Regex bug with line endings handling

[RusKnyaz]

Execute the code:

var rheaders = /^(.*?):[ \t]*([^\r\n]*)$/mg;
var headersString = 'X-AspNetMvc-Version: 4.0\r\nX-Powered-By: ASP.NET\r\n\r\n';
var arr = []
while ( match = rheaders.exec( headersString ) ) { 
arr.push(match[1].toLowerCase());
arr.push(match[ 2 ]);
}

Expected: arr is array of ["x-aspnetmvc-version", "4.0", "x-powered-by", "ASP.NET"] Observed: arr is empty.

[paulbartrum]

Looks like the issue is that in .NET the $ character matches \n (in multiline mode):

new Regex("^.*$", RegexOptions.Multiline).Matches("one\r\ntwo\r\n")[0].Value
// returns "one\r"

Whereas in javascript it matches \r or \n:

'one\r\ntwo\r\n'.match(/^.*$/m)[0]
// returns "one"

[paulbartrum]

The workaround is to change your regular expression:

var rheaders = /^(.*?):[ \t]*([^\r\n]*)\r?$/mg;

[RusKnyaz]

Unfortunately I cannot use a workaround because this code is from jquery. And there are probably tons of code on web pages that use regular expressions. Please look at the similar issue in jint

[Taritsyn]

@RusKnyaz For implementation of regular expressions in the Jurassic, Jint and NiL.JS engines are used a System.Text.RegularExpressions.Regex class, which is not fully compatible with ECMAScript (see the “Regular expression parsing error” issue).

[RusKnyaz]

@Taritsyn I kown it. And It is possible to fix. Please look the link I provided in previous comment.

[paulbartrum]

I've checked in a fix, let me know if it works for you :-)

[paulbartrum]

@RusKnyaz You'll notice that my fix is more complicated than the Jint one. It seems the fix in Jint is not correct. 'one\r\ntwo'.match(/^.*$/mg).toString() should return "one,,two" but in Jint it returns "one\r,two".

[kpreisser]

@RusKnyaz You'll notice that my fix is more complicated than the Jint one. It seems the fix in Jint is not correct. 'one\r\ntwo'.match(/^.*$/mg).toString() should return "one,,two" but in Jint it returns "one\r,two".

It seems the fix in Jurassic is also not 100% correct :wink: E.g. ('one\\\r'.match(/^.*\\$/mg) || []).toString() should return "one\\" but returns "".

[paulbartrum]

Ha, you're right, darn it.

[paulbartrum]

I checked in a fix for the escaping issue.