Python 2.7 Regex capture groups not working as predicted

Question

I am trying to pattern match and replace first person with second person with Python 2.7.

string = re.sub(r'(\W)I(\W)', '\g<1>you\g<2>',string)
string = re.sub(r'(\W)(me)(\W)', '\g<1>you\g<3>',string)
# but does NOT work
string = re.sub(r'(\W)I|(me)(\W)', '\g<1>you\g<3>',string)

I want to use the last regex, but somehow the capture groups are all messed up and even doing a \g<0> shows strange, irregular matches. I would think that capture group 3 would be the last word boundary, but it doesn't appear to be.

A sample sentence could be: I like candy.

I am not interested very much in the correctness of the replacement (me will never actually be selected since I goes first), but I don't know why the capture groups don't work as I would expect.

Thanks!

Then, there are at least 2 issues with the 3rd pattern: wrong alternation boundaries, and non-participating group replacement. Don't you get `"unmatched group"` error? — Wiktor Stribiżew, Apr 06 '16 at 10:20
I do that's exactly it. Can you explain where I went wrong? :) — Nick Anderson, Apr 06 '16 at 10:44
I can, I actually posted a long answer on that some time ago. See [*Empty string instead of unmatched group error*](http://stackoverflow.com/questions/35516298/empty-string-instead-of-unmatched-group-error/35516425#35516425). If you still do not understand the root cause of your pattern issue, I will post an answer with explanations. — Wiktor Stribiżew, Apr 06 '16 at 10:50
I'm looking over your answer and will upvote the original thread. Thank you so much! — Nick Anderson, Apr 06 '16 at 12:33
@WiktorStribiżew: Did I spotted **greedy** approach there ;-) ? P.S I was talking about cap of 200 :D IYKWIM ! — , Apr 06 '16 at 13:00

score 2 · Answer 1 · 2016-04-06T10:13:23.130

2

Try with following regex.

Regex: \b(I|me)\b

Explanation:

\b on both sides marks the word boundary.
(I|me) matches either I OR me.

Note:- You can make it case insensitive using i flag.

Regex101 Demo

edited Apr 06 '16 at 10:13

answered Apr 06 '16 at 10:12

That's a way better answer. But do you know why the alternation seems to mess up the capture groups such that capture group 3 is no longer the one I would expect? – Nick Anderson Apr 06 '16 at 10:13
@NickAnderson: Since you used `\W` instead of word boundary it fails when `I` is at beginning of string. `(\W)(me)|I(\W)` will work too [Demo](https://regex101.com/r/gS0iA7/2) but using word boundary is safer. – Apr 06 '16 at 10:19
This is true. But comparing: string = re.sub(r'(\W)I|(me)(\W)', '\g<1>you\g<3>',string) and string = re.sub(r'(\W)(I|me)(\W)', '\g<1>you\g<3>',string), why are the groups different? Why does making the capture group nested under the alternation operator change how it works? I saw on the Python docs that you can use <2> to reference a nested level also, so does the meaning change based on context? – Nick Anderson Apr 06 '16 at 10:19
It's not about capturing groups. It will work even if you don't use one around `me`. You will have to use for grouping all alteration cases. `(I|me)`. – Apr 06 '16 at 10:24
1

Thank you; that's very helpful! – Nick Anderson Apr 06 '16 at 12:33

Python 2.7 Regex capture groups not working as predicted

1 Answers1