Question

I have this expression for valid number only…

^[0-9,-\.]+$

This works fine, but I need to ensure that a maximum of 1 , and 1 . can be entered? This needs to cater for different cultures so it doesn’t matter to me which order the , and . come in.

Can anyone help with this?

Answer 1

If your regex flavor supports lookaheads you can use:

^(?=[^,],[^,]$)(?=[^.]\.[^.]$)[\d,.-]+$

If not, then you could do something like this:

^(?:[\d-]*,?[\d-]*\.?[\d-]*|[\d-]*\.[\d-]*,[\d-]*)$

But I find this is rather ugly.

Note that I changed your character class. [0-9] can be shortened to \d (note that dependent on your regex flavor this might match other Unicode digits than just 0-9), . does not need to be escaped inside character classes, and - belong at the beginning or the end of character class, otherwise they need to be escaped. Funnily enough, in your case it does not even make a difference, because ,-. creates a range of characters from , to . (in ASCII order). However all the characters in this range are incidentally ,, - and .. Lucky you 😉

However, both answers really just do what you are explicitly asking for. Your pattern will still match --.--34,---1--2 and stuff like that. If you really want to match a valid number with culture-independent decimal point you should probably use this instead:

^-?\d*([.,]\d+)?$

Of if you want to match a valid number with . decimal point and at most one , thousand-separator (or vice-versa), this:

^-?\d{0,3}(?:(?:,\d{3})?(?:\.\d+)?|(?:\.\d{3})?(?:\,\d+)?)$

I don’t think there is a shorter version, really. Do not worry about all the ?:. They just make the subpatterns non-capturing. Since you do not need that anyway, it is usually a significant performance increase.