Class: Spark::Serializer::AutoBatched

Inherits:
Batched show all
Defined in:
lib/spark/serializer/auto_batched.rb

Overview

AutoBatched serializator

Batch size is computed automatically. Simillar to Python’s AutoBatchedSerializer.

Constant Summary collapse

MAX_RATIO = 
10

Instance Attribute Summary

Attributes inherited from Batched

#serializer

Instance Method Summary collapse

Methods inherited from Batched

#dump, #load, #load_from_io, #to_s

Methods inherited from Base

#==, #check_each, #error, #inspect, #load_from_file, #load_from_io, #to_s

Constructor Details

#initialize(serializer, best_size = 65536) ⇒ AutoBatched

Returns a new instance of AutoBatched.



12
13
14
15
16
17
 
# File 'lib/spark/serializer/auto_batched.rb', line 12

def initialize(serializer, best_size=65536)
  @serializer = serializer
  @best_size = best_size.to_i

  error('Batch size must be greater than 1') if @best_size < 2
end

Instance Method Details

#batched? ⇒ Boolean

Returns:

  • (Boolean) 


19
20
21
 
# File 'lib/spark/serializer/auto_batched.rb', line 19

def batched?
  true
end

#dump_to_io(data, io) ⇒ Object 



30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
 
# File 'lib/spark/serializer/auto_batched.rb', line 30

def dump_to_io(data, io)
  check_each(data)

  # Only Array have .slice
  data = data.to_a

  index = 0
  batch = 2
  max = @best_size * MAX_RATIO

  loop do
    chunk = data.slice(index, batch)
    if chunk.nil? || chunk.empty?
      break
    end

    serialized = @serializer.dump(chunk)
    io.write_string(serialized)

    index += batch

    size = serialized.bytesize
    if size < @best_size
      batch *= 2
    elsif size > max && batch > 1
      batch /= 2
    end
  end

  io.flush
end

#name ⇒ Object 



26
27
28
 
# File 'lib/spark/serializer/auto_batched.rb', line 26

def name
  "AutoBatched(#{@best_size})"
end

#unbatch! ⇒ Object 



23
24
 
# File 'lib/spark/serializer/auto_batched.rb', line 23

def unbatch!
end