Question

I have a custom regular expression which I use to detect whole numbers, fractions and floats.

var regEx = new RegExp("^((^[1-9]|(0\.)|(\.))([0-9]+)?((\s|\.)[0-9]+(/[0-9])?)?)$");
var quantity = 'd';
var matched = quantity.match(regEx);
alert(matched);

​
(The code is also found here: http://jsfiddle.net/aNb3L/ .)

The problem is that for a single letter it matches, and I can’t figure out why. But for more letters it fails(which is good).

Disclaimer: I am new to regular expressions, although in http://gskinner.com/RegExr/ it doesn’t match a single letter

Answer 1

It’s easier to use straight regular expression syntax:

var regEx = /^((^[1-9]|(0\.)|(\.))([0-9]+)?((\s|\.)[0-9]+(\/[0-9])?)?)$/;

When you use the RegExp constructor, you have to double-up on the backslashes. As it is, your code only has single backslashes, so the \. subexpressions are being treated as . — and that’s how single non-digit characters are slipping through.

Thus yours would also work this way:

var regEx = new RegExp("^((^[1-9]|(0\\.)|(\\.))([0-9]+)?((\\s|\\.)[0-9]+(/[0-9])?)?)$");

This happens because the string syntax also uses backslash as a quoting mechanism. When your regular expression is first parsed as a string constant, those backslashes are stripped out if you don’t double them. When the string is then passed to the regular expression parser, they’re gone.

The only time you really need to use the RegExp constructor is when you’re building up the regular expression dynamically or when it’s delivered to your code via JSON or something.