Class: Jrf::Stage

Inherits:
Object
  • Object
show all
Defined in:
lib/jrf/stage.rb

Defined Under Namespace

Classes: MapReducer, ReducerToken

Instance Attribute Summary collapse

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(block, src: nil) ⇒ Stage

Returns a new instance of Stage.



26
27
28
29
30
31
32
33
34
35
36
# File 'lib/jrf/stage.rb', line 26

def initialize(block, src: nil)
  @src = src
  @reducers = []
  @cursor = 0
  @template = nil
  @mode = nil # nil=unknown, :reducer, :passthrough
  @map_transforms = {}
  @ctx = Class.new(RowContext) do
    define_method(:__jrf_expr__, &block)
  end.new
end

Instance Attribute Details

#srcObject (readonly)

Returns the value of attribute src.



11
12
13
# File 'lib/jrf/stage.rb', line 11

def src
  @src
end

Class Method Details

.resolve_template(template, reducers) ⇒ Object



13
14
15
16
17
18
19
20
21
22
23
24
# File 'lib/jrf/stage.rb', line 13

def self.resolve_template(template, reducers)
  if template.is_a?(ReducerToken)
    rows = reducers.fetch(template.index).finish
    rows.length == 1 ? rows.first : rows
  elsif template.is_a?(Array)
    template.map { |v| resolve_template(v, reducers) }
  elsif template.is_a?(Hash)
    template.transform_values { |v| resolve_template(v, reducers) }
  else
    template
  end
end

Instance Method Details

#call(input) ⇒ Object



38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
# File 'lib/jrf/stage.rb', line 38

def call(input)
  @ctx.reset(input)
  @cursor = 0
  @ctx.__jrf_current_stage = self
  result = @ctx.__jrf_expr__

  if @mode.nil?
    if @reducers.any?
      @mode = :reducer
      @template = result
    else
      @mode = :passthrough
    end
  end

  (@mode == :reducer) ? Control::DROPPED : result
end

#decomposable?Boolean

Returns true if all reducers in this stage are DecomposableReduce instances, meaning partial accumulators from parallel workers can be merged.

Returns:

  • (Boolean)


178
179
180
181
# File 'lib/jrf/stage.rb', line 178

def decomposable?
  @mode == :reducer && @reducers.any? &&
    @reducers.all? { |r| r.is_a?(Reducers::DecomposableReduce) }
end

#finishObject



166
167
168
169
170
171
172
173
174
# File 'lib/jrf/stage.rb', line 166

def finish
  return [] unless @mode == :reducer && @reducers.any?

  if @template.is_a?(ReducerToken)
    @reducers.fetch(@template.index).finish
  else
    [self.class.resolve_template(@template, @reducers)]
  end
end

#merge_partials!(other_partials) ⇒ Object

Merges an array of partial accumulators (from another worker) into this stage’s reducers.



196
197
198
199
200
# File 'lib/jrf/stage.rb', line 196

def merge_partials!(other_partials)
  @reducers.each_with_index do |reducer, i|
    reducer.merge_partial(other_partials[i])
  end
end

#partial_accumulatorsObject

Returns an array of raw accumulator values, one per reducer.



184
185
186
# File 'lib/jrf/stage.rb', line 184

def partial_accumulators
  @reducers.map(&:partial)
end

#replace_accumulators!(partials) ⇒ Object

Replaces all reducer accumulators with the given values.



189
190
191
192
193
# File 'lib/jrf/stage.rb', line 189

def replace_accumulators!(partials)
  @reducers.each_with_index do |reducer, i|
    reducer.instance_variable_set(:@acc, partials[i])
  end
end

#step_apply(collection, &block) ⇒ Object

Raises:

  • (TypeError)


122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
# File 'lib/jrf/stage.rb', line 122

def step_apply(collection, &block)
  raise TypeError, "apply expects Array, got #{collection.class}" unless collection.is_a?(Array)

  apply_reducers = []
  template = nil
  results = []

  collection.each do |v|
    with_scoped_reducers(apply_reducers) do
      result = @ctx.send(:__jrf_with_current_input, v) { block.call(v) }
      template ||= result
      results << result
    end
  end

  if apply_reducers.any?
    self.class.resolve_template(template, apply_reducers)
  else
    results.each_with_object([]) do |mapped, arr|
      next if mapped.equal?(Control::DROPPED)
      if mapped.is_a?(Control::Flat)
        arr.concat(Array(mapped.value))
      else
        arr << mapped
      end
    end
  end
end

#step_group_by(key, &block) ⇒ Object



151
152
153
154
155
156
157
158
159
160
161
162
163
164
# File 'lib/jrf/stage.rb', line 151

def step_group_by(key, &block)
  idx = @cursor
  map_reducer = (@reducers[idx] ||= MapReducer.new(:group_by, false))

  row = @ctx._
  slot = map_reducer.slot(key)
  with_scoped_reducers(slot.reducers) do
    result = @ctx.send(:__jrf_with_current_input, row) { block.call(row) }
    slot.template ||= result
  end

  @cursor += 1
  ReducerToken.new(idx)
end

#step_map(builtin, collection, &block) ⇒ Object



77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
# File 'lib/jrf/stage.rb', line 77

def step_map(builtin, collection, &block)
  idx = @cursor
  @cursor += 1

  if collection.is_a?(Array)
    raise TypeError, "map_values expects Hash, got Array" if builtin == :map_values
  elsif !collection.is_a?(Hash)
    raise TypeError, "#{builtin} expects #{builtin == :map_values ? "Hash" : "Array or Hash"}, got #{collection.class}"
  end

  # Transformation mode (detected on first call)
  if @map_transforms[idx]
    return transform_collection(builtin, collection, &block)
  end

  map_reducer = (@reducers[idx] ||= MapReducer.new(builtin, collection.is_a?(Array)))

  if collection.is_a?(Array)
    collection.each_with_index do |v, i|
      slot = map_reducer.slot(i)
      with_scoped_reducers(slot.reducers) do
        result = @ctx.send(:__jrf_with_current_input, v) { block.call(v) }
        slot.template ||= result
      end
    end
  else
    collection.each do |k, v|
      slot = map_reducer.slot(k)
      with_scoped_reducers(slot.reducers) do
        result = @ctx.send(:__jrf_with_current_input, v) { invoke_block(builtin, block, k, v) }
        slot.template ||= result
      end
    end
  end

  # Detect transformation: no reducers were allocated in any slot
  if @mode.nil? && map_reducer.slots.values.all? { |s| s.reducers.empty? }
    @map_transforms[idx] = true
    @reducers[idx] = nil
    return transformed_slots(builtin, map_reducer)
  end

  ReducerToken.new(idx)
end

#step_reduce(value, initial:, finish: nil, merge: nil, step_fn: nil, &step_block) ⇒ Object



56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
# File 'lib/jrf/stage.rb', line 56

def step_reduce(value, initial:, finish: nil, merge: nil, step_fn: nil, &step_block)
  idx = @cursor
  step_fn ||= step_block

  if @reducers[idx].nil?
    finish_rows = finish || ->(acc) { [acc] }
    @reducers[idx] = if merge
      Reducers.decomposable_reduce(initial, merge: merge, finish: finish_rows, &step_fn)
    else
      Reducers.reduce(initial, finish: finish_rows, &step_fn)
    end
    result = ReducerToken.new(idx)
  else
    result = Control::DROPPED
  end

  @reducers[idx].step(value)
  @cursor = idx + 1
  result
end