Class: MapRedus::Reducer

Inherits:
QueueProcess show all
Defined in:
lib/mapredus/reducer.rb

Overview

Reduce is a function that takes in “all” the values for a single given key and outputs a list of values or a single value that usually “reduces” the initial given value set.

The output of the reduce shall always be

reduce(values) = [ reduced value, reduced value, ... ]

and it will often only be a single element array

The input values and the output values of the reduce will always be a string. As described in the paper, it is up to the client to define how to deal with this restriction.

Direct Known Subclasses

Adder

Constant Summary collapse

DEFAULT_WAIT =

After a recoverable fail this describes how much time we shall wait before readding the reducer back on to the queue.

10

Class Method Summary collapse

Methods inherited from QueueProcess

queue

Class Method Details

.perform(pid, key) ⇒ Object

Doesn’t handle redundant workers and fault tolerance

TODO: Resque::AutoRetry might mess this up.



27
28
29
30
31
32
33
34
35
36
37
 
# File 'lib/mapredus/reducer.rb', line 27

def self.perform(pid, key)
  process = Process.open(pid)
  reduce(process.map_values(key)) do |reduce_val|
    process.emit( key, reduce_val )
  end
rescue MapRedus::RecoverableFail
  Master.enslave_later_reduce(process, key)
ensure
  Master.free_slave(pid)
  process.next_state
end

.reduce(values) ⇒ Object

Raises:



22
 
# File 'lib/mapredus/reducer.rb', line 22

def self.reduce(values); raise InvalidReducer; end

.wait ⇒ Object

seconds



20
 
# File 'lib/mapredus/reducer.rb', line 20

def self.wait; DEFAULT_WAIT; end