Improve my regex for numbers in a string

Question

I have created a bulky .net regex expression that works, but is quite inefficient. I'd have to assume that there is a more condensed way to write it, but I can't seem to find one.

What I need it to do:

match positive and negative integers
match positive and negative decimals with or without any numbers preceding the decimal
work for .NET since it is being used in a WPF application

Examples of matches from arbitrary example strings:

"7" from "7in"
"-4" from "-4 mm"
".345" from " .345"
"-.43" from " -.43 cm"
"0.16" from "$0.16"
"-10.04" "-10.04"

What I need to avoid:

matching the second number in a range as a negative
*ex: matching "-3500" in "1000-3500 RPM" (should match 1000 and 3500)

My regex:

(?<![0-9])(-)\.\d+|(?<![0-9])(-)\d+\.\d+|(?<![0-9])(-)\d+|\.\d+|\d+\.\d+|\d+

The Breakdown:

?<![0-9])(-)\.\d+ - a negative decimal without numbers before the decimal and then make sure there is no number before the dash

OR

(?<![0-9])(-)\d+\.\d+ - a negative decimal with numbers before the decimal and then make sure there is no number before the dash

OR

?<![0-9])(-)\d+ - a negative integer without a number preceding the dash

OR

\.\d+ - find a positive decimal with no preceding numbers

OR

\d+\.\d+ - find a positive decimal with preceding numbers

OR

\d+ - find an positive integer

Now, from what I understand, this should be able to be combined into fewer "OR" segments. Id guess that you can at least combine the positive and negative cases and probably the optional numbers proceeding a decimal. Any increased performance would be much appreciated.

Thanks.

Take a look at [this post](http://stackoverflow.com/q/4246077/1679863). You don't need anything else. — Rohit Jain, Aug 02 '13 at 16:31
Is this a rang `-1--3`? negative 1 to negative 3. Can ranges have whitespace between the hyphen? — Daniel Moses, Aug 02 '13 at 16:44

score 0 · Answer 1 · answered Aug 02 '13 at 16:35

0

How about this one: -?(\d*\.)?\d+ ?

answered Aug 02 '13 at 16:35

PoByBolek

3,405
2
16
22

this will match the case he wants to avoid. – user428517 Aug 02 '13 at 16:37
Close, but it doesn't solve the issue with the range situation. Thanks – Scott P Aug 02 '13 at 16:38
1

It looks like adding the lookback to the optional dash will do it though ((? – Scott P Aug 02 '13 at 16:43

score 0 · Accepted Answer · answered Aug 02 '13 at 16:44

0

Try (?<!\d)\s*-?\d*(\.\d+)?

Uses a negative lookbehind to assert there are no digits before the dash, also only allows only one decimal point (if any).

answered Aug 02 '13 at 16:44

Song Gao

656
4
14

sharktamer · Answer 3 · 2013-08-02T17:05:15.343

0

Looks like this does the job: (?<![0-9])-?\.?([0-9]+)?\.?[0-9]+

Tested here

edit- Condensed slightly: (?<![0-9])-?\.?[0-9]*\.?[0-9]+

Tested here

edited Aug 02 '13 at 17:05

answered Aug 02 '13 at 16:49

sharktamer

126
8

this would allow the matching of two periods – Song Gao Aug 02 '13 at 19:42

Improve my regex for numbers in a string

3 Answers3