: attributedict._reference contains a pure-Python implementation of the same contract. The test suite and benchmarks use it as the cross-check baseline; it is not the public import path.

Construction¶

AttributeDict # empty
AttributeDict(mapping) # from a mapping
AttributeDict(iterable_of_pairs)
AttributeDict(**kwargs)
AttributeDict(mapping, **kwargs)

All forms match dict. Nested mappings are recursively converted.

Attribute / Mapping semantics¶

Operation	Behavior
`d[key]`	dict getitem
`d[key] = v`	dict setitem
`del d[key]`	dict delitem
`d.name`	real type attribute first; else `d["name"]` if key; else AttributeError
`d.name = v`	`d["name"] = v`
`del d.name`	delete key `"name"`; `AttributeError` if absent
`d.items` (key `"items"` exists)	the dict method (type attribute wins)
`dict.items(d)`	the mapping view

repr / str¶

repr(AttributeDict({'a': 1})) # "AttributeDict({'a': 1})"

Nested values use their own repr; recursive structures render with ....

Equality / hash¶

==/!= follow dict semantics. AttributeDict is unhashable like dict.

Copy / pickle¶

copy.copy, copy.deepcopy, pickle.dumps/loads are supported. Nested AttributeDict instances stay AttributeDict; self-references and cycles work.

Errors¶

Missing key → KeyError
Missing attribute → AttributeError
Unhashable key → TypeError
del d.missing → AttributeError (documented deviation from del d["missing"] → KeyError)

Typing¶

py.typed is shipped. Attribute access can't be fully represented statically, so typing is best-effort — and honestly documented as such.