Use Typhoeus to help forked/threaded users

[brucek]

Fixes #241.

This pull introduces/changes:

• Using Typhoeus gem for http connections
• Removed net-http-persistent

Pings:
@cheerfulstoic
@subvertallchris

[brucek]

Not sure if you guys want to do this, but I get much better stability in my resque-pool multi-worker application using Typhoeus. See my comments in drbrain/net-http-persistent#37

[brucek]

Also I'm using this on the 6.1.x branch so I can only give data from there, but I ported it over to 7.0.x and it looks to be working.

[coveralls]

Coverage increased (+0.002%) to 92.363% when pulling fdbf94a on brucek:typhoeus_7.0.x into 7abe23c on neo4jrb:master.

[subvertallchris]

I seem to remember benchmarking the two adapters and finding
NetHttpPersistent much faster when using SSL. We should probably check that
out again using a remote server if we're going to switch the default;
alternatively, what about making it configurable, let the user find the one
that's best for them?

On Wednesday, November 16, 2016, Coveralls <[email protected]
javascript:_e(%7B%7D,'cvml','[email protected]');> wrote:

[image: Coverage Status] https://coveralls.io/builds/8852408

Coverage increased (+0.002%) to 92.363% when pulling fdbf94a
fdbf94a
on brucek:typhoeus_7.0.x into 7abe23c
7abe23c
on neo4jrb:master.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#272 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AD6E9xNDo-tQ8Zj1x5QicDDsuuXx-muuks5q-vpqgaJpZM4KzzTI
.

[brucek]

So I did some quick benchmarking on OSX / Ruby 2.3.1 and it looks like Typhoeus is basically as fast or faster (at least on a normal https:// web connection).

I didn't spend a lot of time optimizing, so it's very possible I left something out so definitely have a look.

The configurable approach is probably ideal, but it might be a little bit before I can look at that, however.

[cheerfulstoic]

I'm all for Typhoeus if it's at least as fast.

As for configuration, I wonder is it possible to configure the adaptor via the init_params (passed in as the :initialize key to the last argument of open which I think is also configurable in the railtie)

[brucek]

Closing in favor of #273