diff --git a/Lib/test/test_traceback.py b/Lib/test/test_traceback.py index 40663e7d887e14b..bcae787ec9e2d51 100644 --- a/Lib/test/test_traceback.py +++ b/Lib/test/test_traceback.py @@ -148,6 +148,40 @@ def test_caret_wide_char(self): # "女女女 for 女女女 in range(30)" has display width 30 self.assertEqual(err[2].count("^"), 30) + def test_mismatched_bracket_carets(self): + cases = ( + ('obj{"some long expression"]', 3, 26, + ' ^ ^\n'), + ('obj[func(x[store[arg, a[i], b[j])])[y]]', 16, 32, + ' ^ ^\n'), + ('x = ("é女"]', 4, 9, + ' ^ ^\n'), + ('α = obj{1]', 7, 9, + ' ^ ^\n'), + ("f'{a[4)}'", 4, 6, + ' ^ ^\n'), + ) + for source, opening, closing, expected_carets in cases: + with self.subTest(source=source): + with self.assertRaises(SyntaxError) as cm: + compile(source, '', 'exec') + exc = cm.exception + self.assertEqual(exc.offset, closing + 1) + self.assertEqual(exc._opening_offset, opening + 1) + formatted = traceback.format_exception_only(exc) + self.assertEqual(formatted[2], expected_carets) + colorized = ''.join(traceback.format_exception_only(exc, colorize=True)) + self.assertEqual( + re.sub(r'\x1b\[[0-9;]*m', '', colorized), ''.join(formatted) + ) + + # The existing one-caret display remains for brackets on different lines. + with self.assertRaises(SyntaxError) as cm: + compile('x = (\n 1 + ]', '', 'exec') + exc = cm.exception + self.assertFalse(hasattr(exc, '_opening_offset')) + self.assertEqual(traceback.format_exception_only(exc)[2], ' ^\n') + def test_nocaret(self): exc = SyntaxError("error", ("x.py", 23, None, "bad syntax")) err = traceback.format_exception_only(SyntaxError, exc) diff --git a/Lib/traceback.py b/Lib/traceback.py index 3ecb1dd318fbcb7..812e32e8e307d69 100644 --- a/Lib/traceback.py +++ b/Lib/traceback.py @@ -1183,6 +1183,7 @@ def __init__(self, exc_type, exc_value, exc_traceback, *, limit=None, self.msg = exc_value.msg self._is_syntax_error = True self._exc_metadata = getattr(exc_value, "_metadata", None) + self._opening_offset = getattr(exc_value, "_opening_offset", None) elif exc_type and issubclass(exc_type, ImportError) and \ getattr(exc_value, "name_from", None) is not None: wrong_name = getattr(exc_value, "name_from", None) @@ -1592,21 +1593,41 @@ def _format_syntax_error(self, stype, **kwargs): dp_colno = _display_width(ltext, colno) highlighted = ltext[colno:end_colno] caret_count = _display_width(highlighted) if highlighted else (end_colno - colno) + opening_colno = None + if ( + isinstance(self._opening_offset, int) + and self.lineno == self.end_lineno + ): + opening_colno = self._opening_offset - 1 - spaces + if not 0 <= opening_colno < colno: + opening_colno = None + carets = '^' * caret_count + if opening_colno is not None: + dp_opening_colno = _display_width(ltext, opening_colno) + carets = '^' + ' ' * (dp_colno - dp_opening_colno - 1) + carets + dp_colno = dp_opening_colno start_color = end_color = "" if colorize: - # colorize from colno to end_colno + # Colorize the later range first so the earlier + # opening column is unaffected by escape sequences. ltext = ( ltext[:colno] + theme.error_highlight + ltext[colno:end_colno] + theme.reset + ltext[end_colno:] ) + if opening_colno is not None: + ltext = ( + ltext[:opening_colno] + + theme.error_highlight + ltext[opening_colno] + theme.reset + + ltext[opening_colno + 1:] + ) start_color = theme.error_highlight end_color = theme.reset yield ' {}\n'.format(ltext) yield ' {}{}{}{}\n'.format( ' ' * dp_colno, start_color, - '^' * caret_count, + carets, end_color, ) else: diff --git a/Misc/NEWS.d/next/Core_and_Builtins/2026-11-13-00-00-00.gh-issue-98278.4vK9pC.rst b/Misc/NEWS.d/next/Core_and_Builtins/2026-11-13-00-00-00.gh-issue-98278.4vK9pC.rst new file mode 100644 index 000000000000000..f2f381347279a44 --- /dev/null +++ b/Misc/NEWS.d/next/Core_and_Builtins/2026-11-13-00-00-00.gh-issue-98278.4vK9pC.rst @@ -0,0 +1,3 @@ +When mismatched brackets or parentheses are on the same line, the +:exc:`SyntaxError` display now points to both the opening and closing +delimiters. diff --git a/Parser/lexer/lexer.c b/Parser/lexer/lexer.c index 9049a0487c4ac20..362cc2ae77e2d80 100644 --- a/Parser/lexer/lexer.c +++ b/Parser/lexer/lexer.c @@ -472,7 +472,8 @@ _PyLexer_get_normal(struct tok_state *tok, ftstring_state *current, struct token c, opening, tok->parenlinenostack[tok->level])); } else { - return MAKE_TOKEN(_PyTokenizer_syntaxerror(tok, + return MAKE_TOKEN(_PyTokenizer_syntaxerror_with_opening( + tok, tok->parencolstack[tok->level], "closing parenthesis '%c' does not match " "opening parenthesis '%c'", c, opening)); diff --git a/Parser/tokenizer/helpers.c b/Parser/tokenizer/helpers.c index c39e6756476b97c..081d69daca4f6fd 100644 --- a/Parser/tokenizer/helpers.c +++ b/Parser/tokenizer/helpers.c @@ -98,6 +98,39 @@ _PyTokenizer_syntaxerror(struct tok_state *tok, const char *format, ...) return ret; } +int +_PyTokenizer_syntaxerror_with_opening(struct tok_state *tok, + int opening_col_offset, + const char *format, ...) +{ + va_list vargs; + va_start(vargs, format); + const char *line = _PyTok_SourcePointer(&tok->source, tok->line_start); + int ret = _syntaxerror_range( + tok, line, tok->cur - tok->line_start, tok->lineno, + format, -1, -1, vargs); + va_end(vargs); + + /* Leave the error location on the closing bracket, but make the + opening bracket's character column available to traceback. */ + if (PyErr_ExceptionMatches(PyExc_SyntaxError)) { + PyObject *exc = PyErr_GetRaisedException(); + PyObject *offset = PyLong_FromLong( + byte_col_to_char_col(line, opening_col_offset + 1)); + if (offset != NULL) { + if (PyObject_SetAttrString(exc, "_opening_offset", offset) < 0) { + PyErr_Clear(); + } + Py_DECREF(offset); + } + else { + PyErr_Clear(); + } + PyErr_SetRaisedException(exc); + } + return ret; +} + int _PyTokenizer_syntaxerror_known_range(struct tok_state *tok, int col_offset, int end_col_offset, diff --git a/Parser/tokenizer/helpers.h b/Parser/tokenizer/helpers.h index f33d8f55ed1d17c..6146d600c2feaae 100644 --- a/Parser/tokenizer/helpers.h +++ b/Parser/tokenizer/helpers.h @@ -6,6 +6,9 @@ #include "../lexer/state.h" int _PyTokenizer_syntaxerror(struct tok_state *tok, const char *format, ...); +/* opening_col_offset is a 0-based byte column on the current line. */ +int _PyTokenizer_syntaxerror_with_opening( + struct tok_state *tok, int opening_col_offset, const char *format, ...); /* Positive range columns are 1-based byte columns. A start column of -1 derives the character column from the reporting cursor; an end column of -1 uses the start column. */