From 01f7fafe8f2f33576037694c1001e5a438652e64 Mon Sep 17 00:00:00 2001
From: Jean Abou Samra <jean@abou-samra.fr>
Date: Wed, 14 Jun 2023 21:16:25 +0200
Subject: [PATCH 1/9] Support 't' specifier in keywords

Fixes #1014
---
 babel/messages/extract.py       | 92 ++++++++++++++++++---------------
 babel/messages/frontend.py      | 36 +++++++++----
 tests/messages/test_frontend.py | 11 ++++
 3 files changed, 88 insertions(+), 51 deletions(-)

diff --git a/babel/messages/extract.py b/babel/messages/extract.py
index b6dce6fdb..1a4def03d 100644
--- a/babel/messages/extract.py
+++ b/babel/messages/extract.py
@@ -55,7 +55,8 @@ class _FileObj(SupportsRead[bytes], SupportsReadline[bytes], Protocol):
         def seek(self, __offset: int, __whence: int = ...) -> int: ...
         def tell(self) -> int: ...
 
-    _Keyword: TypeAlias = tuple[int | tuple[int, int] | tuple[int, str], ...] | None
+    _SimpleKeyword: TypeAlias = tuple[int | tuple[int, int] | tuple[int, str], ...] | None
+    _Keyword: TypeAlias = dict[int | None, _SimpleKeyword] | _SimpleKeyword
 
     # 5-tuple of (filename, lineno, messages, comments, context)
     _FileExtractionResult: TypeAlias = tuple[str, int, str | tuple[str, ...], list[str], str | None]
@@ -400,56 +401,65 @@ def extract(
                    options=options or {})
 
     for lineno, funcname, messages, comments in results:
-        spec = keywords[funcname] or (1,) if funcname else (1,)
+        specs = keywords[funcname] or (1,) if funcname else (1,)
+        if not isinstance(specs, dict): # for backwards compatibility
+            specs = {None: specs}
         if not isinstance(messages, (list, tuple)):
             messages = [messages]
         if not messages:
             continue
 
         # Validate the messages against the keyword's specification
-        context = None
-        msgs = []
-        invalid = False
-        # last_index is 1 based like the keyword spec
-        last_index = len(messages)
-        for index in spec:
-            if isinstance(index, tuple):
-                context = messages[index[0] - 1]
+
+        # None matches all arities.
+        for arity in (None, len(messages)):
+            try:
+                spec = specs[arity]
+            except KeyError:
+                continue
+            context = None
+            msgs = []
+            invalid = False
+            # last_index is 1 based like the keyword spec
+            last_index = len(messages)
+            for index in spec:
+                if isinstance(index, tuple):
+                    context = messages[index[0] - 1]
+                    continue
+                if last_index < index:
+                    # Not enough arguments
+                    invalid = True
+                    break
+                message = messages[index - 1]
+                if message is None:
+                    invalid = True
+                    break
+                msgs.append(message)
+            if invalid:
                 continue
-            if last_index < index:
-                # Not enough arguments
-                invalid = True
-                break
-            message = messages[index - 1]
-            if message is None:
-                invalid = True
-                break
-            msgs.append(message)
-        if invalid:
-            continue
 
-        # keyword spec indexes are 1 based, therefore '-1'
-        if isinstance(spec[0], tuple):
-            # context-aware *gettext method
-            first_msg_index = spec[1] - 1
-        else:
-            first_msg_index = spec[0] - 1
-        if not messages[first_msg_index]:
-            # An empty string msgid isn't valid, emit a warning
-            filename = (getattr(fileobj, "name", None) or "(unknown)")
-            sys.stderr.write(
-                f"{filename}:{lineno}: warning: Empty msgid.  It is reserved by GNU gettext: gettext(\"\") "
-                f"returns the header entry with meta information, not the empty string.\n"
-            )
-            continue
+            # keyword spec indexes are 1 based, therefore '-1'
+            if isinstance(spec[0], tuple):
+                # context-aware *gettext method
+                first_msg_index = spec[1] - 1
+            else:
+                first_msg_index = spec[0] - 1
+            if not messages[first_msg_index]:
+                # An empty string msgid isn't valid, emit a warning
+                filename = (getattr(fileobj, "name", None) or "(unknown)")
+                sys.stderr.write(
+                    f"{filename}:{lineno}: warning: Empty msgid.  It is reserved by GNU gettext: gettext(\"\") "
+                    f"returns the header entry with meta information, not the empty string.\n"
+                )
+                continue
 
