Skip to content
Docs Pydantic Docs

hashlib module

Every algorithm in hashlib.algorithms_guaranteed is implemented, with no OpenSSL behind it, and a hash object is ordinary session state: a dump taken between update() calls restores it mid-stream.

Not implemented

  • file_digest(). It reads from a file object in a loop, which a Monty builtin cannot do (see open.md).
  • scrypt().
  • hmac is not importable (see modules.md); pbkdf2_hmac() is the only HMAC construction.

algorithms_available and new()

algorithms_available is the same set as algorithms_guaranteed. In CPython it also lists what OpenSSL provides (sha512_256, sha512_224, ripemd160, md5-sha1, sm3, …), and new() accepts those names; Monty raises ValueError: unsupported hash type sha512_256.

new('blake2b512') returns a _blake2.blake2b object and new('blake2s256') a _blake2.blake2s; in CPython each is an OpenSSL-backed _hashlib.HASH whose name is also 'blake2b' / 'blake2s'.

pbkdf2_hmac() with an unsupported name raises ValueError: unsupported hash type <name> where CPython raises its ValueError subclass _hashlib.UnsupportedDigestmodError with the OpenSSL message [digital envelope routines] unsupported. With a SHAKE it raises ValueError: key length must be greater than 0. when dklen is omitted and ValueError: [Provider routines] xof digests not allowed when it is positive; CPython’s messages for these come from OpenSSL and vary between builds.

SHAKE digest lengths

digest(length) and hexdigest(length) on shake_128 / shake_256 objects reject a bad length with the messages of CPython’s _sha3 module, where CPython’s OpenSSL-backed objects behave differently:

  • a negative length raises ValueError: Cannot convert negative int, where CPython raises SystemError (digest) or MemoryError (hexdigest);
  • a length of 2**29 or more raises ValueError: length is too large, where CPython attempts the allocation and raises MemoryError;
  • a length beyond 2**64 - 1 raises OverflowError: Python int too large for C unsigned long, where CPython says Python int too large to convert to C ssize_t.

Smaller lengths beyond the session’s memory limit raise MemoryError (see resource_limits.md).

Windows integer widths

Monty’s messages for out-of-range integers are the same on every platform; CPython’s depend on the width of a C long, which is 32 bits on Windows:

  • BLAKE2 leaf_size above 2**32 - 1 raises OverflowError: leaf_size is too large; on Windows CPython the conversion overflows first with Python int too large for C unsigned long.
  • pbkdf2_hmac() iterations or dklen above 2**31 - 1 raises OverflowError: iteration value is too great. or key length is too great.; on Windows CPython the conversion overflows first with Python int too large to convert to C long.

Input types

bytes is the only bytes-like input: Monty has no bytearray, memoryview or array. The int parameters take int and bool; a class defining __index__ raises TypeError: 'X' object cannot be interpreted as an integer (see classes.md).

Method objects

h.update cannot be bound and called later: methods on hash objects can only be called in place (h.update(b'...')), as for the other builtin types (see classes.md).