Fix spurious E231 for nested f-string format specifiers

When a replacement field appears inside an f-string format spec
(e.g. `f"{x:0.{digits:d}f}"`), the `:` in the nested `{digits:d}` is a
format specifier, not a dict colon.  Previously pycodestyle reported a
false-positive E231 on it because the brace-stack pattern `['f', '{']`
only matched the top-level replacement field.

Track `FSTRING_MIDDLE` / `TSTRING_MIDDLE` tokens: when `{` is opened
immediately after one of these tokens it is a nested replacement field
inside a format spec.  Push `'F'` (instead of `'{'`) onto the brace
stack, then suppress E231 for `:` when the top of the stack is `'F'`.

Fixes #1241.

Author: Friday

pycodestyle.py

@@ -874,9 +874,17 @@ def missing_whitespace(logical_line, tokens):
     prev_text = prev_end = None
     operator_types = (tokenize.OP, tokenize.NAME)
     brace_stack = []
+    _after_fstring_middle = False
     for token_type, text, start, end, line in tokens:
         if token_type == tokenize.OP and text in {'[', '(', '{'}:
-            brace_stack.append(text)
+            # After FSTRING_MIDDLE/TSTRING_MIDDLE, `{` opens a nested
+            # replacement field inside a format spec.  Mark it with 'F'
+            # so the `:` in e.g. f"{x:0.{digits:d}f}" is recognised as
+            # a format-specifier colon, not a dict colon.
+            if _after_fstring_middle and text == '{':  # pragma: >=3.12 cover
+                brace_stack.append('F')
+            else:
+                brace_stack.append(text)
         elif token_type == FSTRING_START:  # pragma: >=3.12 cover
             brace_stack.append('f')
         elif token_type == TSTRING_START:  # pragma: >=3.14 cover
@@ -897,6 +905,16 @@ def missing_whitespace(logical_line, tokens):
             ):
                 brace_stack.pop()
 
+        # Track whether the previous meaningful token was FSTRING_MIDDLE
+        # or TSTRING_MIDDLE (i.e. we just entered a format spec region).
+        if (  # pragma: >=3.12 cover
+                token_type == FSTRING_MIDDLE or
+                token_type == TSTRING_MIDDLE
+        ):
+            _after_fstring_middle = True
+        elif token_type not in SKIP_COMMENTS:
+            _after_fstring_middle = False
+
         if token_type in SKIP_COMMENTS:
             continue
 
@@ -912,6 +930,9 @@ def missing_whitespace(logical_line, tokens):
                 # 3.14+ tstring format specifier
                 elif text == ':' and brace_stack[-2:] == ['t', '{']:  # pragma: >=3.14 cover  # noqa: E501
                     pass
+                # nested replacement field in f/t-string format spec
+                elif text == ':' and brace_stack[-1:] == ['F']:  # pragma: >=3.12 cover  # noqa: E501
+                    pass
                 # tuple (and list for some reason?)
                 elif text == ',' and next_char in ')]':
                     pass

testing/data/python312.py

@@ -29,3 +29,8 @@ def g[T: str, U: int](x: T, y: U) -> dict[T, U]:
 f'{ an_error_now }'
 #: Okay
 f'{x:02x}'
+#: Okay
+# nested replacement field in format spec (issue #1241)
+f"{x:0.{digits:d}f}"
+#: Okay
+f'{value:{fill}{align}{width}.{precision}f}'