adjust to new wrapper, replace activerecord with datamapper (remove activerecord code), is_classification to feature_type

5 files changed, 194 insertions, 81 deletions

diff --git a/lib/active_record_setup.rb b/lib/active_record_setup.rb
index 3682c7a..404bfec 100644..100755
--- a/lib/active_record_setup.rb
+++ b/lib/active_record_setup.rb
@@ -1,17 +1,17 @@
 
-gem "activerecord", "= 2.3.8"
-gem "ar-extensions", "= 0.9.2"
-['rubygems', 'logger', 'active_record', 'ar-extensions', 'opentox-ruby-api-wrapper' ].each do |g|
+#gem "activerecord", "= 2.3.8"
+#gem "ar-extensions", "= 0.9.2"
+['rubygems', 'logger', 'active_record', 'opentox-ruby' ].each do |g| #'ar-extensions',
     require g
 end
 
 unless ActiveRecord::Base.connected?
   ActiveRecord::Base.establish_connection(  
-     :adapter => @@config[:database][:adapter],
-     :host => @@config[:database][:host],
-     :database => @@config[:database][:database],
-     :username => @@config[:database][:username],
-     :password => @@config[:database][:password]
+     :adapter => CONFIG[:database][:adapter],
+     :host => CONFIG[:database][:host],
+     :database => CONFIG[:database][:database],
+     :username => CONFIG[:database][:username],
+     :password => CONFIG[:database][:password]
   )
   ActiveRecord::Base.logger = Logger.new("/dev/null")
 end
@@ -19,6 +19,9 @@ end
 class ActiveRecord::Base
   
   def self.find_like(filter_params)
+    
+    raise "find like removed"
+    
     #puts "params before "+filter_params.inspect
     filter_params.keys.each do |k|
       key = k.to_s
diff --git a/lib/ot_predictions.rb b/lib/ot_predictions.rb
index 0175a0c..52683a6 100644..100755
--- a/lib/ot_predictions.rb
+++ b/lib/ot_predictions.rb
@@ -15,7 +15,7 @@ module Lib
       return @compounds[instance_index]
     end
   
-    def initialize(is_classification, test_dataset_uri, test_target_dataset_uri, 
+    def initialize(feature_type, test_dataset_uri, test_target_dataset_uri, 
       prediction_feature, prediction_dataset_uri, predicted_variable, task=nil)
       
         LOGGER.debug("loading prediciton via test-dataset:'"+test_dataset_uri.to_s+
@@ -53,38 +53,47 @@ module Lib
           raise "prediction_feature not found in test_target_dataset\n"+
                 "prediction_feature: '"+prediction_feature.to_s+"'\n"+
                 "test_target_dataset: '"+test_target_dataset_uri.to_s+"'\n"+
-                "available features are: "+test_target_dataset.features.inspect if test_target_dataset.features.index(prediction_feature)==nil
+                "available features are: "+test_target_dataset.features.inspect if test_target_dataset.features.keys.index(prediction_feature)==nil
         end
         
+        test_dataset.load_all
         @compounds = test_dataset.compounds
         LOGGER.debug "test dataset size: "+@compounds.size.to_s
         raise "test dataset is empty" unless @compounds.size>0
-        class_values = is_classification ? OpenTox::Feature.domain(prediction_feature) : nil
+        class_values = feature_type=="classification" ? OpenTox::Feature.new(prediction_feature).domain : nil
         
         actual_values = []
         @compounds.each do |c|
-          value = test_target_dataset.get_value(c, prediction_feature)
-          
-          if is_classification
-            value = value.to_s unless value==nil
-            raise "illegal class_value of actual value "+value.to_s+" class: "+
-              value.class.to_s unless value==nil or class_values.index(value)!=nil
-            actual_values.push class_values.index(value) 
-          else
-            begin
-              value = value.to_f unless value==nil or value.is_a?(Numeric)
-            rescue
-              LOGGER.warn "no numeric value for regression: '"+value.to_s+"'"
-              value = nil
-            end
-            actual_values.push value
+          case feature_type
+          when "classification"
+            actual_values << classification_value(test_target_dataset, c, prediction_feature, class_values)
+          when "regression"
+            actual_values << regression_value(test_target_dataset, c, prediction_feature)
           end
         end
         task.progress(40) if task # loaded actual values
         
