[3.12] bpo-43950: handle wide unicode characters in tracebacks (GH-28150)

isidentical · pablogsal · commit d5ee672aacc5 · 2023-10-26T16:14:54.000+09:00
(cherry picked from commit 78e6d72) Co-authored-by: Batuhan Taskaya <isidentical@gmail.com>
diff --git a/Lib/test/test_traceback.py b/Lib/test/test_traceback.py
@@ -922,8 +922,63 @@ def f():
             f"  File \"{__file__}\", line {self.callable_line}, in get_exception",
             "    callable()",
             f"  File \"{__file__}\", line {f.__code__.co_firstlineno + 4}, in f",
-            "    print(1, ｗｗｗ(",
-            "             ^^^^",
+            f"    print(1, ｗｗｗ(",
+            f"             ^^^^^^^",
+        ]
+        self.assertEqual(actual, expected)
+
+    def test_byte_offset_with_wide_characters_term_highlight(self):
+        def f():
+            说明说明 = 1
+            şçöğıĤellö = 0 # not wide but still non-ascii
+            return 说明说明 / şçöğıĤellö
+
+        actual = self.get_exception(f)
+        expected = [
+            f"Traceback (most recent call last):",
+            f"  File \"{__file__}\", line {self.callable_line}, in get_exception",
+            f"    callable()",
+            f"  File \"{__file__}\", line {f.__code__.co_firstlineno + 3}, in f",
+            f"    return 说明说明 / şçöğıĤellö",
+            f"           ~~~~~~~~~^~~~~~~~~~~~",
+        ]
+        self.assertEqual(actual, expected)
+
+    def test_byte_offset_with_emojis_term_highlight(self):
+        def f():
+            return "✨🐍" + func_说明说明("📗🚛",
+                "📗🚛") + "🐍"
+
+        actual = self.get_exception(f)
+        expected = [
+            f"Traceback (most recent call last):",
+            f"  File \"{__file__}\", line {self.callable_line}, in get_exception",
+            f"    callable()",
+            f"  File \"{__file__}\", line {f.__code__.co_firstlineno + 1}, in f",
+            f'    return "✨🐍" + func_说明说明("📗🚛",',
+            f"                    ^^^^^^^^^^^^^",
+        ]
+        self.assertEqual(actual, expected)
+
+    def test_byte_offset_wide_chars_subscript(self):
+        def f():
+            my_dct = {
+                "✨🚛✨": {
+                    "说明": {
+                        "🐍🐍🐍": None
+                    }
+                }
+            }
+            return my_dct["✨🚛✨"]["说明"]["🐍"]["说明"]["🐍🐍"]
+
+        actual = self.get_exception(f)
+        expected = [
+            f"Traceback (most recent call last):",
+            f"  File \"{__file__}\", line {self.callable_line}, in get_exception",
+            f"    callable()",
+            f"  File \"{__file__}\", line {f.__code__.co_firstlineno + 8}, in f",
+            f'    return my_dct["✨🚛✨"]["说明"]["🐍"]["说明"]["🐍🐍"]',
+            f"           ~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^",
         ]
         self.assertEqual(actual, expected)
 
diff --git a/Lib/traceback.py b/Lib/traceback.py
@@ -470,39 +470,49 @@ def format_frame_summary(self, frame_summary):
             stripped_line = frame_summary.line.strip()
             row.append('    {}\n'.format(stripped_line))
 
-            orig_line_len = len(frame_summary._original_line)
+            line = frame_summary._original_line
+            orig_line_len = len(line)
             frame_line_len = len(frame_summary.line.lstrip())
             stripped_characters = orig_line_len - frame_line_len
             if (
                 frame_summary.colno is not None
                 and frame_summary.end_colno is not None
             ):
                 start_offset = _byte_offset_to_character_offset(
-                    frame_summary._original_line, frame_summary.colno) + 1
+                    line, frame_summary.colno)
                 end_offset = _byte_offset_to_character_offset(
-                    frame_summary._original_line, frame_summary.end_colno) + 1
+                    line, frame_summary.end_colno)
+                code_segment = line[start_offset:end_offset]
 
                 anchors = None
                 if frame_summary.lineno == frame_summary.end_lineno:
                     with suppress(Exception):
-                        anchors = _extract_caret_anchors_from_line_segment(
-                            frame_summary._original_line[start_offset - 1:end_offset - 1]
-                        )
+                        anchors = _extract_caret_anchors_from_line_segment(code_segment)
                 else:
-                    end_offset = stripped_characters + len(stripped_line)
+                    # Don't count the newline since the anchors only need to
+                    # go up until the last character of the line.
+                    end_offset = len(line.rstrip())
 
                 # show indicators if primary char doesn't span the frame line
                 if end_offset - start_offset < len(stripped_line) or (
                         anchors and anchors.right_start_offset - anchors.left_end_offset > 0):
+                    # When showing this on a terminal, some of the non-ASCII characters
+                    # might be rendered as double-width characters, so we need to take
+                    # that into account when calculating the length of the line.
+                    dp_start_offset = _display_width(line, start_offset) + 1
+                    dp_end_offset = _display_width(line, end_offset) + 1
+
                     row.append('    ')
-                    row.append(' ' * (start_offset - stripped_characters))
+                    row.append(' ' * (dp_start_offset - stripped_characters))
 
                     if anchors:
-                        row.append(anchors.primary_char * (anchors.left_end_offset))
-                        row.append(anchors.secondary_char * (anchors.right_start_offset - anchors.left_end_offset))
-                        row.append(anchors.primary_char * (end_offset - start_offset - anchors.right_start_offset))
+                        dp_left_end_offset = _display_width(code_segment, anchors.left_end_offset)
+                        dp_right_start_offset = _display_width(code_segment, anchors.right_start_offset)
+                        row.append(anchors.primary_char * dp_left_end_offset)
+                        row.append(anchors.secondary_char * (dp_right_start_offset - dp_left_end_offset))
+                        row.append(anchors.primary_char * (dp_end_offset - dp_start_offset - dp_right_start_offset))
                     else:
-                        row.append('^' * (end_offset - start_offset))
+                        row.append('^' * (dp_end_offset - dp_start_offset))
 
                     row.append('\n')
 
@@ -623,6 +633,25 @@ def _extract_caret_anchors_from_line_segment(segment):
 
     return None
 
+_WIDE_CHAR_SPECIFIERS = "WF"
+
+def _display_width(line, offset):
+    """Calculate the extra amount of width space the given source
+    code segment might take if it were to be displayed on a fixed
+    width output device. Supports wide unicode characters and emojis."""
+
+    # Fast track for ASCII-only strings
+    if line.isascii():
+        return offset
+
+    import unicodedata
+
+    return sum(
+        2 if unicodedata.east_asian_width(char) in _WIDE_CHAR_SPECIFIERS else 1
+        for char in line[:offset]
+    )
+
+
 
 class _ExceptionPrintContext:
     def __init__(self):
