in src/latex2sympy2_extended/math_normalization.py [0:0]
def extract_boxed_content(text: str, mode: Literal["last", "all"] = "last") -> str:
"""
Find and extract all \\boxed{...} or \\fbox{...} elements from a string, searching from right to left.
If mode is "last", return content up to the last valid separator.
If mode is "all", return all boxed contents joined by commas.
"""
def find_content_boundaries(text: str, opening_brace_pos: int, max_pos: int) -> tuple[int, int] | None:
# Start searching for closing brace from the opening brace position
i = opening_brace_pos
num_left_braces_open = 1 # We start after the opening brace
while i + 1 < max_pos: # Check if next position is within bounds and max_pos
i += 1
if text[i] == "{":
num_left_braces_open += 1
elif text[i] == "}":
num_left_braces_open -= 1
if num_left_braces_open == 0:
return opening_brace_pos, i
return None
def has_valid_separator(text: str, content_end: int, next_boxed_start: int) -> bool:
between_text = text[content_end + 1:next_boxed_start]
# Making regex for it not worth it so this works
return len(between_text) < 70 and bool(VALID_SEPARATOR_PATTERN.search(between_text))
results = []
current_pos = len(text)
last_boxed_start = None
max_pos = len(text)
while True:
boxed_idx = text.rfind("\\boxed", 0, current_pos)
fbox_idx = text.rfind("\\fbox", 0, current_pos)
if boxed_idx < 0 and fbox_idx < 0:
break
start_idx = max(boxed_idx, fbox_idx)
command_end = start_idx + (6 if boxed_idx > fbox_idx else 5)
# Find opening brace
next_char_pos = command_end
while next_char_pos < max_pos and text[next_char_pos].isspace():
next_char_pos += 1
if next_char_pos >= max_pos:
break
if text[next_char_pos] == "{":
boundaries = find_content_boundaries(text, next_char_pos, max_pos)
if not boundaries:
# This is our last box
if len(results) == 0:
results.append(text[next_char_pos:])
break
content_start, content_end = boundaries
content = text[content_start + 1:content_end].strip()
if mode == "last" and last_boxed_start is not None:
if not has_valid_separator(text, content_end, last_boxed_start):
break
results.append(content)
last_boxed_start = start_idx
max_pos = start_idx
else:
# This is our last box
if len(results) == 0:
results.append(text[next_char_pos:])
# Otherwise we just ignore it
break
current_pos = start_idx
if not results:
return text
return ",".join(reversed(results))