Repository navigation
pybabel extract crashes when it encounters a byte string #1190
Description
(Happens on current master)
Sample file:
_(b'foo')now run pybabel extract test.py -o messages.pot and you'll get this error:
> pybabel extract test.py -o messages.pot
extracting messages from test.py
Traceback (most recent call last):
File "/home/dev/babel/venv/bin/pybabel", line 8, in <module>
sys.exit(main())
^^^^^^
File "/home/dev/babel/babel/messages/frontend.py", line 998, in main
return CommandLineInterface().run(sys.argv)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/dev/babel/babel/messages/frontend.py", line 924, in run
return cmdinst.run()
^^^^^^^^^^^^^
File "/home/dev/babel/babel/messages/frontend.py", line 522, in run
for filename, lineno, message, comments, context in extracted:
^^^^^^^^^
File "/home/dev/babel/babel/messages/extract.py", line 282, in check_and_call_extract_file
for message_tuple in extract_from_file(
^^^^^^^^^^^^^^^^^^
File "/home/dev/babel/babel/messages/extract.py", line 324, in extract_from_file
return list(extract(method, fileobj, keywords, comment_tags,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/dev/babel/babel/messages/extract.py", line 451, in extract
for lineno, funcname, messages, comments in results:
^^^^^^^
File "/home/dev/babel/babel/messages/extract.py", line 562, in extract_python
messages.append(''.join(buf))
^^^^^^^^^^^^
TypeError: sequence item 0: expected str instance, bytes foundIt seems like we're trying to call .join() on a byte string. I suggest just fixing the crash but keeping byte strings disallowed - they won't work with gettext anyway because gettext uses normal strings internally.
Activity
akx commented on Mar 3, 2025
Agreed. We should warn about that and leave it non-extracted.
tomasr8 commented on Mar 5, 2025
I'll put it on my todo list for this weekend
Mukller commented on Jul 29, 2026
Root cause and fix:
_parse_python_string() is annotated as -> str | None, but when it encounters a byte-string literal (e.g. b'foo'), it evaluates to an ast.Constant whose .value is a bytes object and returns bytes unchanged. The caller in extract_python appends this to buf (a list of string fragments), then calls ''.join(buf) — which fails because str.join() requires all items to be str:
Fix: Add an isinstance(body.value, str) guard in _parse_python_string so that byte-string literals are silently skipped (returning None), matching the behaviour for all other non-extractable expressions:
Byte strings cannot be used as gettext messages anyway, so ignoring them is the correct behaviour. The crash is avoided and callers calling _(b'foo') will simply have no entry extracted, with no error.
I have a PR ready.