Thursday, July 1, 2010

Using Java Regex Functions on CFML Strings

I ran into this today while working on ColdTonica, and since it's something I'm still surprised people forget (including myself) I thought I'd share.

ColdTonica is a CFML clone of StatusNet (formerly Laconica), which is an open source PHP-based microblogging service similar (although vastly superior) to Twitter. As you might imagine, those simple 140-character notices you spend way too much of your day posting go through a lot of transformations before reaching the final form in which they are displayed, because the notices need to be parsed and manipulated to add things like links to tags, links to @ replies, shortening URLs, and so on. Honestly when I started studying the StatusNet code and saw what all goes on behind the scenes for such a seemingly simple service, I have to admit I was a bit surprised.

All the text manipulation of course involves a lot of regular expressions, and since for ColdTonica we're porting the PHP code over to CFML, it saves us a ton of time since all the regular expressions have already been written. Unfortunately there are some syntax issues with the regular expressions that render them incompatible with CFML and even Java, so it did take a bit of research and help from a friend of mine to start to unravel and convert them.

One of the issues I ran into while moving the PHP regular expressions over to CFML is that CFML doesn't have Unicode support in regular expressions (some nice info about Unicode in regex here, although the information about Java is dated), at least not without first converting Unicode to ASCII values and wrapping them all in Chr(). This is what I've discovered while messing with this at least; if this isn't correct I'm happy to be proven wrong.

Since the PHP regular expressions use the Perl syntax of x{hex_value_here}, which CFML doesn't support, converting the regular expressions was getting a bit messy. Java, however, does support the x syntax (though it didn't used to), but with a slightly different syntax. You can read more about Java regex syntax in Java 6 here.

During the course of this I was reminded of the fact that under the hood, CFML strings are Java strings, which means that rather than using functions like REReplaceNoCase() in CFML and converting the hex codes into something usable by Chr(), I can simply use Java's replaceAll() function on the String class. This lets me keep the PHP syntax more intact and do a lot less conversion research.

So the original PHP looks like this:

$r = preg_replace('/[x{0}-x{8}x{b}-x{c}x{e}-x{19}]/', '', $r);

And the CFML version using replaceAll() on the String class looks like this:

r.replaceAll("/[x00-x08x0B-x0Cx0E-x19]/", "");

At least I think that's right. ;-) I still need to test all of this out, but as I convert the rest of these it'll be much simpler to go this route and keep things in hex as opposed to converting everything to CFML-compatible Unicode regex syntax.

The moral of the story is you can do a lot in CFML by leveraging the underlying Java functionality, and this doesn't apply only to the String class. So if you run into things that are a bit weird to try and accomplish in CFML check the Java docs and see what additional functionality you have available. You'll probably be surprised at what you learn!

No comments: