From: Niklas Frykholm <r2d2@...>
Date: 2002-07-01T01:48:42+09:00
Subject: Re: select loop question

[Joseph McDonald]:
> NF> Just do
> NF>         answer[0].each {|r| readfds.delete(r)}
> 
> Thanks,  I do notice that Array.delete seems to be very slow, If I run this:
> 
> hits = (200..300).to_a
> 1.upto(1000) do
> 
>   arr = []
>   1.upto(500) do |x|
>     arr.push(x)
>   end
> 
>   # hits.each do |h|
>   #  arr.delete(h)
>   # end
> 
> end
> 
> I get:
>      0.47 real         0.46 user         0.00 sys
> If I uncomment out the delete portion I get:
>       8.74 real         8.23 user         0.00 sys
>       
> My guess is that it is doing the cartesion product thing, interesting
> that it is 8 times slower than:
> 
>   arr.each do |x|
>     if x >= 200 and x <= 300
>       arr.slice!(x)
>     end
>   end
> 
> which gives:  1.09 real         1.06 user         0.00 sys

If you do many deletes, it is slow. It has to search the array for
the element to delete and then move all the data that follows it.

Note that slice! is not the same thing, when you do slice!(x) you remove
the element at position x, not the element with content x. So the
array does not have to be searched.

If you want to do fast deletes based on content, you can use a hash to store
the data. Store the data in the key of the hash and a dummy value such as
"true" in the value.  You get fast lookups and there is no need to move
elements.

// Niklas