ansaurus

Question

C# Regex for a username with a few restrictions

Answer 1

+7 A:

It is not so clean to express a set of unrelated rules in a single regular expression, but it can be done by using lookaround assertions (Rubular):

@"^(?=[A-Za-z0-9])(?!.*[._()\[\]-]{2})[A-Za-z0-9._()\[\]-]{3,15}$"

Explanation:

(?=[A-Za-z0-9])            Must start with a letter or number
(?!.*[._()\[\]-]{2})       Cannot contain two consecutive symbols
[A-Za-z0-9._()\[\]-]{3,15} Must consist of between 3 to 15 allowed characters

You might want to consider if this would be easier to read and more maintable as a list of simpler regular expressions, all of which must validate successfully, or else write it in ordinary C# code.

Mark Byers 2010-08-27 22:51:34

Using a regex for a length check is inefficient, unclear, and subject to breakage. (I think Mark's code is broken, as it looks like it requires a letter or number followed by 3-15 MORE characters, so it actually matches 4-16 characters.) Using a separate test for each requirement will be far easier to understand and maintain.

Ben Voigt 2010-08-27 22:58:08

Thank you for the speedy reply. This did work too, but the period is missing from the (?!.*[_()\[\]-\{2}) part.should be (?!.*[._()\[\]-]{2}) otherwise a period can follow any symbol

Marlon 2010-08-27 22:58:43

@Ben: unclear, yes, but Mark's code uses zero-width assertion (it consumes no characters), so the length constraint is preserved.

Chris Schmich 2010-08-27 22:59:41

@Marlon: You are correct, that was a copy/paste error. I've updated my post.

Mark Byers 2010-08-27 23:02:16

I think this is a clear and succinct way to describe the rules. The only ugliness is that all the rules are mashed together in one string, rather than being split up into 3 variables. To mitigate this, you could use string concatenation and split up the different rules into multiple lines: `"^" <newline> + "(?=[A-Z...])" <newline> + "(?!...)" <newline> + "[A-Z...]{3,15}" <newline> + "$"`

Merlyn Morgan-Graham 2010-08-28 01:13:29

Answer 2

+5 A:

As an optimization to Mark's answer:

^(?=.{3,15}$)([A-Za-z0-9][._()\[\]-]?)*$

Explanation:

(?=.{3,15}$)                   Must be 3-15 characters in the string
([A-Za-z0-9][._()\[\]-]?)*   The string is a sequence of alphanumerics,
                               each of which may be followed by a symbol

This one permits Unicode alphanumerics:

^(?=.{3,15}$)((\p{L}|\p{N})[._()\[\]-]?)*$

This one is the Unicode variant, plus uses non-capturing groups:

^(?=.{3,15}$)(?:(?:\p{L}|\p{N})[._()\[\]-]?)*$

Ben Voigt 2010-08-27 23:11:22

+1 That's quite a useful variation because it avoids repeating the definitions of "letter or number" and "symbol". You don't need the innermost parentheses.

Mark Byers 2010-08-27 23:15:05

You're right, that was a parenthesized atom. Redundant parens removed.

Ben Voigt 2010-08-27 23:19:51

It should also be faster, since there's only one lookahead, and it's much simpler.

Ben Voigt 2010-08-27 23:24:16

ansaurus

tags:

views:

answers:

C# Regex for a username with a few restrictions

related questions