From: Erik Veenstra <erikveen@...>
Date: 2007-02-15T22:35:46+09:00
Subject: Re: sorting an array of hash

>  mime_types.sort!{|mime_type1,mime_type2|
>    mime_type1['type'] <=> mime_type2['type']
>  }
>
> it's working very well ;-)

But it's inefficient... (See benchmark below.)

You're doing a lookup in both hashes in order to compare the
types. A more efficient technique is to cache theses types.

Sort_by constructs a temporary array, where each element is an
array containing the type along with the mime type. Sort_by
than sorts this array, and extracts the mime type from the
result.

gegroet,
Erik V. - http://www.erikveen.dds.nl/

----------------------------------------------------------------

 $ cat test.rb
 require "benchmark"

 array =
 (1..100_000).collect do
   {:thing => rand}
 end

 Benchmark.bmbm do |bm|
   bm.report("sort") do
     10.times do
       array.sort do |h1, h2|
         h1[:thing] <=> h2[:thing]
       end
     end
   end

   bm.report("sort_by") do
     10.times do
       array.sort_by do |h|
         h[:thing]
       end
     end
   end
 end

----------------------------------------------------------------

 $ ruby test.rb
 Rehearsal -------------------------------------------
 sort     76.360000   0.000000  76.360000 ( 76.510000)
 sort_by   7.391000   0.000000   7.391000 (  7.391000)
 --------------------------------- total: 83.751000sec

               user     system      total        real
 sort     46.967000   0.000000  46.967000 ( 47.047000)
 sort_by   7.191000   0.000000   7.191000 (  7.190000)

----------------------------------------------------------------