ASCII//TRANSLIT not found with python 3.9

[bodograumann]

In python 3.9 the ASCII//TRANSLIT codec cannot be loaded anymore:

> python -m unittest
.....EE
======================================================================
ERROR: test_incremental_encode (test_iconvcodec.TestIconvcodecModule)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/bodo/Libraries/python-iconv/test_iconvcodec.py", line 21, in test_incremental_encode
    encoder = codecs.getincrementalencoder("ASCII//TRANSLIT")()
  File "/usr/lib/python3.9/codecs.py", line 986, in getincrementalencoder
    encoder = lookup(encoding).incrementalencoder
LookupError: unknown encoding: ASCII//TRANSLIT

======================================================================
ERROR: test_transliterate (test_iconvcodec.TestIconvcodecModule)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/bodo/Libraries/python-iconv/test_iconvcodec.py", line 17, in test_transliterate
    bytestring = string.encode("ASCII//TRANSLIT")
LookupError: unknown encoding: ASCII//TRANSLIT

----------------------------------------------------------------------
Ran 7 tests in 0.003s

FAILED (errors=2)

[bodograumann]

This is due to a change in python codecs handling from version 3.8 to 3.9, where codec names are normalized.
Cf. https://bugs.python.org/issue37751

[bodograumann]

Submitted a new bug report: https://bugs.python.org/issue44723

[bodograumann]

Another related report: https://bugs.python.org/issue46508

[Javrd]

I use this workaround to get it working in case anyone interested:

import codecs
import iconvcodec
import encodings

# Hack alias for resolving https://github.com/bodograumann/python-iconv/issues/4
# as charset names are normalized
def normalized_alias(charset):
    codec = 'ASCII//TRANSLIT' # It works also for other codecs that contains normalized characters like '-' or ':'
    result = None
    if charset == encodings.normalize_encoding(codec.lower()):
        result = iconvcodec.lookup(codec)
    return result

codecs.register(normalized_alias)

[bodograumann]

Current upstream issue: python/cpython#88886

The breaking change has been reverted, but looks like this will probably only land in python 3.15:

python/cpython#137167

[StanFromIreland]

Yes we didn't backport, I suggest the tests are modified here for the time being.