rspec users - Nov 2011 - Using require_relative to speed up rspec require time.

Hi,

I noticed recently that require ''rspec'' on my machine was
taking close
to half a second. That''s not a huge amount of time, but it is still
the single slowest part of my test suite.

It boils down to Ruby 1.9''s rather slow require. I''m using
1.9.3, but
I''d still like to shave off some of the require time.

As an experiment, I went into rspec-core and rspec-expectations (the
two biggest offenders) and replaced all require calls with
require_relative. The benefits are actually quite impressive:

rspec-core:
before: 0.16s
after: 0.10s

rspec-expectations:
before: 0.16s
after: 0.05s

Applying this to rspec-mocks also, my total require time for
''rspec''
has gone from 0.5s to 0.21s.

These are just quick a nasty timings, but there''s obviously some win to
be had.

Would David, or the RSpec developers in general accept a patch to use
require_relative if 1.9 is detected? Are there scenarios under which
require_relative would not be possible?

Cheers
Ian

On Nov 12, 2011, at 12:58 PM, Ian Leitch wrote:
> Hi,
> 
> I noticed recently that require ''rspec'' on my machine was
taking close
> to half a second. That''s not a huge amount of time, but it is
still
> the single slowest part of my test suite.
> 
> It boils down to Ruby 1.9''s rather slow require. I''m
using 1.9.3, but
> I''d still like to shave off some of the require time.
> 
> As an experiment, I went into rspec-core and rspec-expectations (the
> two biggest offenders) and replaced all require calls with
> require_relative. The benefits are actually quite impressive:
> 
> rspec-core:
> before: 0.16s
> after: 0.10s
> 
> rspec-expectations:
> before: 0.16s
> after: 0.05s
> 
> Applying this to rspec-mocks also, my total require time for
''rspec''
> has gone from 0.5s to 0.21s.
> 
> These are just quick a nasty timings, but there''s obviously some
win to be had.
> 
> Would David, or the RSpec developers in general accept a patch to use
> require_relative if 1.9 is detected? Are there scenarios under which
> require_relative would not be possible?
I can''t think of a reason why this would be problematic, as long as
we''re not redefining existing methods that other code might be using.

My only concern is the details of the implementation, and I''m not sure
what I''d like to see there. Would you please file a pull request with
just one of the libs (rspec-core is probably best) and once we get settled on
the implementation details we can do it for the others.

Cheers,
David