         prediction_dataset = OpenTox::Dataset.find prediction_dataset_uri
         raise "prediction dataset not found: '"+prediction_dataset_uri.to_s+"'" unless prediction_dataset
-        raise "prediction-feature not found: '"+predicted_variable+"' in prediction-dataset: "+prediction_dataset_uri.to_s+", available features: "+prediction_dataset.features.inspect if prediction_dataset.features.index(predicted_variable)==nil
+        
+        # TODO: remove LAZAR_PREDICTION_DATASET_HACK
+        no_prediction_feature = prediction_dataset.features.keys.index(predicted_variable)==nil
+        if no_prediction_feature
+          one_entry_per_compound = prediction_dataset.data_entries.keys.size == @compounds.size
+          @compounds.each do |c|
+            if prediction_dataset.data_entries[c].size != 1
+              one_entry_per_compound = false
+              break
+            end
+          end
+          msg = "prediction-feature not found: '"+predicted_variable+"' in prediction-dataset: "+prediction_dataset_uri.to_s+", available features: "+
+            prediction_dataset.features.keys.inspect
+          if one_entry_per_compound
+            LOGGER.warn msg
+          else
+            raise msg
+          end
+        end
         
         raise "more predicted than test compounds test:"+@compounds.size.to_s+" < prediction:"+
           prediction_dataset.compounds.size.to_s if @compounds.size < prediction_dataset.compounds.size
@@ -102,38 +111,79 @@ module Lib
             predicted_values << nil
             confidence_values << nil
           else
-            if is_classification
-              value = prediction_dataset.get_predicted_class(c, predicted_variable)
-              value = value.to_s unless value==nil
-              raise "illegal class_value of predicted value "+value.to_s+" class: "+value.class.to_s unless value==nil or class_values.index(value)!=nil
-              predicted_values << class_values.index(value)
-            else
-              value = prediction_dataset.get_predicted_regression(c, predicted_variable)
-              begin
-                value = value.to_f unless value==nil or value.is_a?(Numeric)
-              rescue
-                LOGGER.warn "no numeric value for regression: '"+value.to_s+"'"
-                value = nil
-              end
-              predicted_values << value
+            case feature_type
+            when "classification"
+              # TODO: remove LAZAR_PREDICTION_DATASET_HACK
+              predicted_values << classification_value(prediction_dataset, c, no_prediction_feature ? nil : predicted_variable, class_values)
+            when "regression"
+              predicted_values << regression_value(prediction_dataset, c, no_prediction_feature ? nil : predicted_variable)
+            end
+            # TODO confidence_values << prediction_dataset.get_prediction_confidence(c, predicted_variable)
+            conf = 1
+            begin
+              feature = prediction_dataset.data_entries[c].keys[0]
+              feature_data = prediction_dataset.features[feature]
+              conf = feature_data[OT.confidence] if feature_data[OT.confidence]!=nil 
+            rescue
+              LOGGER.warn "could not get confidence"
             end
-            confidence_values << prediction_dataset.get_prediction_confidence(c, predicted_variable)
+            confidence_values << conf
           end
         end
         task.progress(80) if task # loaded predicted values and confidence
         
-        super(predicted_values, actual_values, confidence_values, is_classification, class_values)
+        super(predicted_values, actual_values, confidence_values, feature_type, class_values)
         raise "illegal num compounds "+num_info if  @compounds.size != @predicted_values.size
         task.progress(100) if task # done with the mathmatics
     end
     
+    private
+    def regression_value(dataset, compound, feature)
+      v = value(dataset, compound, feature)
+      begin
+        v = v.to_f unless v==nil or v.is_a?(Numeric)
+        v
+      rescue
+        LOGGER.warn "no numeric value for regression: '"+v.to_s+"'"
+        nil
+      end
+    end
+    
+    def classification_value(dataset, compound, feature, class_values)
+      v = value(dataset, compound, feature)
+      i = class_values.index(v)
+      raise "illegal class_value of predicted value "+v.to_s+" class: "+v.class.to_s unless v==nil or i!=nil
+      i
+    end
+    
+    def value(dataset, compound, feature)
+      
+      if feature==nil
+        v = dataset.data_entries[compound].values[0]
+      else
+        v = dataset.data_entries[compound][feature]
+      end
+      raise "no array" unless v.is_a?(Array)
+      if v.size>1
+        raise "multiple values"
+      elsif v.size==1
+        v = v[0]
+      else
+        v = nil
+      end
+      raise "array" if v.is_a?(Array)
+      v = nil if v.to_s.size==0
+      v
+    end
 
+    public
     def compute_stats
     
       res = {}
-      if @is_classification
+      case @feature_type
+      when "classification"
         (Lib::VAL_CLASS_PROPS).each{ |s| res[s] = send(s)}  
-      else
+      when "regression"
         (Lib::VAL_REGR_PROPS).each{ |s| res[s] = send(s) }  
       end
       return res
@@ -152,16 +202,18 @@ module Lib
           
           #PENDING!
           begin
-            a.push( "http://ambit.uni-plovdiv.bg:8080/ambit2/depict/cdk?search="+
-              URI.encode(OpenTox::Compound.new(:uri=>p.identifier(i)).smiles) ) if add_pic
+            #a.push( "http://ambit.uni-plovdiv.bg:8080/ambit2/depict/cdk?search="+
+            #  URI.encode(OpenTox::Compound.new(:uri=>p.identifier(i)).smiles) ) if add_pic
+            a << p.identifier(i)+"/image"
           rescue => ex
+            raise ex
             #a.push("Could not add pic: "+ex.message)
-            a.push(p.identifier(i))
+            #a.push(p.identifier(i))
           end
           
           a << (format ? p.actual_value(i).to_nice_s : p.actual_value(i))
           a << (format ? p.predicted_value(i).to_nice_s : p.predicted_value(i))
-          if p.classification?
+          if p.feature_type=="classification"
             if (p.predicted_value(i)!=nil and p.actual_value(i)!=nil)
               a << (p.classification_miss?(i) ? 1 : 0)
             else
@@ -180,7 +232,7 @@ module Lib
       header << "compound" if add_pic
       header << "actual value"
       header << "predicted value"
-      header << "missclassified" if predictions[0].classification?
+      header << "missclassified" if predictions[0].feature_type=="classification"
       header << "confidence value" if predictions[0].confidence_values_available?
       header << "compound-uri"
       res.insert(0, header)
diff --git a/lib/predictions.rb b/lib/predictions.rb
index 6e50e94..e73dda6 100644..100755
--- a/lib/predictions.rb
+++ b/lib/predictions.rb
@@ -22,13 +22,13 @@ module Lib
     def initialize( predicted_values, 
                     actual_values, 
                     confidence_values, 
-                    is_classification, 
+                    feature_type, 
                     class_domain=nil )
                     
       @predicted_values = predicted_values
       @actual_values = actual_values
       @confidence_values = confidence_values
-      @is_classification = is_classification
+      @feature_type = feature_type
       @class_domain = class_domain
       @num_classes = 1
       
@@ -36,6 +36,8 @@ module Lib
       #puts "actual:     "+actual_values.inspect
       #puts "confidence: "+confidence_values.inspect
       
+      raise "unknown feature_type: "+@feature_type.to_s unless 
+        @feature_type=="classification" || @feature_type=="regression"
       raise "no predictions" if @predicted_values.size == 0
       num_info = "predicted:"+@predicted_values.size.to_s+
         " confidence:"+@confidence_values.size.to_s+" actual:"+@actual_values.size.to_s
@@ -54,15 +56,16 @@ module Lib
       #  @confidence_values=nil
       #end
       
