Replaced 'alias' calls with an instance variable.

[fimmtiu]

This fixes two problems with batch:

1. Calling alias blows out the Ruby method cache for the current class and all its subclasses, so for performance reasons you don't want to be calling alias on a regular basis.

2. If an exception occurred while sending the batched data, it would get stuck in batched mode.

Also fixed up a few typos in the specs for fun.

[degemer]

Hi @fimmtiu, thanks for the contribution!

I didn't know about 1, do you know in this particular case what is the impact?
Thanks for fixing 2 and the typos. 🙂

My main concern is sacrificing the performance of the non-batched send_stats over the batch one.
We don't have a performance test right now, I think we'll need one in order to decide. I unfortunately won't have time to work on this in the next few weeks, if you want (and have time!) to do it before, I'll be glad to review and talk about it, otherwise it will have to wait.

Thanks again for raising this issue!

[fimmtiu]

@degemer Good suggestion about the performance test! I've written a quick benchmark script, and the results seem to indicate that this PR is about the same for most cases, and faster in the case of small batches.

On master:

dennis@concertina ~/dogstatsd-ruby$ ./performance.rb
                       user     system      total        real
regular           18.860000  19.870000  38.730000 ( 38.795239)
batched single    31.590000  23.100000  54.690000 ( 54.854231)
batched 1k         9.020000   0.520000   9.540000 (  9.551664)
batched 10k        9.070000   0.510000   9.580000 (  9.588308)

On the dont-alias-all-the-things branch:

dennis@concertina ~/dogstatsd-ruby$ ./performance.rb
                       user     system      total        real
regular           18.990000  19.410000  38.400000 ( 38.532748)
batched single    24.100000  20.680000  44.780000 ( 44.883200)
batched 1k         9.290000   0.550000   9.840000 (  9.872589)
batched 10k        9.290000   0.500000   9.790000 (  9.806697)

Here's the script I was testing with:

#!/usr/bin/env ruby

require 'benchmark'
require 'rubygems'
$: << './lib'
require 'datadog/statsd'

statsd = Datadog::Statsd.new('localhost', 8125)

n = 5_000_000
Benchmark.bm(16) do |x|
  x.report("regular") do
    n.times { statsd.increment('woo') }
  end
  x.report("batched single") do
    n.times do
      statsd.batch { |s| s.increment('woo') }
    end
  end
  x.report("batched 1k") do
    (n / 1_000).times do
      statsd.batch do |s|
        1_000.times { s.increment('woo') }
      end
    end
  end
  x.report("batched 10k") do
    (n / 10_000).times do
      statsd.batch do |s|
        10_000.times { s.increment('woo') }
      end
    end
  end
end

[degemer]

Thanks a lot for the benchmark! I re-ran it locally and obtained the same results (small improvements for small batches, equivalent everywhere else).
Moreover the code is way easier to understand this way.
Merging, this will be part of the next release (targeting end of next week).