Remove define_singleton_method in favor of opening multiple eigen classes for 1.8.7 support

[andrewGarson]

This is used to create "enum" classes so that we can enforce a bit of sanity around all the magic numbers that zabbix uses in the api. I would rather not have the lwrps accepting Fixnums for things like "type" because then a person trying to use them basically has to have the api docs open instead of just having a working knowledge of zabbix.

@laradji I can "fix" this to work on 1.8, but it has a bit of strangeness to it. Basically, in order to add the methods without define_singleton_method i have to use the following trick:

eigen_class = class << self; self; end
eigen_class.send(:define_method .....)

instead of

self.send(:define_singleton_method)

The difference is that each call to the enum method (1 per value in the enumeration) will open a new eigen class and put one method into it

class Chicken
  include Enumeration
  enum :tuna, 0
  enum :laser_beams, 1
end

self#define_singleton_method will produce [ Chicken ] < [ Chicken' ] < [ Object ] with #tuna and #laser_beams eigen_class#define_method will produce [ Chicken ] < [ Chicken' ] < [ Chicken'' ] < [ Object ] with one method in each ' class

This should not cause any functional difference (except a slightly slower method resolution) but could cause strangeness with stacktraces or debugging.

[andrewGarson]

Should fix #76

[guilhem]

Good for me, I was also working on something like this:

self.class.instance_eval do
  define_method(name) { @enumeration_values[name] }
end

But your solution seems better.

Can you add a comment in code (for futur contributor to know that this is for 1.8 compatibility) and you miss an update in metadata.rb ;)

[andrewGarson]

@guilhem i'm not 100% certain that code works. I believe it would pollute the Class class (not 100% certain without testing).

Just so everybody is aware, the reason I went with this complex enum stuff instead of just using Constants (as would be normal) is that I am generating recipes off of xml template dumps and needed to maintain the name<->fixnum mapping in memory.

I'll add that comment.

[guilhem]

@andrewGarson no more error on 1.8 for me :)

[guilhem]

can you remove the version change?

[andrewGarson]

oops

the orchestrator we're building required me to bump the version to test quickly. Fixed

[guilhem]

It's good enough for me to merge. @laradji do you want to do this? :)

[laradji]

let's go.