Faster require

[odlp]

Requiring fake_stripe is slow because it boots the Stripe JS server immediately, even when FakeStripe.stub_stripe hasn't been called yet.

This PR proposes:

• Only boot Fake Stripe JS server when needed.
• Still perform the identification of an available port early, so STRIPE_JS_HOST can be populated correctly.

Impact

By deferring spinning up a Capybara server until necessary there's a significant improvement on the time to require the gem.

Running command:

repeat 10 { /usr/bin/time -p bundle exec ruby -e "require 'fake_stripe'" 2>&1 } | grep real

	Before (secs)	After (secs)
	5.85	0.71
	5.74	0.71
	5.75	0.72
	5.81	0.69
	5.81	0.7
	5.72	0.68
	5.72	0.67
	5.81	0.71
	5.79	0.7
	5.81	0.66
Average:	5.781	0.695

Motivation

When I run an non-Stripe related test I don't want to experience the performance penalty. Today one would have to use the following in a Gemfile & explicitly require fake_stripe later:

gem "fake_stripe", require: false

After this PR there's far less performance impact of a vanilla require fake_stripe in the Gemfile like:

gem "fake_stripe"

[derekprior]

👍

[samcday]

@odlp it looks like this (awesome) PR never ended up in a release anywhere. Would you mind cutting a new release to RubyGems sometime?

[odlp]

@samcday good spot - I just released v0.1.0 to Rubygems: https://rubygems.org/gems/fake_stripe/versions/0.1.0