Add unified C API for accessing bytes and bytearray

[salty-horse]

BPO	36065
Nosy	@ronaldoussoren, @salty-horse, @serhiy-storchaka, @matrixise

^{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 = None
closed_at = <Date 2019-02-21.16:01:51.213>
created_at = <Date 2019-02-21.14:20:08.593>
labels = ['interpreter-core', 'type-feature', '3.8']
title = 'Add unified C API for accessing bytes and bytearray'
updated_at = <Date 2019-02-21.16:01:51.212>
user = 'https://github.com/salty-horse'

bugs.python.org fields:

activity = <Date 2019-02-21.16:01:51.212>
actor = 'serhiy.storchaka'
assignee = 'none'
closed = True
closed_date = <Date 2019-02-21.16:01:51.213>
closer = 'serhiy.storchaka'
components = ['Interpreter Core']
creation = <Date 2019-02-21.14:20:08.593>
creator = 'salty-horse'
dependencies = []
files = []
hgrepos = []
issue_num = 36065
keywords = []
message_count = 7.0
messages = ['336218', '336219', '336220', '336222', '336227', '336232', '336233']
nosy_count = 4.0
nosy_names = ['ronaldoussoren', 'salty-horse', 'serhiy.storchaka', 'matrixise']
pr_nums = []
priority = 'normal'
resolution = 'rejected'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue36065'
versions = ['Python 3.8']

[salty-horse]

It would be useful to have a shared API for consuming bytes and bytearrays.

At present, I need to write very similar code twice. Some existing codebases only support bytes (perhaps forgetting bytearrays exist). Adding support for bytearray would be trivial if there was a shared API.

These are the functions/macros that can have "BytesOrByteArray" equivalents:

• PyBytes_Check
• PyBytes_CheckExact
• PyBytes_Size
• PyBytes_GET_SIZE
• PyBytes_AsString
• PyBytes_AS_STRING
• PyBytes_AsStringAndSize

Here are some example implementations for the macros:

#define PyBytesOrByteArray_Check(ob) (PyBytes_Check(ob) || PyByteArray_Check(ob))
#define PyBytesOrByteArray_AS_STRING(ob) (PyBytes_Check(ob) ? PyBytes_AS_STRING(ob) : PyByteArray_AS_STRING(ob))
#define PyBytesOrByteArray_GET_SIZE(ob) #define PyByteArray_GET_SIZE(self) (assert(PyBytesOrByteArray_Check(self)), Py_SIZE(self))

[matrixise]

I remove 2.7 because this branch is in bugfix mode, no new features.

[vstinner]

Stéphane: Please don't add me to the nosy list of bugs.

[ronaldoussoren]

What is your use case for this? Is that something that can use the buffer API instead of these low-level APIs?

[serhiy-storchaka]

The unified C API already exists. It is called the buffer protocol.

https://docs.python.org/3/c-api/buffer.html#buffer-related-functions

[salty-horse]

My use-case is modifying existing code that supports bytes to also support bytearray.

https://github.com/mongodb/mongo-python-driver/blob/9902d239b4e557c2a657e8c8110f7751864cec95/bson/_cbsonmodule.c#L1112

The buffer protocol, which I didn't know of, feels slightly complicated for my use-case. For now I opted to adding these macros myself, only changing 3 lines.

[serhiy-storchaka]

If you need to support only bytes and bytearray, but not other bytes-like object, this is a too special case. It is easy to write your own macros or functions that wrap existing C API. Other option -- duplicate the code and replace PyBytes_ with PyByteArray_. In all cases be aware abot differences between bytes and bytearray: bytarray can change its content and size, saved values of PyByteArray_AS_STRING(ob) and PyByteArray_GET_SIZE(self) can not be used after executing arbitrary code in destructors or releasing GIT.