On Jul 31, 2006, at 11:17 AM, Caio Chassot wrote:
Do you really think it's a good idea to run filters more than once?
How about maybe adding a uniq option to pre/append_filter_to_chain,
that would ensure a filter is never added twice (but related
conditions may be changed)?
It would seem to me most filters that end up running twice or
thrice do so harmlessly, but needlessly. At most, hindering
performance.
I think you're right that most filters should only run once. But I
worry that doing this might cause problems in border cases. What
about putting warnings in the logs when filters are run more than
once rather than prohibiting it all together. Or maybe throw an
exception, but allow that to be suppressed by providing something like:
before_filter :foo, :rerun => true
Perhaps :rerun isn't the best name, but I think you get the idea.
You could throw the exception in before_filter using .include? and
avoid the overhead of keeping a call record or constantly making the
list unique.
-Mat
PS., I ran a quick benchmark for uniq vs. include? and found this:
Rehearsal -----------------------------------------------
uniq 1.430000 0.070000 1.500000 ( 1.531704)
if include? 0.050000 0.000000 0.050000 ( 0.042735)
-------------------------------------- total: 1.550000sec
user system total real
uniq 6.470000 0.430000 6.900000 ( 7.013875)
if include? 0.040000 0.000000 0.040000 ( 0.044006)
here's the benchmark code incase I missed anything:
--- begin uniq_vs_include.rb
require 'benchmark'
n = 5000
Benchmark.bmbm do |b|
array = []
50.times do
array << rand(50)
end
array.uniq
items= []
n.times do
items << rand(50)
end
b.report("uniq") do
n.times do |i|
item = items[i]
array << item
array.uniq
end
end
b.report("if include?") do
n.times do |i|
item = items[i]
if array.include? item
array << item
end
end
end
end
---- end uniq_vs_include.rb
_______________________________________________
Rails-core mailing list
Rails-core@lists.rubyonrails.org
http://lists.rubyonrails.org/mailman/listinfo/rails-core
