eladaus / url-detector Goto Github PK

Adapt existing string-based logic to use Spans and implement performance comparisons.

Possible useful links:

https://github.com/dotnet/corefx/blob/master/src/Common/src/CoreLib/System/Text/ValueStringBuilder.cs

https://github.com/discostu105/stringbuilder-tests/blob/master/StringBuilderTests/Program.cs#L105

Should use StringBuilder.ToString(start, length) overload, to reduce string allocations and theoretically cpu usage

Have found a bug in the parsing logic whereby the UrlDetector.DeadDefault logic gets stuck in an infinite loop and will crash your application in such simple cases as where a string contains a colon followed by an unexpected character combination such that it tests as NOT a username:password pair.

To repro, run this test:

    [Fact]
    public void TestDalesKillerString3() {
        var sourceText = " :u "; // kills loop in UrlDetector.ReadDefault()
        var flags = UrlDetectorOptions.AllowSingleLevelDomain;
        var links = new UrlDetector(sourceText, flags)
            // BEWARE THE INFINITE LOOP THAT HAPPENS HERE:
            .Detect();
        Assert.Equal(0, links.Count);
    }