-        messages = tuple(msgs)
-        if len(messages) == 1:
-            messages = messages[0]
+            messages = tuple(msgs)
+            if len(messages) == 1:
+                messages = messages[0]
 
-        if strip_comment_tags:
-            _strip_comment_tags(comments, comment_tags)
-        yield lineno, messages, comments, context
+            if strip_comment_tags:
+                _strip_comment_tags(comments, comment_tags)
+            yield lineno, messages, comments, context
 
 
 def extract_nothing(
diff --git a/babel/messages/frontend.py b/babel/messages/frontend.py
index af7e1d11d..bc2bd7f32 100644
--- a/babel/messages/frontend.py
+++ b/babel/messages/frontend.py
@@ -1111,13 +1111,14 @@ def parse_mapping(fileobj, filename=None):
 def parse_keywords(strings: Iterable[str] = ()):
     """Parse keywords specifications from the given list of strings.
 
-    >>> kw = sorted(parse_keywords(['_', 'dgettext:2', 'dngettext:2,3', 'pgettext:1c,2']).items())
+    >>> kw = sorted(parse_keywords(['_', 'dgettext:2', 'dngettext:2,3', 'pgettext:1c,2', 'polymorphic:1', 'polymorphic:2,2t', 'polymorphic:3c,3t']).items())
     >>> for keyword, indices in kw:
     ...     print((keyword, indices))
     ('_', None)
     ('dgettext', (2,))
     ('dngettext', (2, 3))
     ('pgettext', ((1, 'c'), 2))
+    ('polymorphic', {None: (1,), 2: (2,), 3: ((3, 'c'),)})
     """
     keywords = {}
     for string in strings:
@@ -1125,16 +1126,31 @@ def parse_keywords(strings: Iterable[str] = ()):
             funcname, indices = string.split(':')
         else:
             funcname, indices = string, None
+        number = None
+        if indices:
+            inds = []
+            for x in indices.split(','):
+                if x[-1] == 't':
+                    number = int(x[:-1])
+                elif x[-1] == 'c':
+                    inds.append((int(x[:-1]), 'c'))
+                else:
+                    inds.append(int(x))
+            inds = tuple(inds)
+        else:
+            inds = None
         if funcname not in keywords:
-            if indices:
-                inds = []
-                for x in indices.split(','):
-                    if x[-1] == 'c':
-                        inds.append((int(x[:-1]), 'c'))
-                    else:
-                        inds.append(int(x))
-                indices = tuple(inds)
-            keywords[funcname] = indices
+            if number is None:
+                # For best backwards compatibility, collapse {None: x} into x.
+                keywords[funcname] = inds
+            else:
+                keywords[funcname] = {number: inds}
+        else:
+            if isinstance(keywords[funcname], tuple) or keywords[funcname] is None:
+                keywords[funcname] = {None: keywords[funcname]}
+            else:
+                assert isinstance(keywords[funcname], dict)
+            keywords[funcname][number] = inds
     return keywords
 
 
diff --git a/tests/messages/test_frontend.py b/tests/messages/test_frontend.py
index 8e1e670ca..20068afdb 100644
--- a/tests/messages/test_frontend.py
+++ b/tests/messages/test_frontend.py
@@ -1320,6 +1320,17 @@ def test_parse_keywords():
     }
 
 
+def test_parse_keywords_with_t():
+    kw = frontend.parse_keywords(['_:1', '_:2,2t', '_:2c,3,3t'])
+
+    assert kw == {
+        '_': {
+            None: (1,),
+            2: (2,),
+            3: ((2, 'c'), 3),
+        }
+    }
+
 def configure_cli_command(cmdline):
     """
     Helper to configure a command class, but not run it just yet.

