lib/dwc-archive/classification_normalizer.rb in dwc-archive-0.4.10 vs lib/dwc-archive/classification_normalizer.rb in dwc-archive-0.4.11

@@ -2,17 +2,18 @@
 require 'parsley-store'
 
 class DarwinCore 
     
   class TaxonNormalized
-    attr_accessor :id, :parent_id, :classification_path, :current_name, :current_name_canonical, :synonyms, :vernacular_names, :rank, :status
+    attr_accessor :id, :parent_id, :classification_path_id, :classification_path, :current_name, :current_name_canonical, :synonyms, :vernacular_names, :rank, :status
 
     def initialize
       @id = @parent_id = @rank = @status = nil
       @current_name = ''
       @current_name_canonical = ''
       @classification_path = []
+      @classification_path_id = []
       @synonyms = []
       @vernacular_names = []
     end
 
   end
@@ -20,23 +21,33 @@
   class SynonymNormalized < Struct.new(:name, :canonical_name, :status);end
   class VernacularNormalized < Struct.new(:name, :language);end
 
   class ClassificationNormalizer
     attr_accessor :verbose
-    attr_reader :error_names
+    attr_reader :error_names, :tree
 
     def initialize(dwc_instance, verbose = false)
       @dwc = dwc_instance
       @core = get_fields(@dwc.core)
       @extensions = @dwc.extensions.map { |e| [e, get_fields(e)] }
       @res = {}
       @parser = ParsleyStore.new(1,2)
       @verbose = verbose
       @verbose_count = 10000
+      @name_strings = {}
       @error_names = []
+      @tree = {}
     end
 
+    def add_name_string(name_string)
+      @name_strings[name_string] = 1 unless @name_strings[name_string]
+    end
+
+    def name_strings
+      @name_strings.keys
+    end
+
     def normalize
       @res = {}
       ingest_core
       calculate_classification_path
       ingest_extensions
@@ -53,10 +64,12 @@
         parsed_name = @parser.parse(a_scientific_name)[:scientificName]
       rescue
         @parser = ParsleyStore.new(1,2)
         parsed_name = @parser.parse(a_scientific_name)[:scientificName]
       end
+      add_name_string(a_scientific_name)
+      add_name_string(parsed_name[:canonical]) if parsed_name[:parsed]
       parsed_name[:parsed] ? parsed_name[:canonical] : a_scientific_name
     end
     
     def get_fields(element)
       data = element.fields.inject({}) { |res, f| res[f[:term].split('/')[-1].downcase.to_sym] = f[:index].to_i; res }
@@ -116,12 +129,15 @@
       end
     end
 
     def get_classification_path(taxon)
       return if !taxon.classification_path.empty?
+      current_node = {taxon.id => {}}
       if DarwinCore.nil_field?(taxon.parent_id)
         taxon.classification_path << taxon.current_name_canonical
+        taxon.classification_path_id << taxon.id
+        @tree.merge!(current_node)
       else
         begin
           parent_cp = @res[taxon.parent_id].classification_path
         rescue NoMethodError #name has a parent which is not a current name
           error = "The parent of the taxon \'#{taxon.current_name}\' is deprecated"
@@ -129,12 +145,18 @@
           raise DarwinCore::ParentNotCurrentError, error
         end
         if parent_cp.empty?
           get_classification_path(@res[taxon.parent_id]) 
           taxon.classification_path += @res[taxon.parent_id].classification_path + [taxon.current_name_canonical]
+          taxon.classification_path_id += @res[taxon.parent_id].classification_path_id + [taxon.id]
+          parent_node = @res[taxon.parent_id].classification_path_id.inject(@tree) {|node, id| node[id]}
+          parent_node.merge!(current_node)
         else
           taxon.classification_path += parent_cp + [taxon.current_name_canonical]
+          taxon.classification_path_id += @res[taxon.parent_id].classification_path_id + [taxon.id]
+          parent_node = @res[taxon.parent_id].classification_path_id.inject(@tree) {|node, id| node[id]}
+          parent_node.merge!(current_node)
         end
       end
     end
 
     def ingest_extensions
@@ -165,9 +187,10 @@
         count = i + 1
         puts "Ingesting %s'th record" % count if @verbose && count % @verbose_count == 0
         @res[r[fields[:id]]].vernacular_names << VernacularNormalized.new(
           r[fields[:vernacularname]],
           fields[:languagecode] ? r[fields[:languagecode]] : nil)
+        add_name_string(r[fields[:vernacularname]])
       end
     end
 
   end
 end