add require statements to top of file

[shicholas]

Based on
#1170 (comment)

[bf4]

I'd rather use the full require active_model/serializer/configuration without further discussion

[NullVoxPopuli]

👍 to full require

[shicholas]

np I'll update this.

EDIT: updated.

[bf4]

Awesome, thanks!

[NullVoxPopuli]

Is there any sort of test you can add for this @shicholas, so your #1170 doesn't happen in the future?

[shicholas]

good question, I'm kind of stuck but I can't think of a test that won't introduce a ton of test dependencies (e.g. a test/dummy rails app that is present in rails engines). Any thoughts? I'm happy to implement it.

[shicholas]

perhaps we can get rid of all autoload like @bf4 mentioned here: http://www.benjaminfleischer.com/2013/07/18/ruby-requires-confusion/.

[NullVoxPopuli]

perhaps. I could get on board with that, even if there is a small performance and memory penalty in certain cases.
@beauby, @joaomdmoura thoughts on that?

[bf4]

perhaps we can get rid of all autoload like @bf4 mentioned here: http://www.benjaminfleischer.com/2013/07/18/ruby-requires-confusion/.

Nice, link! I should note that Rubies 2.0+ are threadsafe with autoload ref: headius/thread_safe#15 (comment)

Here's some links on require vs. require_relative performance in rspec

• http://rubyforge.org/pipermail/rspec-users/2011-November/020760.html
• Add support for optimized requires. rspec/rspec-support#45
• Benchmark require relative and update to new rspec-support method rspec/rspec-core#1348
• Use require_relative inplace of standard require if available. rspec/rspec-core#502

[bf4]

I'm 👍 on getting rid of the autoload, but I'm okay with lazy loading

[jfelchner]

I would ❤️ getting rid of the autoloads. Also require_relative gives almost no performance improvement to require so I think we should just go with require since require_relative has its own problems.

[shicholas]

I pushed an update to use all require statements. Thanks for the links, @bf4, I tried playing around w/ some autoload stuff to no avail (could be user error :)).

[NullVoxPopuli]

Well, I'd say if we are going to do away with autoload, and since the issue was with autoload, and that coming up up with a proper test would be really weird / tricky / hacky, I think this in good to go as in. Good work!

[shicholas]

I'm having a tough time deciphering the build error messages; it looks like simplecov?

[NullVoxPopuli]

Our appveyer build breaks 80% of the time. We're working on it. Lol

[shicholas]

sorry, I realized that the test failures were my fault b/c I only removed the autoload statements in lib/active_model/serializer.rb and not elsewhere. This is now updated and should now pass.

[bf4]

Note of caution, we're not actually using Kernel.autoload but ActiveSupport::Autoload, which cannot co-exist with require. Anyhow once we're using only requires, it won't matter.

Also, requires should always be one direction, never circular

[NullVoxPopuli]

👍