Extract a float

Hello,

I'm trying to extract a float from the following strings:
-10K
10K
2M
<0.2%

Basically, there can be textual information before the float and textual information after it. The regexp i'm using now is:

/(\-?[0-9]+\.?[0-9]+)[^0-9]*/

I'm not very good at regexps so i don't understand, why it does not match "-1K" string, but matches "-10K". I guess it has something to do with the "\.?[0-9]+" part. It should be one separate rule and the [0-9] should not be matched if the \. is not found. How do i make this entire part optional?

I've modified it to:

/(\-?[0-9]+\.[0-9]+|\-?[0-9]+)[^0-9]*/

Seems to work now, but not sure if it won't cause any problems.

You basically just want to pull the number bits out of the string, including a sign (if any), right? You're overcomplicating it:
-Dan

Well, almost. I also wanted to extract the part, comming after the number so i just added the brackets to the last part [^0-9], but then i realised i was probably also overcomplicating it and i could've done it with /[0-9]+([^0-9]*)$/ (or probably even without the [0-9]+ part)

Oh you want the bits afterward too. You don't even have to do [^0-9] you can just use a dot:

That puts the number (with any sign) into $1, and the rest into $2

-Dan

Thanks, that's a neat expression. But now i remembered why i went through all the trouble. Sometimes there's a weird string like "1-2-3" and this should never be interpreted as a number (shouldn't be matched). Should i use something like [0-9]\-[0-9]\-[0-9] and add an additional expression for this situation or is it possible to modify your idea? Basically the rules generally are:

* the string starts with a non-numeric character or nothing at all
* followed by minus sign or not
* followed by a number (possibly with a floating point)
* the rest can't contain any numbers or minus signs

I think an expression, based on these rules, should be able to match all the variants. But then again, i'm not that good in regexps, maybe i'm not thinking the way i should be thinking.


Update

I've modified your version of regexp to this:
/^[^0-9\-]*(-?[0-9]+(\.[0-9]+)?)([^0-9\-]*)$/

Seems to work, but i still had to change the .+? part to [^0-9\-] to make it work. If i didn't miss anything, that's probably the final solution. I'll add some more comments when i fully test it out.

Update 2

Well i've tested every possible situation that might occur (if i haven't forgotten something) and it works ;]

Thanks a lot for your help!