PHP in Action

Readable regular expressions

dagfinn | 18 March, 2007 21:41

It bothers me: I can dream up more ideas in one afternoon than I can write down in a week. So now I've decided to write down as much of today's ideas as I can before it gets too late in the evening. It might be only a fraction of what I've been thinking, but I guess I'll have to learn to live with that.

Anyway, when I use regular expressions (regexes), I often feel the need to make them more readable. I do all sorts of things to achieve that. Commenting them is a pretty basic way.

The x modifier allows you to insert comments into the regex itself. All whitespace is ignored, and so is text after # at the end of a line:

$regex =
'/(w+s+){6} # Word followed by spaces, repeated six times
w+ # Last word
/x';

I always thought this was wonderful, but today I finally realized it's completely unnecessary. Using concatenation to split the regex costs us a few more characters, but it's more flexible, since we can split it anywhere we want:

$regex =
'/(w+s+)'. // Word followed by spaces
'{6}'. // Repeated six times
'w+'. // Last word
'/'.

In ordinary program code, it's often useful to extract well-named methods instead of using comments. So naturally I think this way with regexes as well. It's possible to name the parts of the expression. The plainest way, just to show the principle, is using variables for example:

$word = "w+";
$spaces = "s+";
$regex = "/($word$spaces){6}$word/";

But why aren't regular expressions slightly more verbose in the first place? Wouldn't that make them more readable? I came across this blog post by Joshua Flanagan. He has constructed an object-oriented API for generating regular expressions, and there's a lot of interesting discussion in the comments.

Apart from all the points made there, there's one that no one seems to have spotted: Simple regular expressions gain readability by being similar to the strings they match. Take this expression to match a date and time in ISO format:

'dddd-dd-dd dd:dd'