-      if @is_classification
+      case @feature_type
+      when "classification"
         raise "class_domain missing while performing classification" unless @class_domain
         @num_classes = @class_domain.size
         raise "num classes < 2" if @num_classes<2
         { "predicted"=>@predicted_values, "actual"=>@actual_values }.each do |s,values|
           values.each{ |v| raise "illegal "+s+" classification-value ("+v.to_s+"),"+
-            "has to be either nil or index of predicted-values" if v!=nil and (v<0 or v>@num_classes)}
+            "has to be either nil or index of predicted-values" if v!=nil and (!v.is_a?(Numeric) or v<0 or v>@num_classes)}
         end
-      else
+      when "regresssion"
         raise "class_domain != nil while performing regression" if @class_domain
         { "predicted"=>@predicted_values, "actual"=>@actual_values }.each do |s,values|
           values.each{ |v| raise "illegal "+s+" regression-value ("+v.to_s+"),"+
@@ -83,7 +86,8 @@ module Lib
       @num_predicted = 0
       @num_unpredicted = 0
       
-      if @is_classification
+      case @feature_type
+      when "classification"
         @confusion_matrix = []
         @class_domain.each do |v|
           @confusion_matrix.push( Array.new( @num_classes, 0 ) )
@@ -91,7 +95,7 @@ module Lib
         
         @num_correct = 0
         @num_incorrect = 0
-      else
+      when "regression"
         @sum_error = 0
         @sum_abs_error = 0
         @sum_squared_error = 0
@@ -122,14 +126,15 @@ module Lib
         else
           @num_predicted += 1
           
-          if @is_classification
+          case @feature_type
+          when "classification"
             @confusion_matrix[actual_value][predicted_value] += 1
             if (predicted_value == actual_value)
               @num_correct += 1
             else
               @num_incorrect += 1
             end
-          else
+          when "regression"
             delta = predicted_value - actual_value
             @sum_error += delta
             @sum_abs_error += delta.abs
@@ -156,13 +161,13 @@ module Lib
     end
     
     def percent_correct
-      raise "no classification" unless @is_classification
+      raise "no classification" unless @feature_type=="classification"
       return 0 if @num_with_actual_value==0
       return 100 * @num_correct / @num_with_actual_value.to_f
     end
     
     def percent_incorrect
-      raise "no classification" unless @is_classification
+      raise "no classification" unless @feature_type=="classification"
       return 0 if @num_with_actual_value==0
       return 100 * @num_incorrect / @num_with_actual_value.to_f
     end
@@ -190,17 +195,17 @@ module Lib
     end
 
     def num_correct
-      raise "no classification" unless @is_classification
+      raise "no classification" unless @feature_type=="classification"
       return @num_correct
     end
 
     def num_incorrect
-      raise "no classification" unless @is_classification
+      raise "no classification" unless @feature_type=="classification"
       return @num_incorrect
     end
     
     def num_unclassified
-      raise "no classification" unless @is_classification
+      raise "no classification" unless @feature_type=="classification"
       return @num_unpredicted
     end
     
@@ -209,7 +214,7 @@ module Lib
     #     and values: <int-value>
     def confusion_matrix
       
-      raise "no classification" unless @is_classification
+      raise "no classification" unless @feature_type=="classification"
       res = {}
       (0..@num_classes-1).each do |actual|
           (0..@num_classes-1).each do |predicted|
@@ -505,9 +510,10 @@ module Lib
     end
   
     def predicted_value(instance_index)
-      if @is_classification
+      case @feature_type 
+      when "classification"
         @predicted_values[instance_index]==nil ? nil : @class_domain[@predicted_values[instance_index]]
-      else
+      when "regression"
         @predicted_values[instance_index]
       end
     end
@@ -517,9 +523,10 @@ module Lib
     end
     
     def actual_value(instance_index)
-      if @is_classification
+      case @feature_type 
+      when "classification"
         @actual_values[instance_index]==nil ? nil : @class_domain[@actual_values[instance_index]]
-      else
+      when "regression"
         @actual_values[instance_index]
       end
     end
@@ -529,13 +536,13 @@ module Lib
     end      
     
     def classification_miss?(instance_index)
-      raise "no classification" unless @is_classification
+      raise "no classification" unless @feature_type=="classification"
       return false if predicted_value(instance_index)==nil or actual_value(instance_index)==nil
       return predicted_value(instance_index) != actual_value(instance_index)
     end
     
