Can my Regex be improved?

Question

Yes, another Regex question. You're welcome ;-P

This is the first time I've written my own regex for some simple string validation in C#. I think I've got it working but as a learning exercise I was wondering if it could be improved and whether I have made any mistakes.

The strings will all look something like this:

T20160307.0001

Rules:

• Begin with the letter T.
• Date in the format YYYYMMDD.
• A full stop.
• Last 4 characters are always numeric. There should be exactly 4.

Here is my regex (fiddle):

^(?i)[T]20[0-9]{2}[0-1][0-9][0-3][0-9].\d{4}$

• ^ Assert the start of the string.
• (?i)[T] Check that we have a letter T, case insensitive.
• 20 YYYY begins with 20 (I'll be dead by 2100 so I don't care about anything further :-P)
• [0-9]{2} Any number between 0 and 99 for second part of YYYY.
• [0-1][0-9] 0 or 1 for first part of month, 0-9 for second part of month.
• [0-3][0-9] 0-3 for first part of day, 0-9 for second part of day.
• . Full stop.
• \d{4} 4 numerical characters.
• $ Assert end of string.

One pitfall I can already see is date validation. 20161935 (the 35th day of the 19th month) is considered valid. I've read some / other / posts about achieving this which I believe match on number ranges but I was unable to understand the format.

I would accept an answer that simply solved the date issue if someone would be kind enough to ELI5 how this works, but other improvements would be a welcome bonus.

Edit: To avoid further confusion I should state that I know about DateTime.TryParse etc. As mentioned I'm using this as an opportunity to learn Regex and felt this was a good starting point. Sorry to anyone who's time I wasted, I should have made this clear in the original post.

Answer 1

The things you can do are:

• avoid the \d character class that matches all the unicode digits (since you only need the ascii digits)
• instead of [0-1] you can write [01]
• escape the dot to figure a literal dot (and not any characters)
• no need to put T in a character class if it is the only character
• eventually you can remove the inline modifier and use [Tt] in place of T

^(?i)T20[0-9]{2}[01][0-9][0-3][0-9]\.[0-9]{4}$

or

^[Tt]20[0-9]{2}[01][0-9][0-3][0-9]\.[0-9]{4}$

Other thing: do you really need to add extra checking for the date since you can't really test if the date is well formatted? (Think a minute about leap years) So why not:

^(?i)T(20[0-9]{6})\.[0-9]{4}$

and if you want to know if the date really exists, capture it and test it with DateTime.TryParse method.

Answer 2

Why even use Regex just use the DateTime.TryParseExact Method. I'd implement it like so with extra checks for the other characters:

bool IsCorrectFormat(string input)
{
    //14 is a magic number for the length of the expected format
    if (input.Length == 14 && input.StartsWith("T", StringComparison.OrdinalIgnoreCase))
    {
        DateTime dt;
        if (DateTime.TryParseExact(input.Substring(1), "yyyyMMdd.ffff", CultureInfo.InvariantCulture, DateTimeStyles.None, out dt))
        {
            return true;
        }
    }

    return false;
}

I don't know if the format is correct but you could always substring from 1 to 6 to get the yyyyMMdd then check for a decimal point and numerics at the last 5 characters.

EDIT: If this must be done with regex

I have used this regex in the past. Note that it does not check for leap years

@"^(((0[1-9]{1}|[1-2][0-9]{1}|3[01]{1})(0[13578]{1}|1[12]{1}))" //For a 31 day month
+ @"|"
+ @"((0[1-9]{1}|[1-2][0-9]{1}|30)(0[469]{1}|11))" //For a 30 day month
+ @"|"
+ @"((0[1-9]{1}|1[0-9]{1}|2[0-8]{1})(02)))" //For a 28 day month(feb)
+ @"([0-9]{4})$"; //For the year

Answer 3

As mentioned I'm using this as an opportunity to learn Regex and felt this was a good starting point.

It's certainly not trivial to validate a date using a regular expression, particularly given the complex rules involved for leap years. But it is possible.

The below expression will match if a valid date is input in YYYYMMdd format:

(?=\p{IsBasicLatin}{8}) # ensures \d matches only 0-9
(?!0000)\d{4} # year any 4-digit year, except 00
(?:0[1-9]\d|1[012]) # month 01-12
(?: 
   # day logic for leap years
   (?:
      (!00)[012]\d # Days 01-29 (we exclude 2/29 later)
      | (?<!02)30  # Day 30 valid for all months except Feb
      | (?<=0[13578]|1[02])31 # Day 31 valid for some months
   )
   # Non-Leap-year logic.  Do not allow 2/29 if the provided year
   # is not a leap year.
   (?<!
     (?:
        [13579] # years ending with odd number are not leap years
        | [02468][26]|[13579][048] # years not divisible by 4
                                     # are not leap years (02, 06, 10, ...)
        | (?:[02468][\d-[048]]|[13579][\d-[26]])00 # years divisible by
                                                 # 100 are not leap years,
                                                 # unless divisible by 400

     )0229)
)

Compile with RegexOptions.IgnorePatternWhitespace. You can use ^T~\.\d{4}$ to match the full string in your example, replacing ~ with the above expression.