问题
In a regex replacement pattern, a backreference looks like \1. If you want to include a digit after that backreference, this will fail because the digit is considered to be part of the backreference number:
# replace all twin digits by zeroes, but retain white space in between
re.sub(r"\d(\s*)\d", r"0\10", "0 1")
>>> sre_constants.error: invalid group reference
Substitution pattern r"0\1 0" would work fine but in the failing example back-reference \1 is interpreted as \10.
How can the digit '0' be separated from the back-reference \1 that precedes it?
回答1:
You can use \g<1>, as mentioned in the docs.
回答2:
Instead of using a backreference with a sequence number (\1), you can use named groups and the problem is solved:
# replace all twin digits by zeroes, but retain whitespace in between
re.sub(r"\d(?P<whitespace>\s*)\d", r"0\g<whitespace>0", "0 1")
>>> '0 0'
Turns out this trick is in fact described in the documentation of re.sub.
来源：https://stackoverflow.com/questions/16810523/python-regex-subsitution-separate-backreference-from-digit