First ones I noticed were codecs.encode() and codecs.decode(). Here is a list of other candidates, although they are not all documented, so maybe should not all be in __all__.

>>> import codecs
>>> public = (a for a in dir(codecs) if not a.startswith("_"))
>>> sorted(set(public).difference(codecs.__all__))
['BufferedIncrementalDecoder', 'BufferedIncrementalEncoder', 'Codec', 'CodecInfo', 'IncrementalDecoder', 'IncrementalEncoder', 'StreamReader', 'StreamReaderWriter', 'StreamRecoder', 'StreamWriter', 'ascii_decode', 'ascii_encode', 'backslashreplace_errors', 'builtins', 'charmap_build', 'charmap_decode', 'charmap_encode', 'decode', 'encode', 'escape_decode', 'escape_encode', 'getdecoder', 'getencoder', 'getincrementaldecoder', 'getincrementalencoder', 'getreader', 'getwriter', 'iterdecode', 'iterencode', 'latin_1_decode', 'latin_1_encode', 'make_encoding_map', 'make_identity_dict', 'raw_unicode_escape_decode', 'raw_unicode_escape_encode', 'readbuffer_encode', 'sys', 'unicode_escape_decode', 'unicode_escape_encode', 'unicode_internal_decode', 'unicode_internal_encode', 'utf_16_be_decode', 'utf_16_be_encode', 'utf_16_decode', 'utf_16_encode', 'utf_16_ex_decode', 'utf_16_le_decode', 'utf_16_le_encode', 'utf_32_be_decode', 'utf_32_be_encode', 'utf_32_decode', 'utf_32_encode', 'utf_32_ex_decode', 'utf_32_le_decode', 'utf_32_le_encode', 'utf_7_decode', 'utf_7_encode', 'utf_8_decode', 'utf_8_encode']

Yes, only documented names should be in __all__. Do you want to provide a patch Martin? Suggestion: first filter names which are contained in codecs.rst.

Okay I will try and make a patch for this once I have finished a patch to revise the documentation for bpo-19548

Here is the patch (against the default branch)

Thanks! Could you also add a test? See PublicAPITests in Lib/test/test_shutil.py as an example.

This patch includes a new test; although this kind of testing won’t detect when someone adds a new API and forgets to add it to __all__.

LGTM.

+1

New changeset c6491d91d59a by Serhiy Storchaka in branch '2.7':
Issue bpo-23071: Added missing names to codecs.__all__. Patch by Martin Panter.
https://hg.python.org/cpython/rev/c6491d91d59a

New changeset 2b642f2ca391 by Serhiy Storchaka in branch '3.4':
Issue bpo-23071: Added missing names to codecs.__all__. Patch by Martin Panter.
https://hg.python.org/cpython/rev/2b642f2ca391

New changeset 73086376ed3c by Serhiy Storchaka in branch 'default':
Issue bpo-23071: Added missing names to codecs.__all__. Patch by Martin Panter.
https://hg.python.org/cpython/rev/73086376ed3c

Thank you for your patch Martin.

In 3.4, "namereplace_errors" needs to be removed in the test:

https://hg.python.org/cpython/rev/2b642f2ca391#l2.17

======================================================================
FAIL: test_all (test.test_codecs.CodecsModuleTest)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/usr/home/buildbot/python/3.4.koobs-freebsd9/build/Lib/test/test_codecs.py", line 1668, in test_all
    self.assertCountEqual(api, codecs.__all__)
AssertionError: Element counts were not equal:
First has 1, Second has 0:  'namereplace_errors'

http://buildbot.python.org/all/builders/AMD64%20FreeBSD%209.x%203.4/builds/645/steps/test/logs/stdio

New changeset 1ab04bcd03bf by Serhiy Storchaka in branch '3.4':
Issue bpo-23071: "namereplace_errors" was added only in 3.5.
https://hg.python.org/cpython/rev/1ab04bcd03bf

Hmm, I have no ideas how I missed this when ran tests. Thank you Berker.