New method to post UDP events to the stream with Ruby

[adrienDog]

Documentation updated using Yard.
2 tests created in statsd_spec.
Refs: #1

Was greatly inspired by the python version.

[miketheman]

I'd like to see some more tests around the behavior added of other flags, to ensure that the addition or omission, ordering, etc behaves as expected.

[miketheman]

Overall, it looks like string is being rewritten multiple times as you flow through the code, building up a new object each time, leading to memory bloat.

[1] pry(main)> string = "foo"
=> "foo"
[2] pry(main)> string.object_id
=> 70154227837960
[3] pry(main)> string = "#{string}|bar"
=> "foo|bar"
[4] pry(main)> string.object_id
=> 70154235031700

You could use the shovel operator to append to the existing object, like so:

[1] pry(main)> string = "foo"
=> "foo"
[2] pry(main)> string.object_id
=> 70270745414420
[3] pry(main)> string << '|bar'
=> "foo|bar"
[4] pry(main)> string.object_id
=> 70270745414420

Another cleanup is for any single-line if/end statements, you can put the evaluation inline with the code, like so:

date_happened = opts[:date_happened] || nil
string << "|d:#{date_happened}" if date_happened

[miketheman]

Final note:
using string as the main variable to pass to the socket might be easily mistaken for String, which is the top-level object class.
Naming it to something like event_message or whatever in this code could prevent future mistakes, as well as declare what the purpose of the object is.

[adrienDog]

So I have made more tests, I don't know if they are sufficient.
I left it possible to change the hostname for an event, since it's available in the python file.

All modifications asked should be done I hope :)

[adrienDog]

I have created a dictionary to link parameter names to their respective 1-letter key so that in the future it's faster to add more parameters.

[miketheman]

@adrienDog This is looking better and better.

As you can see, the tests failed. :) For travis to not install the yard/redcarpet (which it doesn't need), you can exclude them in .travis.yml.

For the tests, see the failures: most of them are due to the packet size being over 8K - so consider this:
Instead of running the entire test suite through 100 iterations of potential failures due to too much randomness, have a test that confirms the correct failure mode when the size is too large, and have the other tests behave correctly when the payload size is correct.

All in all, I like the way this is going, keep it up!

[adrienDog]

Ok thanks Mike I will work on that !

[wang-arthur]

I think you could rename this file "test-statsd.rb" (http://www.deathoftheauthor.com/).

[miketheman]

I think this file was mistakenly added and committed. @adrienDog ?

[adrienDog]

Yeah @miketheman it doesn't really need to be there. But it was a way for me to really send events to my localhost.
Do you think I should leave it there and rename it ? Or just not commit it ?

[miketheman]

I think it's good to have an example handy to test, so I would clean it up, place it in an examples folder, with notes on why and when it can be used.

[adrienDog]

Ok I created the folder examples, renamed the test file in it more accurately I think.
I hope everything is fine now

[adrienDog]

Yeah! Travis is my friend again

[miketheman]

This defines an array of the same elements that are represented in the above hash.
Is there any danger in sorting the payload's opt_keys before passing to .each instead?

[clofresh]

ops_keys never changes, so it should be a constant defined once in the global scope, instead of each time this function is called. Also, you're defining a hash which is unordered, and sorting it each time, which is wasteful. You could instead just have an array of arrays and avoid resorting:

OPTS_KEYS = [
      ['date_happened', 'd'],
      ['hostname', 'h'],
      ['aggregation_key', 'k'],
      ['priority', 'p'],
      ['source_type_name', 's'],
      ['alert_type', 't']
]

[adrienDog]

@clofresh Is it better like this ?