Date Range Validation in BizTalk EDI Schema

Question

I need to validate the date range in an element in a BizTalk schema. I have received dates with start dates before end dates (20130521-20130501). I know I could easily pars

Answer 1

Rather than writing your own custom component to do the validation you could use the BizTalk Business Rules Engine Pipeline Framework in conjunction with a BRE Policy to validate the date range.

Full Disclosure: This framework is written by a colleague of mine.

Answer 2

_{Disclaimer: I'm using PHP, which means I'm using the PCRE regex flavor.}

The problem

It seems that you are unaware of the limits of regular expressions, but that's ok.
What the question really comes to is the following solution: check if x =< y and match it.

The limit

Why ? Well you want to check if start date =< end date. The idea of regular expressions is to match certain characters following a certain regular pattern. Regex alone can't check if x < y since regex doesn't have logical operators > <.

Bypassing certain limit

What regex could do is to check if x = y. Say for example I've the following string and I want to get all the lines where x = y:

10 = 10
15 = 51
33 = 31
100 = 101
780 = 780

We could use the following regex: ^(\d+)\s*=\s*\1$ with the m modifier. What does this mean ?

• ^ : start of line
• (\d+) : group and match any digit one or more times
• \s*=\s* : match white space 0 or more times, and then = and then any white space 0 or more times
• \1 : referring to group 1, so match only if it's the same as what was matched in group 1.
• $ : end of line
• m modifier : multi-line. Make ^ and $ match start and end of line respectively
  Online demo.

Proof of concept

Let's hack further. For this POC, we are going to match the following: x-y where 0 =< x =< 9 and 0 =< y =< 9 and x =< y.
What we can do is trying to match all possibilites where x =< y. So if x=0 then y=[0-9], if x=1 then y=[1-9], if x=2 then y=[2-9] and so forth. Since regex has the or statement, we could write the following regex:
0-[0-9]|1-[1-9]|2-[2-9]|3-[3-9]|4-[4-9]|5-[5-9]|6-[6-9]|7-[7-9]|8-[8-9]|9-9
Online demo
You see ? It's actually so long for a simple comparison ! That's why any sane person would parse and validate it with built in language tools.

Breaking the laws of regex

We're going to use PHP to generate a regex expression:

$start = strtotime('2013-01-01'); // Start date
$end = strtotime('2013-03-01'); // End date
$range = array_map(function($v){return date('Ymd', $v);}, range($start, $end, 86400)); // Creating the range of dates
$result = ''; // Declaring an empty variable to store our regex in it

for($i=$start;$i<=$end;$i+=86400){ // loop each day
    $result .= '(?:' . date('Ymd', $i) . '-(?:'. implode('|', $range) . '))|'; // building our regex
    array_shift($range); // removing first element of range
}
$result = substr($result, 0, -1); // removing last | since we don't need it
echo $result; // Output

The above code will generate a regex that can validate a date between 2013-01-01 and 2013-03-01 where x =< y in the form of x-y. This regex is not optimised and is about 17KB. So imagine the size of this regex if I configured it to validate a range of 10 years ? Note that the size grows exponentially. I tried with an interval of 4 months but I got an error/warning saying that the expression is too long.
Since the regex is too long, I can't make an online demo of it, but here's the code in PHP:

$string = '20130101-20130101
20130201-20130101
20130105-20130120
20130201-20130301
20130210-20130215
20130301-20130301
20130301-20130201
'; // A sample

$regex = file_get_contents('regex.txt'); // Get the regex from a file (which we generated previously)
preg_match_all('#'.$regex.'#', $string, $matches); // Let's regex !
print_r($matches[0]); // Printing the matches ...

Output:

Array
(
    [0] => 20130101-20130101
    [1] => 20130105-20130120
    [2] => 20130201-20130301
    [3] => 20130210-20130215
    [4] => 20130301-20130301
)

Online dump of the regex   |   Online PHP demo

Conclusion

Please don't ever think about using regex for this task, otherwise you would have 10 problems :)