verify input string is valid decimal number

Question

Can anybody provide an algorithm that checks whether an input string is a decimal number in correct shape and form?

Rules of correct shape and form:

1. At most two decimal places.
2. For all practical purposes, the largest number is 99,999,999.99
3. The integral part can use a space, or a comma, or a dot for a group separator.
4. The decimal part can use a comma, or a dot for a separator.

Examples of correct shape and form:

1,234,567.89   // English style
1.234.567,89   // French style
1 234 567,89   // German style
1234567.89     // English mathematical style
1234567,89     // European mathematical style
12.00
12.0
12

Please, resist your temptation to propose Decimal.Parse or Decimal.TryParse. Either method merrily accepts strings such as “1,2,3,4”, which is not what I want.

Answer 1

I answered a very similar question on Friday, and the regex to do this is going to be more complicated than you think. I strongly suggest setting up a separate regex for each style - not because it can't be done in one line, but because a one-liner for this is going to be big and pretty tough to maintain.

Here are the patterns I'd use:

English: ^[-+]?\d{1,3}(,\d{3})*(\.\d+)?$
French: ^[-+]?\d{1,3}(\.\d{3})*(,\d+)?$
German: ^[-+]?\d{1,3}(\s\d{3})*(,\d+)?$
English mathematical: ^[-+]?\d+(\.\d+)?$
European mathematical: ^[-+]?\d+(,\d+)?$

These can be combined into something like:

^[-+]?(\d{1,3}((,\d{3})*(\.\d+)?|([.\s]\d{3})*(,\d+)?)|\d+([,\.]\d+)?)$

Hopefully that one-liner is appropriately terrifying to you. This is a very common, important task, but it's also one that's more complicated than it appears.

Tested at Rubular with your example inputs: http://rubular.com/r/Dipvyrf6C8 (note that I made the groups non-capturing for clarity of the results).

Edit: Added [-+]? clause to allow for negative numbers.