From ab2f84195fb39246206e511d2ab1cad8c4d52c8a Mon Sep 17 00:00:00 2001
From: Jean Abou Samra <jean@abou-samra.fr>
Date: Sat, 1 Jul 2023 18:59:56 +0200
Subject: [PATCH 2/9] Fix variable reuse across loop bug and add test

---
 babel/messages/extract.py       |  8 ++++----
 tests/messages/test_frontend.py | 22 ++++++++++++++++++++--
 2 files changed, 24 insertions(+), 6 deletions(-)

diff --git a/babel/messages/extract.py b/babel/messages/extract.py
index 1a4def03d..cca5c9b53 100644
--- a/babel/messages/extract.py
+++ b/babel/messages/extract.py
@@ -453,13 +453,13 @@ def extract(
                 )
                 continue
 
-            messages = tuple(msgs)
-            if len(messages) == 1:
-                messages = messages[0]
+            msgs = tuple(msgs)
+            if len(msgs) == 1:
+                msgs = msgs[0]
 
             if strip_comment_tags:
                 _strip_comment_tags(comments, comment_tags)
-            yield lineno, messages, comments, context
+            yield lineno, msgs, comments, context
 
 
 def extract_nothing(
diff --git a/tests/messages/test_frontend.py b/tests/messages/test_frontend.py
index 20068afdb..ed8bf6d5c 100644
--- a/tests/messages/test_frontend.py
+++ b/tests/messages/test_frontend.py
@@ -17,7 +17,7 @@
 import time
 import unittest
 from datetime import datetime
-from io import StringIO
+from io import BytesIO, StringIO
 
 import pytest
 from freezegun import freeze_time
@@ -25,7 +25,7 @@
 
 from babel import __version__ as VERSION
 from babel.dates import format_datetime
-from babel.messages import Catalog, frontend
+from babel.messages import Catalog, extract, frontend
 from babel.messages.frontend import (
     BaseError,
     CommandLineInterface,
@@ -1331,6 +1331,24 @@ def test_parse_keywords_with_t():
         }
     }
 
+def test_extract_messages_with_t():
+    content = rb"""
+_("1 arg, arg 1")
+_("2 args, arg 1", "2 args, arg 2")
+_("3 args, arg 1", "3 args, arg 2", "3 args, arg 3")
+_("4 args, arg 1", "4 args, arg 2", "4 args, arg 3", "4 args, arg 4")
+"""
+    kw = frontend.parse_keywords(['_:1', '_:2,2t', '_:2c,3,3t'])
+    result = list(extract.extract("python", BytesIO(content), kw))
+    expected = [(2, '1 arg, arg 1', [], None),
+                (3, '2 args, arg 1', [], None),
+                (3, '2 args, arg 2', [], None),
+                (4, '3 args, arg 1', [], None),
+                (4, '3 args, arg 3', [], '3 args, arg 2'),
+                (5, '4 args, arg 1', [], None)]
+    assert result == expected
+
+
 def configure_cli_command(cmdline):
     """
     Helper to configure a command class, but not run it just yet.

From 907ff98adc5711f01a23dd392000db9856a16d2e Mon Sep 17 00:00:00 2001
From: Jean Abou Samra <jean@abou-samra.fr>
Date: Tue, 1 Aug 2023 11:33:06 +0200
Subject: [PATCH 3/9] Document --keyword format

---
 docs/cmdline.rst | 39 +++++++++++++++++++++++++++++++++++++++
 1 file changed, 39 insertions(+)

diff --git a/docs/cmdline.rst b/docs/cmdline.rst
index 8d9742fbf..d314f2e93 100644
--- a/docs/cmdline.rst
+++ b/docs/cmdline.rst
@@ -133,6 +133,45 @@ a collection of source files::
                             header comment for the catalog
 
 
+The meaning of ``--keyword`` values is as follows:
+
+- Pass a simple identifier like ``_`` to extract the first (and only the first)
+  argument of all function calls to ``_``,
+
+- To extract other arguments than the first, add a colon and the argument
+  indices separated by commas. For example, the ``dngettext`` function
+  typically expects translatable strings as second and third arguments,
+  so you could pass ``dngettext:2,3``.
+
+- Some arguments should not be interpreted as translatable strings, but
+  context strings. For that, append "c" to the argument index. For example:
+  ``pgettext:1c,2``.
+
+- In C++ and Python, you may have functions that behave differently
+  depending on how many arguments they take. For this use case, you can
+  add an integer followed by "t" after the colon. In this case, the
+  keyword will only match a function invocation if it has the specified
+  total number of arguments.  For example, if you have a function that
+  behaves as ``gettext`` (argument is a message) or ``pgettext``
+  (arguments are a context and a message) depending on whether it takes
+  one or two arguments, you can pass
+  ``--keyword=gettext:1t --keyword=pgettext:1c,2,2t``.
+
+The default keywords are equivalent to passing ::
+
+  --keyword=_
+  --keyword=gettext
+  --keyword=ngettext:1,2
+  --keyword=ugettext
+  --keyword=ungettext:1,2
+  --keyword=dgettext:2
+  --keyword=dngettext:2,3
+  --keyword=N_
+  --keyword=pgettext:1c,2
+  --keyword=npgettext:1c,2,3
+
+
+
 init
 ====
 

