bytes.fromhex() should parse a bytes

[lordmauve]

Feature or enhancement

Proposal:

bytes.fromhex() should accept a bytes:

>>> bytes.fromhex(b'8a3218def90a84cb4373beed87d9ba1ccc7d90d1')
b'\x8a2\x18\xde\xf9\n\x84\xcbCs\xbe\xed\x87\xd9\xba\x1c\xcc}\x90\xd1'

Background:

bytes.fromhex() accepts a str:

>>> bytes.fromhex('8a3218def90a84cb4373beed87d9ba1ccc7d90d1')
b'\x8a2\x18\xde\xf9\n\x84\xcbCs\xbe\xed\x87\xd9\xba\x1c\xcc}\x90\xd1'

However, it refuses to parse a byte string:

>>> bytes.fromhex(b'8a3218def90a84cb4373beed87d9ba1ccc7d90d1')
Traceback (most recent call last):
  File "<python-input-0>", line 1, in <module>
    bytes.fromhex(b'8a3218def90a84cb4373beed87d9ba1ccc7d90d1')
    ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: fromhex() argument must be str, not bytes

This requires an extra .decode(), which is rather wasteful given that the str is not of any real use.

This came up for me in parsing the output of git cat-file --batch, which must be a binary stream because it contains bytes, but includes header lines like

8a3218def90a84cb4373beed87d9ba1ccc7d90d1 100644 1394

The integers are parseable directly from bytes:

>>> int(b'100644', 8)
33188

so it seems like an omission that the SHAs are not.

Has this already been discussed elsewhere?

This is a minor feature, which does not need previous discussion elsewhere

Links to previous discussion of this feature:

No response

Linked PRs

• gh-129349: Accept bytes in bytes.fromhex()/bytearray.fromhex() #129844

[serhiy-storchaka]

You have to look at the discussion that preceded introduction of bytes.fromhex() or ask the authors of the feature why it only supports str.

BTW, float.fromhex() also only supports str.

[serhiy-storchaka]

For reference, bytes.fromhex() was added by @birkenfeld in 0b9b9e0. bytes at that time was a mutable type, later it was renamed to bytearray, and str became new bytes. Initially, fromhex() accepted 8-bit and Unicode strings. It stopped support of 8-bit strings only after making new bytes and new str not interoperable in Python 3.0. So I do not see reasons against supporting bytes here.

[vstinner]

Implemented in e0637ce.

[serhiy-storchaka]

I had doubts about supporting the general buffer protocol (see #779 and #71759). The buffer protocol is for binary data, and here we have text data. But it is difficult to decide in which cases to support only bytes, in which bytes and bytearray, and in which any "bytes-like" object.

Note also that the complex constructor only supports str, not bytes (there may be open or closed issue for this).

[vstinner]

IMO it's convenient to accept bytes when the "text" is ASCII only.