Create method to Validator that accepts data argument.

[glennpratt]

Currently the Validator class requires data be passed into the constructor, though it doesn't do anything important with it then.

In my case, I'm validating in a server and I see no reason to parse the schema over and over for different inputs, especially since it uses two mutexes in the initialize method.

Let me know if I missed something and Validator simply must accept data at construction.

Is #validate_data(data) the way to go, or should I change the signature of #validate to take an optional argument? The former seemed like a clean change with little API change.

[glennpratt]

I wrote this little script to look for any obvious threading issues.

https://gist.github.com/glennpratt/bf254a896c9b0f15b812

No errors on 2.1.6 and JRuby.

[RST-J]

To me this looks fine. 👍

[glennpratt]

This needs work, have issues with shared @errors instance var in testing.

[glennpratt]

OK, pushed a commit that should fix that issue.

Need more tests around that apparently.

[glennpratt]

Note: This PR is sparse to help decide if this is a good direction to move it.

If it is, I'll update usage, documentation and tests.

So, do people think this is the right direction?

(Thanks @RST-J)

[jpmckinney]

Yes, this is the right direction. Re-initializing a schema everytime you validate a record is extremely slow. The schema should be initialized once, and then used to validate any number of records.

[iainbeeston]

Yes I agree with the direction completely. I would actually change the signature of the validate method. We already have a 3.x branch for exactly this kind of backwards incompatible change.

Once you're happy with it could open a new pull request on the 3.x branch?