From b879f5001ee77396f57791858854d2c32ad9a081 Mon Sep 17 00:00:00 2001
From: Jean Abou Samra <jean@abou-samra.fr>
Date: Tue, 1 Aug 2023 11:33:20 +0200
Subject: [PATCH 4/9] Document parse_keywords output format and make logic more
 readable

---
 babel/messages/frontend.py | 83 ++++++++++++++++++++++----------------
 1 file changed, 48 insertions(+), 35 deletions(-)

diff --git a/babel/messages/frontend.py b/babel/messages/frontend.py
index bc2bd7f32..1034a9195 100644
--- a/babel/messages/frontend.py
+++ b/babel/messages/frontend.py
@@ -1107,50 +1107,63 @@ def parse_mapping(fileobj, filename=None):
 
     return method_map, options_map
 
+def _parse_spec(s):
+    inds = []
+    number = None
+    for x in s.split(','):
+        if x[-1] == 't':
+            number = int(x[:-1])
+        elif x[-1] == 'c':
+            inds.append((int(x[:-1]), 'c'))
+        else:
+            inds.append(int(x))
+    return number, tuple(inds)
 
 def parse_keywords(strings: Iterable[str] = ()):
     """Parse keywords specifications from the given list of strings.
 
-    >>> kw = sorted(parse_keywords(['_', 'dgettext:2', 'dngettext:2,3', 'pgettext:1c,2', 'polymorphic:1', 'polymorphic:2,2t', 'polymorphic:3c,3t']).items())
-    >>> for keyword, indices in kw:
-    ...     print((keyword, indices))
-    ('_', None)
-    ('dgettext', (2,))
-    ('dngettext', (2, 3))
-    ('pgettext', ((1, 'c'), 2))
-    ('polymorphic', {None: (1,), 2: (2,), 3: ((3, 'c'),)})
+    >>> import pprint
+    >>> keywords = ['_', 'dgettext:2', 'dngettext:2,3', 'pgettext:1c,2',
+    ...             'polymorphic:1', 'polymorphic:2,2t', 'polymorphic:3c,3t']
+    >>> pprint.pprint(parse_keywords(keywords))
+    {'_': None,
+     'dgettext': (2,),
+     'dngettext': (2, 3),
+     'pgettext': ((1, 'c'), 2),
+     'polymorphic': {None: (1,), 2: (2,), 3: ((3, 'c'),)}}
+
+    The input keywords are in GNU Gettext style; see :doc:`cmdline` for details.
+
+    The output is a dictionary mapping keyword names to a dictionary of specifications.
+    Keys in this dictionary are numbers of arguments, where ``None`` means that all numbers
+    of arguments are matched, and a number means only calls with that number of arguments
+    are matched (which happens when using the "t" specifier). However, as a special
+    case for backwards compatibility, if the dictionary of specifications would
+    be ``{None: x}``, i.e., there is only one specification and it matches all argument
+    counts, then it is collapsed into just ``x``.
+
+    A specification is either a tuple or None. If a tuple, each element can be either a number
+    ``n``, meaning that the nth argument should be extracted as a message, or the tuple
+    ``(n, 'c')``, meaning that the nth argument should be extracted as context for the
+    messages. A ``None`` specification is equivalent to ``(1,)``, extracting the first
+    argument.
     """
     keywords = {}
     for string in strings:
         if ':' in string:
-            funcname, indices = string.split(':')
-        else:
-            funcname, indices = string, None
-        number = None
-        if indices:
-            inds = []
-            for x in indices.split(','):
-                if x[-1] == 't':
-                    number = int(x[:-1])
-                elif x[-1] == 'c':
-                    inds.append((int(x[:-1]), 'c'))
-                else:
-                    inds.append(int(x))
-            inds = tuple(inds)
-        else:
-            inds = None
-        if funcname not in keywords:
-            if number is None:
-                # For best backwards compatibility, collapse {None: x} into x.
-                keywords[funcname] = inds
-            else:
-                keywords[funcname] = {number: inds}
+            funcname, spec_str = string.split(':')
+            number, spec = _parse_spec(spec_str)
         else:
-            if isinstance(keywords[funcname], tuple) or keywords[funcname] is None:
-                keywords[funcname] = {None: keywords[funcname]}
-            else:
-                assert isinstance(keywords[funcname], dict)
-            keywords[funcname][number] = inds
+            funcname = string
+            number = None
+            spec = None
+        keywords.setdefault(funcname, {})[number] = spec
+
+    # For best backwards compatibility, collapse {None: x} into x.
+    for k, v in keywords.items():
+        if set(v.keys()) == {None}:
+            keywords[k] = v[None]
+
     return keywords
 
