Create hierarchical dates for opengeometadata records

lib/earthworks/harvester.rb

@@ -47,10 +47,16 @@ def transform_record(record, path)
         record.update({ 'schema_provider_s' => transformed_provider })
       end
 
+      # Filter out themes that are not in the controlled vocabulary from OGM
       if (themes = record['dcat_theme_sm'])
-        # Filter out themes that are not in the controlled vocabulary from OGM
         record.update({ 'dcat_theme_sm' => themes.select { |theme| Settings.ALLOWED_OGM_THEMES.include?(theme) } })
       end
+
+      # Add a hierarchicalized version of the dates for the year facet
+      if (years = record['gbl_indexYear_im'])
+        record.update({ 'date_hierarchy_sm' => hierarchicalize_year_list(years) })
+      end
+
       record
     end
 
@@ -63,5 +69,43 @@ def record_repo(path)
     def repositories
       @repositories ||= @ogm_repos ? super.compact.select { |repo| @ogm_repos.key?(repo) } : super
     end
+
+    # NOTE: this duplicates logic in searchworks_traject_indexer, see:
+    # https://github.com/sul-dlss/searchworks_traject_indexer/pull/1521
+    def hierarchicalize_year_list(years)
+      centuries = Set.new
+      decades = Set.new
+
+      hierarchicalized_years = years.map do |year|
+        century, decade = centimate_and_decimate(year)
+        centuries << century
+        decades << [century, decade].join(':')
+        [century, decade, year].join(':')
+      end
+
+      centuries.to_a + decades.to_a + hierarchicalized_years
+    end
+
+    # NOTE: this duplicates logic in searchworks_traject_indexer, see:
+    # https://github.com/sul-dlss/searchworks_traject_indexer/pull/1521
+    def centimate_and_decimate(maybe_year)
+      parsed_date = Date.new(maybe_year.to_i)
+      [century_from_date(parsed_date), decade_from_date(parsed_date)]
+    rescue Date::Error
+      %w[unknown_century unknown_decade] # guess not
+    end
+
+    # NOTE: this duplicates logic in searchworks_traject_indexer, see:
+    # https://github.com/sul-dlss/searchworks_traject_indexer/pull/1521
+    def century_from_date(date)
+      date.strftime('%C00-%C99')
+    end
+
+    # NOTE: this duplicates logic in searchworks_traject_indexer, see:
+    # https://github.com/sul-dlss/searchworks_traject_indexer/pull/1521
+    def decade_from_date(date)
+      decade_prefix = (date.strftime('%Y').to_i / 10).to_s
+      "#{decade_prefix}0-#{decade_prefix}9"
+    end
   end
 end

spec/lib/earthworks/harvester_spec.rb

@@ -61,14 +61,31 @@
 
   describe '#docs_to_index' do
     # Provenance value will be transformed by our ogm_repos config
-    let(:psu_doc) { { schema_provider_s: 'Pennsylvania State University', gbl_mdVersion_s: 'Aardvark' }.to_json }
+    let(:psu_doc) do
+      {
+        schema_provider_s: 'Pennsylvania State University',
+        gbl_mdVersion_s: 'Aardvark',
+        gbl_indexYear_im: [1701, 1980, 1991, 1995]
+      }.to_json
+    end
     let(:psu_path) { "#{ogm_path}/edu.psu/metadata-aardvark/Maps/08d-01/geoblacklight.json" }
 
     # PolicyMap records have placeholder data and should be skipped
-    let(:policymap_doc) { { schema_provider_s: 'Geoblacklight', gbl_mdVersion_s: 'Aardvark' }.to_json }
+    let(:policymap_doc) do
+      {
+        schema_provider_s: 'Geoblacklight',
+        gbl_mdVersion_s: 'Aardvark'
+      }.to_json
+    end
     let(:policymap_path) { "#{ogm_path}/shared-repository/gbl-policymap/records/geoblacklight.json" }
 
-    let(:psu_doc2_hash) { { schema_provider_s: 'Pennsylvania State University', gbl_mdVersion_s: 'Aardvark' } }
+    let(:psu_doc2_hash) do
+      {
+        schema_provider_s: 'Pennsylvania State University',
+        gbl_mdVersion_s: 'Aardvark',
+        gbl_indexYear_im: [1538]
+      }
+    end
     let(:psu_doc2) { psu_doc2_hash.to_json }
     let(:psu_path2) { "#{ogm_path}/edu.psu/metadata-aardvark/Maps/08d-01/geoblacklight_2.json" }
 
@@ -101,6 +118,19 @@
       expect(docs.first.first['schema_provider_s']).to eq('Penn State')
     end
 
+    it 'generates hierarchical years for the year facet' do
+      docs = harvester.docs_to_index.to_a
+      expect(docs.first.first['date_hierarchy_sm']).to contain_exactly(
+        '1700-1799', '1900-1999',
+        '1700-1799:1700-1709', '1900-1999:1980-1989', '1900-1999:1990-1999',
+        '1700-1799:1700-1709:1701', '1900-1999:1980-1989:1980', '1900-1999:1990-1999:1991', '1900-1999:1990-1999:1995'
+      )
+      expect(docs.last.first['date_hierarchy_sm']).to contain_exactly(
+        '1500-1599', '1500-1599:1530-1539',
+        '1500-1599:1530-1539:1538'
+      )
+    end
+
     context 'when record contains themes outside the controlled vocabulary' do
       let(:psu_doc) do
         psu_doc2_hash.merge({ dcat_theme_sm: %w[Agriculture Biota Farming] }).to_json