email validator with empty string slow?

[okeydoke]

Hey, I've been writing some unit tests for my node.js project and one of the tests is showing as slow quite often, 40-100ms. Basically its validating an empty string against an email. eg:

validate({email:""},  {
  email: {
    presence: true,
    email: true,
    length: {
      minimum: 6,
      maximum: 255
    }
  }
});

Running in the browser it seems to be about 15-20ms which isn't as bad but still seems pretty slow?

[ansman]

Test code

// This line is node specific
var validate = require("validate.js");
console.time("validate");
validate({email:""},  {
  email: {
    presence: true,
    email: true,
    length: {
      minimum: 6,
      maximum: 255
    }
  }
});
console.timeEnd("validate");

Results

           | Chrome  | Node.js |
-----------+---------+---------+
    Run #1 | 7.328ms |     2ms |
    Run #2 | 0.246ms |     0ms |
    Run #3 | 0.094ms |     0ms |
    Run #4 | 0.085ms |     0ms |
    Run #5 | 0.109ms |     0ms |
    Run #6 | 0.103ms |     1ms |
    Run #7 | 0.242ms |     0ms |
    Run #8 | 0.125ms |     0ms |
    Run #9 | 0.159ms |     0ms |
   Run #10 | 0.159ms |     0ms |
-----------+---------+---------+
      Mean | 0.865ms |     0ms |
 Std. dev. | 2.272ms |     1ms |

Conclusion

Measuring performance is tricky.
Chrome has lots of JIT compilation which speeds up subsequent calls.
Node seems to be consistently fast.

How did you perform your tests?

[okeydoke]

Sorry didn't get round to this before but heres a jsperf test http://jsperf.com/js-validate-empty-string

Noticeably slower across the 3 browsers I tested. I would of though the performance would be comparable to null/undefined

I think the very slow times I was seeing before was due to a background process generally slowing down my laptop as I'm getting results more in the 3-5ms time now when doing an initial test (subsequent tests end up being cached/optimized)

[ansman]

This particular problem was fixed in d9944f4 but the problem probably comes from the RegExp that parses the email.

It is bound to be pretty slow. Is this causing any actual problems?

[okeydoke]

Yeah I guess its not a huge deal, but eking out a little more performance would be nice. But just had a quick look at the source and it doesn't look as easy as something like email === "" return false though

[ansman]

If you rerun the test against the 0.4.0 source it should be faster for empty strings

On Mon, Feb 9, 2015 at 12:10 AM, Jesse notifications@github.com wrote:

Closed #25.

Reply to this email directly or view it on GitHub:
#25 (comment)