 

From 3544bc526b3add250313e666b2c29e90ab6eca55 Mon Sep 17 00:00:00 2001
From: Jean Abou Samra <jean@abou-samra.fr>
Date: Tue, 1 Aug 2023 12:26:21 +0200
Subject: [PATCH 5/9] Use a helper function to make extraction logic more
 readable

---
 babel/messages/extract.py | 98 ++++++++++++++++++++-------------------
 1 file changed, 51 insertions(+), 47 deletions(-)

diff --git a/babel/messages/extract.py b/babel/messages/extract.py
index cca5c9b53..e53c3a626 100644
--- a/babel/messages/extract.py
+++ b/babel/messages/extract.py
@@ -316,6 +316,47 @@ def extract_from_file(
                             options, strip_comment_tags))
 
 
+def _match_messages_against_spec(lineno, messages, comments, fileobj, spec):
+    translatable = []
+    context = None
+
+    # last_index is 1 based like the keyword spec
+    last_index = len(messages)
+    for index in spec:
+        if isinstance(index, tuple): # (n, 'c')
+            context = messages[index[0] - 1]
+            continue
+        if last_index < index:
+            # Not enough arguments
+            return
+        message = messages[index - 1]
+        if message is None:
+            return
+        translatable.append(message)
+
+    # An empty string msgid isn't valid, emit a warning
+
+    # keyword spec indexes are 1 based, therefore '-1'
+    if isinstance(spec[0], tuple):
+        # context-aware *gettext method
+        first_msg_index = spec[1] - 1
+    else:
+        first_msg_index = spec[0] - 1
+    if not messages[first_msg_index]:
+        filename = (getattr(fileobj, "name", None) or "(unknown)")
+        sys.stderr.write(
+            f"{filename}:{lineno}: warning: Empty msgid.  It is reserved by GNU gettext: gettext(\"\") "
+            f"returns the header entry with meta information, not the empty string.\n"
+        )
+        return
+
+    translatable = tuple(translatable)
+    if len(translatable) == 1:
+        translatable = translatable[0]
+
+    yield lineno, translatable, comments, context
+
+
 def extract(
     method: _ExtractionMethod,
     fileobj: _FileObj,
@@ -401,15 +442,18 @@ def extract(
                    options=options or {})
 
     for lineno, funcname, messages, comments in results:
-        specs = keywords[funcname] or (1,) if funcname else (1,)
-        if not isinstance(specs, dict): # for backwards compatibility
-            specs = {None: specs}
         if not isinstance(messages, (list, tuple)):
             messages = [messages]
         if not messages:
             continue
 
-        # Validate the messages against the keyword's specification
+        specs = keywords[funcname] or None if funcname else None
+        # {None: x} may be collapsed into x for backwards compatibility.
+        if not isinstance(specs, dict):
+            specs = {None: specs}
+
+        if strip_comment_tags:
+            _strip_comment_tags(comments, comment_tags)
 
