locale.getlocale() fails if locale name doesn't include encoding

[serhiy-storchaka]

BPO	20088
Nosy	@malemburg, @loewis, @tjguk, @zware, @serhiy-storchaka, @eryksun, @zooba, @Kristinita

^{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 = None
created_at = <Date 2013-12-28.09:44:02.941>
labels = ['3.8', 'type-bug', 'library', '3.9', '3.10']
title = "locale.getlocale() fails if locale name doesn't include encoding"
updated_at = <Date 2021-03-04.15:21:05.592>
user = 'https://github.com/serhiy-storchaka'

bugs.python.org fields:

activity = <Date 2021-03-04.15:21:05.592>
actor = 'eryksun'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2013-12-28.09:44:02.941>
creator = 'serhiy.storchaka'
dependencies = []
files = []
hgrepos = []
issue_num = 20088
keywords = []
message_count = 4.0
messages = ['207026', '235823', '235840', '388095']
nosy_count = 9.0
nosy_names = ['lemburg', 'loewis', 'tim.golden', 'Arfrever', 'zach.ware', 'serhiy.storchaka', 'eryksun', 'steve.dower', 'nervov_fan']
pr_nums = []
priority = 'normal'
resolution = None
stage = None
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue20088'
versions = ['Python 3.8', 'Python 3.9', 'Python 3.10']

[serhiy-storchaka]

>>> import locale, _locale
>>> _locale.setlocale(locale.LC_CTYPE, 'en_AG')
'en_AG'
>>> _locale.setlocale(locale.LC_CTYPE)
'en_AG'
>>> locale.getlocale(locale.LC_CTYPE)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/serhiy/py/cpython/Lib/locale.py", line 575, in getlocale
    return _parse_localename(localename)
  File "/home/serhiy/py/cpython/Lib/locale.py", line 484, in _parse_localename
    raise ValueError('unknown locale: %s' % localename)
ValueError: unknown locale: en_AG

One solution is proposed in bpo-20079: map all supported in glibc locale names without encoding to locale names with encoding. But see bpo-20087. And default encoding can be different on other systems (not based on glibc).

Other solution is not guess an encoding, but use locale.nl_langinfo(locale.CODESET) in locale.getlocale(). And left in locale alias table only nonstandard mappings (such as english_uk -> en_GB.ISO8859-1 and sr_yu.iso88595 -> sr_CS.ISO8859-5).

[BreamoreBoy]

With bpo-20079 closed but bpo-20087 still open where do we stand with this issue?

[eryksun]

For 3.5 this affects Windows as well, since the new CRT supports RFC1766 language codes, but only without a codepage spec:

    Python 3.5.0a1 (v3.5.0a1:5d4b6a57d5fd, Feb  7 2015, 18:15:14) 
    [MSC v.1900 64 bit (AMD64)] on win32
    Type "help", "copyright", "credits" or "license" for more information.
    >>> import locale
    >>> locale.setlocale(locale.LC_ALL, 'en-GB')
    'en-GB'

    >>> locale.getlocale()
    Traceback (most recent call last):
      File "<stdin>", line 1, in <module>
      File "C:\Program Files\Python35\lib\locale.py", line 578, in getlocale
        return _parse_localename(localename)
      File "C:\Program Files\Python35\lib\locale.py", line 487, in _parse_localename
        raise ValueError('unknown locale: %s' % localename)
    ValueError: unknown locale: en-GB

On Vista+ (since 3.5 drops XP support) the codepage can be queried easily via GetLocaleInfoEx:

    >>> from ctypes import *
    >>> LOCALE_IDEFAULTANSICODEPAGE = 0x1004
    >>> GetLocaleInfoEx = WinDLL('kernel32').GetLocaleInfoEx
    >>> info = (c_wchar * 100)()
    >>> GetLocaleInfoEx("en-GB", LOCALE_IDEFAULTANSICODEPAGE, info, len(info)) 
    5
    >>> info.value
    '1252'
    >>> GetLocaleInfoEx("zh-CN", LOCALE_IDEFAULTANSICODEPAGE, info, len(info))
    4
    >>> info.value                                                            
    '936'

Note that Windows follows the RFC spec here (not POSIX), using a hyphen instead of an underscore.

This is a bit of tangent, but for the Windows full language_country.codepage form, the X-11 based locale_alias dict is generally useless. So, contrary to the docs, on Windows getlocale doesn't return the language code in RFC 1766 form. In some cases it does, but only by chance.

[eryksun]

The locale_alias database was extended to support "en_AG" and many others, but I'd still prefer Serhiy's suggestion to not guess the codeset when checking the default LC_CTYPE category. Use locale.nl_langinfo(locale.CODESET), if it's available.

In Windows, I'd prefer to never guess since the encoding for a BCP-47 locale name can be directly queried. But this issue can be restricted to POSIX. What to do in Windows is already being considered in more recent issues: bpo-23425, bpo-37945, and bpo-43115.

[StanFromIreland]

This is no longer an issue on Linux, I cannot reproduce in any way, a default is provided (UTF-8), and it seems the windows discussion was moved elsewhere, is there anything to do here?