Skip to content
Docs Pydantic Docs

base64 module

The base64, base32, base16, base85 and ascii85 codecs plus the MIME helpers encodebytes/decodebytes are implemented: b64encode, b64decode, standard_b64encode, standard_b64decode, urlsafe_b64encode, urlsafe_b64decode, b32encode, b32decode, b32hexencode, b32hexdecode, b16encode, b16decode, b85encode, b85decode, z85encode, z85decode, a85encode, a85decode, encodebytes, decodebytes. Output bytes, error types and error messages match CPython 3.14 subject to the divergences below.

Not implemented

  • base64.encode(input, output) and base64.decode(input, output), which read and write binary file objects. These are pure Python in CPython, so they call read/readline/write on whatever they are handed; a Monty builtin reaching back into Python runs the call to completion and cannot pause, and reading a real file is a suspension out to the host. They would therefore work on a duck-typed object and fail on the file open() returns — the same restriction filter and sorted(key=) carry. Accessing them raises AttributeError.
  • The python -m base64 command line interface.

a85encode / a85decode

  • wrapcol reaches CPython’s max() and then a slice expression. Every non-int fails at one or the other, so what differs is which message you get, not whether the call fails. Monty matches CPython for a float ('float' object cannot be interpreted as an integer) and for a type that cannot be ordered against an int. A class defining both __gt__ and __index__ diverges: CPython gets past max() and range() to fail on i + wrapcol (unsupported operand type(s) for +), while Monty rejects it at the comparison, since it does not dispatch ordering dunders at all (see classes.md).

Input types

Encoders take bytes; decoders take bytes or an ASCII-only str. Monty has no bytearray, memoryview or array, so the “bytes-like object” the CPython docs describe is always bytes here.

Module attributes

The attributes CPython leaks as a side effect of how base64 is written — bytes_types, and the binascii and re modules it imports — are absent and raise AttributeError. They are not documented API.

binascii

Every name CPython’s binascii exposes is implemented, so the divergence below is all that separates the two.

repr() of a binascii.Error or binascii.Incomplete uses the qualified name: binascii.Error('Non-hexadecimal digit found') where CPython gives Error('Non-hexadecimal digit found'). type(exc).__name__ and str(type(exc)) both match CPython — see exceptions.md.