         # None matches all arities.
         for arity in (None, len(messages)):
@@ -417,49 +461,9 @@ def extract(
                 spec = specs[arity]
             except KeyError:
                 continue
-            context = None
-            msgs = []
-            invalid = False
-            # last_index is 1 based like the keyword spec
-            last_index = len(messages)
-            for index in spec:
-                if isinstance(index, tuple):
-                    context = messages[index[0] - 1]
-                    continue
-                if last_index < index:
-                    # Not enough arguments
-                    invalid = True
-                    break
-                message = messages[index - 1]
-                if message is None:
-                    invalid = True
-                    break
-                msgs.append(message)
-            if invalid:
-                continue
-
-            # keyword spec indexes are 1 based, therefore '-1'
-            if isinstance(spec[0], tuple):
-                # context-aware *gettext method
-                first_msg_index = spec[1] - 1
-            else:
-                first_msg_index = spec[0] - 1
-            if not messages[first_msg_index]:
-                # An empty string msgid isn't valid, emit a warning
-                filename = (getattr(fileobj, "name", None) or "(unknown)")
-                sys.stderr.write(
-                    f"{filename}:{lineno}: warning: Empty msgid.  It is reserved by GNU gettext: gettext(\"\") "
-                    f"returns the header entry with meta information, not the empty string.\n"
-                )
-                continue
-
-            msgs = tuple(msgs)
-            if len(msgs) == 1:
-                msgs = msgs[0]
-
-            if strip_comment_tags:
-                _strip_comment_tags(comments, comment_tags)
-            yield lineno, msgs, comments, context
+            if spec is None:
+                spec = (1,)
+            yield from _match_messages_against_spec(lineno, messages, comments, fileobj, spec)
 
 
 def extract_nothing(

From 116e3e98a3f8b873476a3d0fea59e8672627e5eb Mon Sep 17 00:00:00 2001
From: Jean Abou Samra <jean@abou-samra.fr>
Date: Tue, 1 Aug 2023 12:30:35 +0200
Subject: [PATCH 6/9] Fix typo

---
 docs/cmdline.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/cmdline.rst b/docs/cmdline.rst
index d314f2e93..c580e908b 100644
--- a/docs/cmdline.rst
+++ b/docs/cmdline.rst
@@ -155,7 +155,7 @@ The meaning of ``--keyword`` values is as follows:
   behaves as ``gettext`` (argument is a message) or ``pgettext``
   (arguments are a context and a message) depending on whether it takes
   one or two arguments, you can pass
-  ``--keyword=gettext:1t --keyword=pgettext:1c,2,2t``.
+  ``--keyword=gettext:1,1t --keyword=pgettext:1c,2,2t``.
 
 The default keywords are equivalent to passing ::
 

From 237278fafeac0c01473682d7b00b2ee9aca6b4dd Mon Sep 17 00:00:00 2001
From: Jean Abou Samra <jean@abou-samra.fr>
Date: Tue, 1 Aug 2023 12:33:15 +0200
Subject: [PATCH 7/9] Still typoed :-)

---
 docs/cmdline.rst | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/docs/cmdline.rst b/docs/cmdline.rst
index c580e908b..e1328fe8f 100644
--- a/docs/cmdline.rst
+++ b/docs/cmdline.rst
@@ -151,11 +151,11 @@ The meaning of ``--keyword`` values is as follows:
   depending on how many arguments they take. For this use case, you can
   add an integer followed by "t" after the colon. In this case, the
   keyword will only match a function invocation if it has the specified
-  total number of arguments.  For example, if you have a function that
-  behaves as ``gettext`` (argument is a message) or ``pgettext``
-  (arguments are a context and a message) depending on whether it takes
-  one or two arguments, you can pass
-  ``--keyword=gettext:1,1t --keyword=pgettext:1c,2,2t``.
+  total number of arguments.  For example, if you have a function
+  ``foo`` that behaves as ``gettext`` (argument is a message) or
+  ``pgettext`` (arguments are a context and a message) depending on
+  whether it takes one or two arguments, you can pass
+  ``--keyword=foo:1,1t --keyword=foo:1c,2,2t``.
 
