Class: Tsuga::Service::Clusterer

Inherits:
Object
  • Object
show all
Defined in:
lib/tsuga/service/clusterer.rb

Defined Under Namespace

Modules: SteppedProgressBar Classes: MutableSet, WriteQueue

Constant Summary collapse

PROXIMITY_RATIO = 
0.15
RUN_SANITY_CHECK = 
false
VERBOSE = 
ENV['VERBOSE']
Tile = 
Tsuga::Model::Tile

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(source: nil, adapter: nil) ⇒ Clusterer

Returns a new instance of Clusterer.



15
16
17
18
19
 
# File 'lib/tsuga/service/clusterer.rb', line 15

def initialize(source: nil, adapter: nil)
  @_source  = source
  @_adapter = adapter
  @_queue   = WriteQueue.new(adapter: adapter)
end

Instance Attribute Details

#_adapter ⇒ Object (readonly)

Returns the value of attribute _adapter.



13
14
15
 
# File 'lib/tsuga/service/clusterer.rb', line 13

def _adapter
  @_adapter
end

#_queue ⇒ Object (readonly)

Returns the value of attribute _queue.



13
14
15
 
# File 'lib/tsuga/service/clusterer.rb', line 13

def _queue
  @_queue
end

#_source ⇒ Object (readonly)

Returns the value of attribute _source.



13
14
15
 
# File 'lib/tsuga/service/clusterer.rb', line 13

def _source
  @_source
end

Instance Method Details

#run ⇒ Object 



21
22
23
24
25
26
27
28
29
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
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
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
121
122
123
124
125
126
127
128
129
130
131
132
133
 
# File 'lib/tsuga/service/clusterer.rb', line 21

def run
  # delete all clusters
  _adapter.delete_all

  # create lowest-level clusters
  _source.find_each do |record|
    _queue.push _adapter.build_from(Tsuga::MAX_DEPTH, record)
  end
  _queue.flush

  # for all depths N from 18 to 3
  (Tsuga::MAX_DEPTH-1).downto(Tsuga::MIN_DEPTH) do |depth|
    progress.log "depth #{depth}"                                       if VERBOSE
    progress.title = "#{depth}.0"                                       if VERBOSE

    # create clusters at this level from children
    # TODO: use a save queue, only run saves if > 100 clusters to write
    cluster_ids = Set.new
    _adapter.at_depth(depth+1).find_each do |child|
      _queue.push _adapter.build_from(depth, child)
    end
    _queue.flush
    cluster_ids = MutableSet.new(_adapter.at_depth(depth).collect_ids)

    if cluster_ids.empty?
      progress.log "nothing to cluster"                                 if VERBOSE
      break
    end

    # TODO: group points to cluster by tile, and run on tiles in parallel.

    progress.title = "#{depth}.1"                                       if VERBOSE
    progress.log "started with #{cluster_ids.length} clusters"          if VERBOSE
    progress.set_phase(depth, 1, cluster_ids.length)                    if VERBOSE
    while cluster_ids.any?
      progress.set_progress(cluster_ids.length)                         if VERBOSE

      cluster = _adapter.find_by_id(cluster_ids.first)
      raise 'internal error: cluster was already removed' if cluster.nil?
      tile = Tile.including(cluster, depth: depth)

      clusters = _adapter.in_tile(*tile.neighbours).to_a
      processed_cluster_ids = clusters.collect(&:id)

      # clusters we aggregate in this loop iteration
      # they are _not_ the same as what we pass to the aggregator,
      # just those inside the fence
      fenced_cluster_ids = _adapter.in_tile(tile).collect_ids
      raise RuntimeError, 'no cluster in fence' if fenced_cluster_ids.empty?

      Aggregator.new(clusters:clusters, ratio:PROXIMITY_RATIO, fence:tile).tap do |aggregator|
        aggregator.run

        if VERBOSE
          progress.log("aggregator: %4d left, %2d processed, %2d in fence, %2d updated, %2d dropped" % [
            cluster_ids.length,
            processed_cluster_ids.length,
            fenced_cluster_ids.length,
            aggregator.updated_clusters.length,
            aggregator.dropped_clusters.length]) 
          if aggregator.updated_clusters.any?
            progress.log("updated: #{aggregator.updated_clusters.collect(&:id).join(', ')}")
          end
          if aggregator.dropped_clusters.any?
            progress.log("dropped: #{aggregator.dropped_clusters.collect(&:id).join(', ')}")
          end
        end

        cluster_ids.remove! fenced_cluster_ids
        # updated clusters may need to be reprocessed (they might have fallen close enough to tile edges)
        # TODO: as further optimisation, do not mark for reprocessing clusters that are still inside the fence
        cluster_ids.merge! aggregator.updated_clusters.collect(&:id)
        # destroyed clusters may include some on the outer fringe of the fence tile
        cluster_ids.remove! aggregator.dropped_clusters.collect(&:id)

        aggregator.dropped_clusters.each(&:destroy)
        _adapter.mass_update(aggregator.updated_clusters)
      end

      if RUN_SANITY_CHECK
        # sanity check: all <cluster_ids> should exist
        not_removed = cluster_ids - _adapter.at_depth(depth).collect_ids
        if not_removed.any?
          raise "cluster_ids contains IDs of deleted clusters: #{not_removed.to_a.join(', ')}"
        end

        # sanity check: sum of weights should match that of lower level
        deeper_weight = _adapter.at_depth(depth+1).sum(:weight)
        this_weight   = _adapter.at_depth(depth).sum(:weight)
        if deeper_weight != this_weight
          raise "mismatch between weight at this depth (#{this_weight}) and deeper level (#{deeper_weight})"
        end
      end
    end

    # set parent_id in the whole tree
    # this is made slightly more complicated by #find_each's scoping
    progress.title = "#{depth}.2"                                       if VERBOSE
    child_mappings = {}
    _adapter.at_depth(depth).find_each do |cluster|
      cluster.children_ids.each do |child_id|
        child_mappings[child_id] = cluster.id
      end
    end
    child_mappings.each_pair do |child_id, parent_id|
      cluster = _adapter.find_by_id(child_id)
      cluster.parent_id = parent_id
      _queue.push cluster
    end
    _queue.flush
  end
  progress.finish                                                       if VERBOSE
end