Message Title

Actually I am getting skeptical about the performance gain claim. I wrote a little performance test using JMH - url-constraint-perf and the validation via the URL constructor comes out ahead. Using the single shot time mode, I get for example:

# Run complete. Total time: 00:32:25

Benchmark                                                                               (url)    Mode  Samples  Score   Error  Units
o.h.v.URLConstraintBenchmark.measureValidationUsingRegExp            http://foo.com/blah_blah  sample  1703711  7.290 ? 0.028  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingRegExp             http://142.42.1.1:8080/  sample  1318943  5.262 ? 0.032  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingRegExp                http://??????.??????  sample  1340940  2.626 ? 0.026  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingRegExp                          http//foo/  sample  1425989  0.362 ? 0.004  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingRegExp                                ///a  sample  1421783  0.204 ? 0.005  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingRegExp                     :// should fail  sample  1420884  0.206 ? 0.007  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingURLConstructor    http://foo.com/blah_blah  sample  1552078  0.341 ? 0.007  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingURLConstructor     http://142.42.1.1:8080/  sample  1245500  0.416 ? 0.012  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingURLConstructor        http://??????.??????  sample  1805705  0.289 ? 0.003  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingURLConstructor                  http//foo/  sample  1675968  2.135 ? 0.009  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingURLConstructor                        ///a  sample  1613901  2.104 ? 0.009  us/op
o.h.v.URLConstraintBenchmark.measureValidationUsingURLConstructor             :// should fail  sample  1734023  2.057 ? 0.006  us/op

Note, that in this particular test regexp seems to e better for invalid urls, whereas URL constructor for valid ones. My guess is that building the throwing the exception for invalid URLs might take some time here. I also checked the URL implementation, the lookup/loading of the handler might take some time, but afaiks it is cached (which one would expect anyways). Even if I change the harness parameters, URL constructor validation seems to come out ahead.

Btw, I noticed some differences in what the regexp vs the URL constructor "consider" valid. I also looked at https://mathiasbynens.be/demo/url-regex and noticed that many of the URLs listed there as invalid would be accepted by the suggested regexp as well as the URL constructor. This is just showing what a mine field this whole url validation is.