Skip to content

Instantly share code, notes, and snippets.

@paneq
Last active Dec 11, 2016
Embed
What would you like to do?
cost of using exceptions for control flow compared to one SQL statement (ruby 2.1.4, rails 4.1.7, sqlite) for rails-refactoring.com . Development mode executed in rails console.
require 'benchmark'
ActiveRecord::Base.logger = nil
Benchmark.bmbm do |bench|
bench.report("SQL query") do
1000.times { Whatever.count }
end
bench.report("exception hit") do
1000.times do
begin
raise StandardError.new
rescue
end
end
end
bench.report("exception miss") do
1000.times do
begin
raise StandardError.new if false
rescue
end
end
end
end
Rehearsal --------------------------------------------------
SQL query 0.180000 0.010000 0.190000 ( 0.199253)
exception hit 0.050000 0.000000 0.050000 ( 0.050654)
exception miss 0.000000 0.000000 0.000000 ( 0.000050)
----------------------------------------- total: 0.240000sec
user system total real
SQL query 0.180000 0.010000 0.190000 ( 0.179669)
exception hit 0.050000 0.000000 0.050000 ( 0.048447)
exception miss 0.000000 0.000000 0.000000 ( 0.000054)
@chastell
Copy link

chastell commented Nov 6, 2016

Meanwhile @andrzejkrzywda duly pointed out that comparing exception handling time with writes (rather than – especially easily cacheable – reads) would be much more telling, so here’s the comparison with Whatever.create(text: 'meh') instead of Whatever.count:

ruby 2.3.1p112 (2016-04-26 revision 54768) [x86_64-linux]
Warming up --------------------------------------
           SQL query    18.000  i/100ms
       exception hit    57.640k i/100ms
      exception miss   269.546k i/100ms
Calculating -------------------------------------
           SQL query    182.124  (±15.9%) i/s -    882.000  in   5.003193s
       exception hit    808.613k (± 4.8%) i/s -      4.035M in   5.004163s
      exception miss     10.129M (± 3.2%) i/s -     50.675M in   5.007747s

Comparison:
      exception miss: 10129279.5 i/s
       exception hit:   808613.1 i/s - 12.53x slower
           SQL query:      182.1 i/s - 55617.43x slower
ruby 2.4.0preview2 (2016-09-09 trunk 56129) [x86_64-linux]
Warming up --------------------------------------
           SQL query    20.000  i/100ms
       exception hit    58.768k i/100ms
      exception miss   344.645k i/100ms
Calculating -------------------------------------
           SQL query    186.218  (±20.4%) i/s -    880.000  in   5.005586s
       exception hit    818.088k (± 6.9%) i/s -      4.114M in   5.069060s
      exception miss     11.347M (± 0.6%) i/s -     56.866M in   5.011838s

Comparison:
      exception miss: 11346771.3 i/s
       exception hit:   818087.5 i/s - 13.87x  slower
           SQL query:      186.2 i/s - 60932.79x  slower

@dpneumo
Copy link

dpneumo commented Nov 15, 2016

For me the take home is that when a (SQL) database is involved efforts to optimize the queries has a greater chance of improving app performance than futzing with flow control techniques. Not a surprise but it's nice to have some real numbers to point to. Thanks.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment