Can this RegEx be simplified?

murphaph

I have a RegEx:
^(\d)+((((,(?!\d+\s))|\s)|((,(?!((\d+\s)|(\s+\d+\s+))))\s))\d+)*$

Its purpose is simple enough so I think I'm probably over complicating it.

It is used as a mask in a stripes (Java frontend framework) validation to validate user input.

The user is expected to enter a list of integers, separated by EITHER one comma OR one space OR one comma followed by one space. No other input formats are acceptable.

The user must be consistent however-no mixing comma separation with space separation (to prevent typos causing problems later). To this end I reject the entire input if there's any instance of a comma being followed by a space being followed by a digit being followed by a space OR any instance of a comma being followed by a digit followed by a space.

I'm fairly satisfied that the RegEx works...but is there a more elegant solution?

duffman85

I got this:

^(\d)+(\s|,|,\s)((\d+)(\2))*\d+

to match

", " => 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
" " => 1 2 3 4 5 6 7 8 9 10
"," => 1,2,3,4,5,6,7,8,9,10,


on the regex tester here: http://www.regular-expressions.info/javascriptexample.html

murphaph

Hi duffman, thx for your efforts.

I had a look but it matched the "1, 2" part from "1, 2 3, 4", which would not be desired as it would be a typo (missing comma-user wanted to enter "1, 2, 3, 4" and the validation mask is supposed to catch this and reject the whole string).I've been asked to also make sure a CSV with a trailing comma is rejected by the validator.

Using your solution with a "$" stuck on the end however seems perfect, and much neater than mine! Cheers.

I'm not great at RegEx-can you explain what the \2 bit does? My googling says \n means nth group/subpattern but I don't get what that means here?

Ah I get it, you use the \2 to say "whatever pattern in the second group (the group with the commas and spaces) matched, use this for all subsequent pattern matching"-very elegant, I'll be using that again, thanks duffman!

duffman85

murphaph wrote: »

Hi duffman, thx for your efforts.

I had a look but it matched the "1, 2" part from "1, 2 3, 4", which would not be desired as it would be a typo (missing comma-user wanted to enter "1, 2, 3, 4" and the validation mask is supposed to catch this and reject the whole string).I've been asked to also make sure a CSV with a trailing comma is rejected by the validator.

Using your solution with a "$" stuck on the end however seems perfect, and much neater than mine! Cheers.

No problem - glad it helped.

murphaph wrote: »

I'm not great at RegEx-can you explain what the \2 bit does? My googling says \n means nth group/subpattern but I don't get what that means here?

Ah I get it, you use the \2 to say "whatever pattern in the second group (the group with the commas and spaces) matched, use this for all subsequent pattern matching"-very elegant, I'll be using that again, thanks duffman!

Yes, that's it - very handy. I'm no regex expert myself, usually just trial & error and some googling.

A proper explanation of grouping and backreferences (\2 etc.) is here: http://www.regular-expressions.info/brackets.html

Procasinator

Would this work:

^\d+((\s\d+)*|(,\s\d+)*|(,\d+)*)$

Should match:

1 2 3 4
1,2,3,4
1, 2, 3, 4

If you don't mind:
1,2, 3, 4 (i.e. inconsistent usage of comma + space, but always a comma) you can change it to:

^\d+((\s\d+)*|(,\s?\d+)*)$

Cool Mo D

Surely the java string.split() method is a more appropriate approach? If you split on each delimiter in turn, and test if you can convert the any of the returned array of strings into an array of integers, you can avoid the messy corner cases the regexes can introduce, and is more readable to boot.

murphaph

It has to be a regex for the Stripes framework validation AFAIK. It's a mask applied to the input. I do something similar to what you suggest inside the ActionBean (Stripes name for a servlet) though. Cheers.