Merge Importer and TaskCreator into one.

Fix the only unit test we have for the Ruby sub-system.

Signed-off-by: Anurag Priyam <[email protected]>

Rakefile

@@ -129,15 +129,6 @@ task 'import', [:annotations_file] do |t, args|
   Importer.new(annotations_file).run
 end
 
-desc 'Create tasks'
-task 'tasks:create', [:annotations_file] do |t, args|
-  require_relative 'app'
-  App.init_config
-  App.load_services
-  annotations_file = args[:annotations_file]
-  TaskCreator.new(annotations_file).run
-end
-
 desc 'IRb.'
 task 'irb' do
   require_relative 'app'

models/genome.rb

@@ -1,3 +1,12 @@
 class Genome < Sequel::Model
   one_to_many :ref_seqs
+
+  # Returns Sequel::Dataset of all mRNAs in this genome.
+  #
+  # Must call `all` on the return value to get an Array.
+  def mRNAs
+    Feature.
+      where(type: 'mRNA').
+      where(ref_seq_id: ref_seqs.map(&:seq_id))
+  end
 end

services/importer.rb

@@ -10,12 +10,12 @@ def initialize(annotations_file)
   end
 
   def run
-    #format
-    #update_tracklist
+    format and update_tracklist
     Genome.db.transaction do
       register_genome
       register_ref_seqs
       register_annotations
+      create_curation_tasks
     end
   end
 
@@ -127,4 +127,55 @@ def nclist_to_features(ref, classes, nclist)
     end
     list
   end
+
+  # Create tasks clubbing overlapping mRNAs into one.
+  def create_curation_tasks
+    # Feature loci on all refs, sorted and grouped by ref.
+    # [
+    #   {
+    #     ref: ...,
+    #     ids: [],
+    #     start_coordinates: [],
+    #     end_coordinates: []
+    #   },
+    #   ...
+    # ]
+    loci_all_ref = genome.mRNAs.
+      select(Sequel.function(:array_agg, Sequel.lit('"id" ORDER BY "start"')).as(:ids),
+      Sequel.function(:array_agg, Sequel.lit('"start" ORDER BY "start"')).as(:start_coordinates),
+      Sequel.function(:array_agg, Sequel.lit('"end" ORDER BY "start"')).as(:end_coordinates),
+      :ref_seq_id).group(:ref_seq_id)
+
+    loci_all_ref.each do |loci_one_ref|
+      groups = call_overlaps loci_one_ref
+      groups.each do |group|
+        ids = group.delete :ids
+        t = Task.create group
+        t.difficulty = ids.length
+        t.save
+      end
+    end
+  end
+
+  # Group overlapping loci together regardless of feature strand.
+  #
+  # About overlapping genes: http://www.biomedcentral.com/1471-2164/9/169.
+  def call_overlaps(loci_one_ref)
+    # Ref being processed.
+    ref = loci_one_ref[:ref_seq_id]
+
+    groups = [] # [{start: , end: , gene_ids: []}, ...]
+    loci_one_ref[:ids].each_with_index do |id, i|
+      start = loci_one_ref[:start_coordinates][i]
+      _end  = loci_one_ref[:end_coordinates][i]
+
+      if not groups.empty? and start < groups.last[:end] # overlap
+        groups.last[:ids] << id
+        groups.last[:end] = [groups.last[:end], _end].max
+      else
+        groups << {ref_seq_id: ref, start: start, end: _end, ids: [id]}
+      end
+    end
+    groups
+  end
 end

tests/rb/test_overlap_detection.rb

@@ -5,28 +5,29 @@
 
   def data
     @data ||= {ref: 'test_ref',
-     feature_ids:               [1, 2, 3, 4, 5, 6, 7],
-     feature_start_coordinates: [10, 30, 40, 50, 100, 110, 130],
-     feature_end_coordinates:   [20, 70, 60, 80, 150, 120, 140]
+     ids:               [1, 2, 3, 4, 5, 6, 7],
+     start_coordinates: [10, 30, 40, 50, 100, 110, 130],
+     end_coordinates:   [20, 70, 60, 80, 150, 120, 140]
     }
   end
 
   def overlaps
-    @groups ||= Importer.new(nil).call_overlaps(data)
+    @groups ||=
+      Importer.new('data/annotations/Solenopsis_invicta/Si_gnF.gff').call_overlaps(data)
   end
 
   it "should work" do
     assert overlaps.length == 3
 
-    assert overlaps[0][:feature_ids].length == 1
+    assert overlaps[0][:ids].length == 1
     assert overlaps[0][:start] == 10
     assert overlaps[0][:end]   == 20
 
-    assert overlaps[1][:feature_ids].length == 3
+    assert overlaps[1][:ids].length == 3
     assert overlaps[1][:start] == 30
     assert overlaps[1][:end]   == 80
 
-    assert overlaps[2][:feature_ids].length == 3
+    assert overlaps[2][:ids].length == 3
     assert overlaps[2][:start] == 100
     assert overlaps[2][:end]   == 150
   end