Different Browsers Use Different Non-Matching Captured RegEx Pattern Values

Posted May 7, 2010 at 8:40 AM by Ben Nadel

Tags: Javascript / DHTML

Yesterday, Alec Perkins brought it to my attention that Chrome was passing in an undefined value for a non-matching, captured group in a regular expression (RegEx) replace function. This is a different behavior than Firefox, which passes in an empty string for non-matching, captured groups. I typically only test my Javascript in Firefox (yes, it's a bad habit); so, after hearing this, I wanted to test this feature in a few different browsers to see how they all behaved. The demo for the experiment is quite simple - I'm just matching a pattern than has an optional, but non-matched group:

  • <!DOCTYPE HTML>
  • <html>
  • <head>
  • <title>Javascript String Replace Data Types</title>
  • <script type="text/javascript">
  •  
  • // Create a string to test.
  • var test = "You are just too sexy!";
  •  
  • // Create a pattern to match a sub-string with an optional
  • // capturing group - "way" is not required.
  • var pattern = new RegExp( "((way )?too sexy)", "gi" );
  •  
  • // Loop over the matches, passing each set of captured groups
  • // to the given function. In this case, the second captured
  • // group will be the OPTIONAL part of the pattern.
  • test.replace(
  • pattern,
  • function( $0, $1, $2 ){
  •  
  • // Alert the TYPE of data type passed-in for the non-
  • // captured group value.
  • alert( typeof( $2 ) );
  •  
  • // Alert the boolean-cast of the group.
  • alert( !!$2 );
  •  
  • }
  • );
  •  
  • </script>
  • </head>
  • <body>
  • <!--- Left intentially blank. --->
  • </body>
  • </html>

Here, I am looking for the regular expression pattern:

"((way )?too sexy)"

In this pattern, the substring, "way ", is an optional, captured group and will, in fact, not be matched in our test string, "You are just too sexy!". In this scenario, we can then see how the different browsers pass in the second captured (but not matched) group.

I tested this in Firefox, Chrome, IE, and Safari and here is what we get:

Firefox:
string
false

Chrome:
undefined
false

Safari:
undefined
false

IE:
undefined
false

As it turns out, Firefox is the odd man out here, being the only browser that passes in an empty string for non-matching groups. All of the other browsers pass in "undefined" / null data types for non-matching groups. While this difference is strange, the upside is that both empty strings and undefined values are "falsey" values in Javascript and can be implicitly cast to False for control flow decisions. I know some people frown upon the use of truthy / falsey values as Conditions, but when different browsers pass in different falsey data types, this feature really helps us to simplify our control flow logic.

Note to self: be sure to test Javascript in at least two browsers!




Reader Comments

Jun 14, 2010 at 10:24 AM // reply »
172 Comments

This is a Firefox bug in the handling of nonparticipating capturing groups. The following (old) post of mine lists additional related browser bugs and other info: http://blog.stevenlevithan.com/archives/npcg-javascript


Jun 14, 2010 at 2:13 PM // reply »
11,243 Comments

@Steve,

Ahh, bug! Thanks for the link. I'll be checking out your regex know-how :)


Post A Comment

Comment Etiquette: Please do not post spam. Please keep the comments on-topic. Please do not post unrelated questions or large chunks of code. And, above all, please be nice to each other - we're trying to have a good conversation here.

Please review the following issues:

Author Name:


Author Email:

Author Website:

Comment:

Supported HTML tags for formatting: <strong>bold</strong>   <em>italic</em>   <code>code</code>







  • Help Wanted - Find Your Next ColdFusion Job
Ben Nadel's Company - Epicenter Consulting Recent Blog Comments
May 23, 2013 at 5:19 AM
Ask Ben: Print Part Of A Web Page With jQuery
How to print also the background color of table cells and table lines ... read »
May 23, 2013 at 3:55 AM
Javascript Array Methods: Unshift(), Shift(), Push(), And Pop()
very interesting and helpful too. ... read »
May 22, 2013 at 5:35 PM
Script Tags, jQuery, And Html(), Text() And Contents()
This is still an issue 2 years later. jQuery is supposed to remediate these cross browser issues, no? I have been unable to find any statement from the jQuery team calling this behavior "by de ... read »
May 22, 2013 at 12:44 PM
Ask Ben: Query Loop Inside CFScript Tags
In cf10, if you call a function that has: local.result = {}; local.result.msg = ""; local.svc = new query(); local.svc.setSQL("SELECT * FROM..."); local.obj = local.svc.exe ... read »
May 22, 2013 at 12:29 PM
Strange Interaction Between DeserializeJson(), ArrayContains(), And Database Values In ColdFusion
@Ben: What version of Java are you using? Also, did you test users.id to see what Java reports as the data type? I wonder if it's not a Java primitive data type, but getting returned as something ... read »
May 22, 2013 at 11:47 AM
Strange Interaction Between DeserializeJson(), ArrayContains(), And Database Values In ColdFusion
@Dana, Awesome - so it looks like this bug was fixed in ColdFusion 10. Thanks so much for double-checking that. ... read »
May 22, 2013 at 11:37 AM
Strange Interaction Between DeserializeJson(), ArrayContains(), And Database Values In ColdFusion
When I c&p and run on cf10, I get: Selected User IDs: 1,4 User 1 selected: YES - YES User 2 selected: NO - NO User 3 selected: NO - NO User 4 selected: YES - YES User 5 selected: NO - ... read »
May 22, 2013 at 11:27 AM
Strange Interaction Between DeserializeJson(), ArrayContains(), And Database Values In ColdFusion
@Tom, Good thought, but no dice. Both of these still exhibit the same behavior: users.id[ users.currentRow ] users[ "id" ][ users.currentRow ] It's just something whacky happening with ... read »
InVision App - Prototyping Made Beautiful With Prototyping Tools