ansaurus

Question

Answer 1

A:

It's counting the lookahead after the digits because you allowed as much with

(?=.*[A-Za-z])

The .* means "after zero or more characters."

If you want to force a letter at the beginning, modify your pattern:

^[A-Za-z][-a-zA-Z0-9@_+.]{5,31}$

Greg Bacon 2010-01-25 19:53:54

So how do I fix that?

John Batdorf 2010-01-25 19:55:09

`(?= )` is a **zero width** positive lookahead assertion, so the "cursor" is being reset after the check.

Richard Szalay 2010-01-25 19:56:48

He doesn't want to assert a letter at the start, only that the string *contains* a letter.

Richard Szalay 2010-01-25 20:01:06

That is correct Richard, thanks.

John Batdorf 2010-01-25 20:57:09

No probs! How'd the regex from my answer fair?

Richard Szalay 2010-01-25 21:41:52

Then we're missing a piece of the puzzle somewhere. The example strings you gave all match the pattern from your question. Do your program's actual inputs, for example, contain more than than the 6- to 32-character fields?

Greg Bacon 2010-01-25 21:44:07

But they don't match. Try it in .NET 1.1 in IE.

John Batdorf 2010-01-26 23:49:38

Richard it didn't work.

John Batdorf 2010-01-26 23:50:09

Answer 2

+2 A:

It's quite possible that it's a bug in .NET 1.1. All of your "fail" examples work using Regex Hero, which is based on Silverlight which uses the .NET 2.0 Regex implementation.

You might try using a positive look-behind assertion instead and see if that gets around the problem:

^[a-zA-Z0-9@\-_\+\.]{6,32}(?<=.*[A-Za-z])$

Edit: Considering this is an ASP.NET validator, you should double check that it's not failing client validation (javascript). Some advanced features (like zero-width look ahead/behinds) are not supported by some browsers.

Try to disable client side validation by setting EnableClientScript to false on the RegularExpressionValidator and see if that fixes the problem. If it does, then it's a browser support issue and I'd recommend splitting your validation into two:

^[a-zA-Z0-9@\-_\+\.]{6,32}$ # must be 6-32 characters
^.*[A-Za-z].*$ # must contain a letter

Richard Szalay 2010-01-25 19:58:59

Does .NET support variable-width lookbehinds?

Anon. 2010-01-25 20:07:23

I don't believe so, though I'm not familiar with the concept.

Richard Szalay 2010-01-25 20:18:55

Yes, .NET regexes support unknown-width lookbehinds. At least, they do in .NET 3.5; whether they did in .NET 1.1, and whether it worked correctly, I couldn't say.

Alan Moore 2010-01-25 22:13:29

I see what you mean now (they are referred to as zero-width lookbehinds in .NET), and yes they are supported.

Richard Szalay 2010-01-26 07:58:37

Actually, "zero-width" merely refers to the fact that they don't consume any characters; lookaheads, lookbehinds, word boundaries (`\b`) and anchors (`^`, `$`, `\A`, `\Z`, `\z`) are all zero-width assertions. "Unknown-width" was my ill-advised attempt to abbreviate "able to match a variable number of characters, with no restrictions". Java and PHP (PCRE) lookbehinds aren't limited to fixed-width matches like most other flavors', but they aren't completely unrestricted like .NET's lookbehinds either.

Alan Moore 2010-01-30 14:20:22

That's a good suggestion about the client-side validation. JavaScript regexes do support lookaheads (which is all the OP is using), but that support is notoriously inconsistent among the myriad browsers/versions. And if your dual-validator idea doesn't work, it's time to get out the sledgehammer. :P

Alan Moore 2010-01-30 18:02:18

Answer 3

+1 A:

Have you tried refactoring the regex? For example:

^(?=[a-zA-Z0-9_@+.-]{6,32}$).*[A-Za-z].*$

Or just "pad" your lookahead so it has to match all the way to the end:

^(?=.*[A-Za-z].*$)[a-zA-Z0-9_@+.-]{6,32}$

Maybe that will reset the match position so the second part can start matching at the beginning. It shouldn't be necessary, of course, but I can't see any reason why your regex wouldn't work as written.

Alan Moore 2010-01-30 15:44:51

ansaurus

tags:

views:

answers:

.NET Regex Not working as it should

related questions