Python regex strange behavior

Question

I've discovered something that I can't explain in Python re module. Compilation of (a*)* or (a*|b)* throws an error:

raise error, v # invalid expression sre_constants.error: nothing to repeat

I've tested this regexp in javascript and it seems to be ok.

Is it a bug?

Related: http://stackoverflow.com/questions/3675144/regex-error-nothing-to-repeat — Kobi, May 03 '11 at 12:49
I would add: logically, none of these makes sense. `(a*)*` is the same as `a*`, and `(a*|b)*` is the same as `[ab]*` (or `(a|b)*`). Is there a good use case, out of curiosity? — Kobi, May 03 '11 at 13:00
@Kobi maybe if it's a greedy match, whereby you match as long a string as possible but fewer instances - but for this you'd want `+` not `*` — theheadofabroom, May 03 '11 at 13:14

score 9 · Accepted Answer · answered May 03 '11 at 12:53

9

Yes, it's a bug (or at least a misfeature). It's complaining that if a* matches nothing, it doesn't know how to capture 0 or more "nothings".

answered May 03 '11 at 12:53

Mu Mind

10,935
4
38
69

score 5 · Answer 2 · answered May 03 '11 at 12:52

5

A bug in Python.

http://bugs.python.org/issue2537

http://bugs.python.org/issue214033

Maybe a "bug" is not the correct word here. Different kind of interpretation...

answered May 03 '11 at 12:52

lzap

16,417
12
71
108

1

Wow! I've found first bug in Python in my life =) – f0b0s May 03 '11 at 12:57

score 5 · Answer 3 · answered May 03 '11 at 12:55

5

a* can be null, giving (null)* which makes no sense to the interpreter. (a*|b) can also be null as it can evaluate to either (b) or (a*).you could use (a+)* and therefore (a+|b)*

answered May 03 '11 at 12:55

theheadofabroom

20,639
5
33
65

I absolutely agree with you, but anyway Mu Mind is right. – f0b0s May 03 '11 at 12:57
@f0b0s Essentially @Mu and I are saying the same thing - the only difference being how explicitly - that's why I upvoted him for getting in first. – theheadofabroom May 03 '11 at 13:11
I still don't get it. If cannot repeat null is a feature. Then why (a?)* is valid then? it can also give (null)*. – Wang Jan 15 '14 at 21:52
@Wang I have no idea why (a?)* would be valid, it seems like logically it should not be – theheadofabroom Jan 15 '14 at 22:22
It is a revert of a fixed bug. This has been fixed again recently. sigh! Python.... – Wang Jan 24 '14 at 05:29

score 1 · Answer 4 · answered Jun 14 '11 at 17:39

There is actually an important reason for Python to reject (a*)*, and (a*|b)*. Since * is greedy, it matches the longest string it can. The problem is that if the regex modified by * is empty, then the regular expression parser tries to match as many repetitions of empty string as possible. This means that it would match any number of empty strings in between any two characters in the string you test against. Since a* is in a capturing group, it would have to capture all of those empty strings, which would be impossible.

score 0 · Answer 5 · edited May 23 '17 at 11:58

0

This seems to be a python issue, see this http://bugs.python.org/issue214033

Also it is on StackO as well regex error - nothing to repeat

edited May 23 '17 at 11:58

Community

1
1

answered May 03 '11 at 12:58

bpgergo

15,669
5
44
68

Python regex strange behavior

5 Answers5

Linked