[ruby-core:42386] Re: require vs require_relative benchmarks

How many Gems do you have installed on your test system? I would guess that, unless you had 1000+ gems, the time it takes to parse the required libraries would easily swamp the lookup time for require/require_relative.

I'll save you the trouble, though. If you look in load.c you'll find that require and require_relative both call through to rb_require_safe, the only difference being that require_relative tacks the file base path in front of the file name being required. So, no, require_relative is not optimized. 


On Monday, February 6, 2012 at 8:59 PM, Trans wrote:

> I recently benchmarked require vs require_relative, and I was
> surprised to see no noticeable difference between the two. This
> doesn't make a whole lot of sense to me since relative require
> shouldn't have to look anything up via RubyGems or the $LOAD_PATH, so
> there should be some appreciable difference, even if relatively small.
> 
> So what am I missing? Is the implementation not optimized? Or is there
> something that has to happen behind the scenes that voids any load
> time gain? Or were my benchmarks anomalous for some reason?
> 
> Note: My benchmarks were simple timing how long it takes to load the
> test-unit library, which has approx 20 or so files to load, vs the
> same library with all the requires changed to relative requires. Both
> averaged around 0.075 seconds.
> 
>