| """One reversible source-coordinate view for hyphenated OCR line wraps.""" | |
| import re | |
| LINE_WRAP = re.compile(r"(?<=[^\W\d_])-\r?\n(?=[^\W\d_])", re.UNICODE) | |
| def dehyphenate(text): | |
| removed = {i for match in LINE_WRAP.finditer(text) for i in range(match.start(), match.end())} | |
| positions = tuple(i for i in range(len(text)) if i not in removed) | |
| return "".join(text[i] for i in positions), positions | |