Significant performance regression with 2.5.1

[thardeck]

So we are using json-schema to verify our json system description.

We are using 2.2.5 by default but if 2.5.1 is installed it will be used automatically if it isn't executed with bundle exec.

These are the results with 2.2.5:
real 0m1.368s

These are the results with 2.5.1:
real 0m12.866s

Even 2.5.0 is slower but not as much:
real 0m5.006s

2.4.1 is as fast as 2.2.5:
real 0m1.551s

These two schemas should be relevant. One for the whole description and the other for the inspector.

I have uploaded a simple json file here, which I used for the above benchmarks.

I don't have time to bisect but if you have an idea what might causing it I can test it. Thx in advance.

[iainbeeston]

Would it be possible for you to try the code from #255 and see if that speeds things up? If so I'll try to get that pull request merged and released (along with #239)

[thardeck]

Thx for the quick reply.

So I have tested the branch and it is much faster then original 2.5.1 but still slower than 2.4.1:
real 0m2.832s

[iainbeeston]

It seems like the switch from ruby URI to addressable has caused this problem. I'm tempted to make it optional - #285 will make this easier

[thardeck]

Thx for looking into this but I have tried release 2.6.0 and it is still ~ four times as slow (I have updated the previous values because it was more validated than just the json):

This is with 2.2.5:

> time bundle exec bin/machinery validate tumbleweed
Validation succeeded.

real    0m2.184s
user    0m2.046s
sys     0m0.133s

This is with 2.6.0:

> time bin/machinery validate tumbleweed
Validation succeeded.

real    0m8.309s
user    0m8.182s
sys     0m0.127s

> gem list  | grep schema
json-schema (2.6.0, 2.2.5)

If you want to test any patch or pull request regarding this let me know.

[thardeck]

I have forgotten to post the example from above: http://pastebin.com/a0nJVvKx

[iainbeeston]

Thanks for sharing that, I'll try to take a look.

[thardeck]

While 2.8.0 is much better than 2.5.1 it is still quite slow in comparison to 2.2.5. These values are generated against the example posted above.

2.8.0

real    0m9.611s                                                                                                                                                                       
user    0m9.368s                                                                                                                                                                       
sys     0m0.238s

2.2.5

real    0m2.202s                                                                                                                                                                       
user    0m2.037s                                                                                                                                                                       
sys     0m0.162s

2.5.1

real    0m23.400s
user    0m23.038s
sys     0m0.347s