 The default keywords are equivalent to passing ::
 

From 464279df363e1730962acb3aa3ab1512b47a0a28 Mon Sep 17 00:00:00 2001
From: Jean Abou Samra <jean@abou-samra.fr>
Date: Mon, 4 Sep 2023 17:46:46 +0200
Subject: [PATCH 8/9] Further code review suggestions

---
 babel/messages/extract.py  | 12 +++++++-----
 babel/messages/frontend.py |  4 ++--
 2 files changed, 9 insertions(+), 7 deletions(-)

diff --git a/babel/messages/extract.py b/babel/messages/extract.py
index e53c3a626..dd78f4a5b 100644
--- a/babel/messages/extract.py
+++ b/babel/messages/extract.py
@@ -316,7 +316,8 @@ def extract_from_file(
                             options, strip_comment_tags))
 
 
-def _match_messages_against_spec(lineno, messages, comments, fileobj, spec):
+def _match_messages_against_spec(lineno: int, messages: list[str|None], comments: list[str],
+                                 fileobj: _FileObj, spec: tuple[int|tuple, ...]):
     translatable = []
     context = None
 
@@ -334,14 +335,13 @@ def _match_messages_against_spec(lineno, messages, comments, fileobj, spec):
             return
         translatable.append(message)
 
-    # An empty string msgid isn't valid, emit a warning
-
     # keyword spec indexes are 1 based, therefore '-1'
     if isinstance(spec[0], tuple):
         # context-aware *gettext method
         first_msg_index = spec[1] - 1
     else:
         first_msg_index = spec[0] - 1
+    # An empty string msgid isn't valid, emit a warning
     if not messages[first_msg_index]:
         filename = (getattr(fileobj, "name", None) or "(unknown)")
         sys.stderr.write(
@@ -354,7 +354,7 @@ def _match_messages_against_spec(lineno, messages, comments, fileobj, spec):
     if len(translatable) == 1:
         translatable = translatable[0]
 
-    yield lineno, translatable, comments, context
+    return lineno, translatable, comments, context
 
 
 def extract(
@@ -463,7 +463,9 @@ def extract(
                 continue
             if spec is None:
                 spec = (1,)
-            yield from _match_messages_against_spec(lineno, messages, comments, fileobj, spec)
+            result = _match_messages_against_spec(lineno, messages, comments, fileobj, spec)
+            if result is not None:
+                yield result
 
 
 def extract_nothing(
diff --git a/babel/messages/frontend.py b/babel/messages/frontend.py
index feeadf025..062b73bdd 100644
--- a/babel/messages/frontend.py
+++ b/babel/messages/frontend.py
@@ -1111,7 +1111,7 @@ def parse_mapping(fileobj, filename=None):
 
     return method_map, options_map
 
-def _parse_spec(s):
+def _parse_spec(s: str) -> tuple[int | None, tuple[int|tuple[int, str], ...]]:
     inds = []
     number = None
     for x in s.split(','):
@@ -1165,7 +1165,7 @@ def parse_keywords(strings: Iterable[str] = ()):
 
     # For best backwards compatibility, collapse {None: x} into x.
     for k, v in keywords.items():
-        if set(v.keys()) == {None}:
+        if set(v) == {None}:
             keywords[k] = v[None]
 
     return keywords

From be33f02f483f5fa0288fd2ce89057027b04ce173 Mon Sep 17 00:00:00 2001
From: Jean Abou Samra <jean@abou-samra.fr>
Date: Wed, 6 Sep 2023 16:42:06 +0200
Subject: [PATCH 9/9] Fix: add "from __future__ import annotations"

---
 babel/messages/extract.py  | 2 +-
 babel/messages/frontend.py | 2 ++
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/babel/messages/extract.py b/babel/messages/extract.py
index dd78f4a5b..4fc3da74b 100644
--- a/babel/messages/extract.py
+++ b/babel/messages/extract.py
@@ -317,7 +317,7 @@ def extract_from_file(
 
 
 def _match_messages_against_spec(lineno: int, messages: list[str|None], comments: list[str],
-                                 fileobj: _FileObj, spec: tuple[int|tuple, ...]):
+                                 fileobj: _FileObj, spec: tuple[int|tuple[int, str], ...]):
     translatable = []
     context = None
 
diff --git a/babel/messages/frontend.py b/babel/messages/frontend.py
index 062b73bdd..0008a9b84 100644
--- a/babel/messages/frontend.py
+++ b/babel/messages/frontend.py
@@ -8,6 +8,8 @@
     :license: BSD, see LICENSE for more details.
 """
 
+from __future__ import annotations
+
 import datetime
 import fnmatch
 import logging