Support the buffer protocol in json.loads()

[fafhrd91]

BPO	30193
Nosy	@rhettinger, @etrepum, @ezio-melotti, @serhiy-storchaka, @fafhrd91
PRs	bpo-30193: Allow to load buffer objects with json.loads() #1334 bpo-30193: Allow to load buffer objects with json.loads() #14977

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/etrepum'
closed_at = None
created_at = <Date 2017-04-27.22:55:44.853>
labels = ['3.7', 'type-feature', 'library']
title = 'Support the buffer protocol in json.loads()'
updated_at = <Date 2019-07-27.09:55:07.816>
user = 'https://github.com/fafhrd91'

bugs.python.org fields:

activity = <Date 2019-07-27.09:55:07.816>
actor = 'flavianhautbois'
assignee = 'bob.ippolito'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2017-04-27.22:55:44.853>
creator = 'fafhrd91'
dependencies = []
files = []
hgrepos = []
issue_num = 30193
keywords = ['patch']
message_count = 5.0
messages = ['292487', '292508', '292509', '293250', '293251']
nosy_count = 5.0
nosy_names = ['rhettinger', 'bob.ippolito', 'ezio.melotti', 'serhiy.storchaka', 'fafhrd91']
pr_nums = ['1334', '14977']
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue30193'
versions = ['Python 3.7']

Linked PRs

• gh-74379: allow json.loads with memoryview arguments #131114

[fafhrd91]

It is not possible to use buffer objects in json.loads()

[serhiy-storchaka]

This isn't so easy. Supporting the buffer protocol can slowdown the common case -- str.

Do you have concrete use case for this feature or add it just as an extension of bytes and bytearray support?

[serhiy-storchaka]

Withdraw my note about slowing down str.

[serhiy-storchaka]

The proposed patch doesn't work with all objects supporting the buffer protocol, for example it doesn't work with array.array('u', '\ufeff[1,2,3]'). This is easy to fix. But I'm not sure that it is worth to complicate the code for adding support of the buffer protocol.

[fafhrd91]

I am fine with any decision. close with "won't fix" is fine too

[kovidgoyal]

Please re-open this. There is currently no way to decode JSON in the stdlib from a memoryview. In my case the memoryview comes from C code, copying it into a temporary bytes object just because json.loads() does not support the buffer protocol is very wasteful and has a significant performance impact for large amounts of JSON data.

Tangentially, the type annotations for json.loads() dont raise an error if you pass in a memoryview into it.

[AlexWaygood]

Tangentially, the type annotations for json.loads() dont raise an error if you pass in a memoryview into it.

Please open an issue at https://github.com/python/typeshed if you want that fixed @kovidgoyal — typeshed deals with all the type annotations for the stdlib!