-    def classification?
-      @is_classification
+    def feature_type
+      @feature_type
     end
     
     def confidence_values_available?
diff --git a/lib/test_util.rb b/lib/test_util.rb
index ecab76c..0deee04 100644..100755
--- a/lib/test_util.rb
+++ b/lib/test_util.rb
@@ -10,11 +10,11 @@ module Lib
     end
     
     def self.wait_for_task(uri)
-      if OpenTox::Utils.task_uri?(uri)
+      if uri.task_uri?
         task = OpenTox::Task.find(uri)
         task.wait_for_completion
         raise "task failed: "+uri.to_s+", error is:\n"+task.description if task.error?
-        uri = task.resultURI
+        uri = task.result_uri
       end
       return uri
     end
diff --git a/lib/validation_db.rb b/lib/validation_db.rb
index cb3ece7..83b7e2f 100644..100755
--- a/lib/validation_db.rb
+++ b/lib/validation_db.rb
@@ -49,11 +49,40 @@ module Lib
   VAL_MERGE_AVG = VAL_PROPS_AVG + VAL_CLASS_PROPS_SINGLE_AVG + VAL_CLASS_PROPS_PER_CLASS_AVG + VAL_REGR_PROPS
   
 
-  class Validation < ActiveRecord::Base
-    serialize :classification_statistics
-    serialize :regression_statistics
+#  class Validation < ActiveRecord::Base
+#    serialize :classification_statistics
+#    serialize :regression_statistics
+#    
+#    alias_attribute :date, :created_at
     
-    alias_attribute :date, :created_at
+  class Validation 
+    include DataMapper::Resource
+  
+    property :id, Serial
+    property :validation_type, String, :length => 255
+    property :model_uri, String, :length => 255
+    property :algorithm_uri, String, :length => 255
+    property :training_dataset_uri, String, :length => 255
+    property :test_target_dataset_uri, String, :length => 255
+    property :test_dataset_uri, String, :length => 255
+    property :prediction_dataset_uri, String, :length => 255
+    property :prediction_feature, String, :length => 255
+    property :created_at, DateTime
+    property :num_instances, Integer
+    property :num_without_class, Integer
+    property :num_unpredicted, Integer
+    property :crossvalidation_id, Integer
+    property :crossvalidation_fold, Integer
+    property :real_runtime, Float
+    property :percent_without_class, Float
+    property :percent_unpredicted, Float
+    property :classification_statistics, Object
+    property :regression_statistics, Object
+    property :finished, Boolean, :default => false
+    
+    def date
+      created_at
+    end
     
     def validation_uri
       $sinatra.url_for("/"+self.id.to_s, :full)
@@ -77,8 +106,23 @@ module Lib
     
   end
   
-  class Crossvalidation < ActiveRecord::Base
-    alias_attribute :date, :created_at
+#  class Crossvalidation < ActiveRecord::Base
+#    alias_attribute :date, :created_at
+  class Crossvalidation 
+    include DataMapper::Resource
+  
+    property :id, Serial
+    property :algorithm_uri, String, :length => 255
+    property :dataset_uri, String, :length => 255
+    property :created_at, DateTime
+    property :num_folds, Integer, :default => 10
+    property :random_seed, Integer, :default => 1
+    property :finished, Boolean, :default => false
+    property :stratified, Boolean, :default => false
+    
+    def date
+      created_at
+    end
     
     def crossvalidation_uri
       $sinatra.url_for("/crossvalidation/"+self.id.to_s, :full) if self.id
@@ -88,7 +132,8 @@ module Lib
     # in terms of dataset_uri,num_folds,stratified,random_seed
     # further conditions can be specified in __conditions__
     def self.find_all_uniq(conditions={})
-      cvs = Lib::Crossvalidation.find(:all, :conditions => conditions)
+      #cvs = Lib::Crossvalidation.find(:all, :conditions => conditions)
+      cvs = Lib::Crossvalidation.all(:conditions => conditions)
       uniq = []
       cvs.each do |cv|
         match = false
@@ -105,3 +150,9 @@ module Lib
     end
   end
 end
+
+
+Lib::Validation.auto_upgrade!
+Lib::Validation.raise_on_save_failure = true
+Lib::Crossvalidation.auto_upgrade!
+Lib::Crossvalidation.raise_on_save_failure = true
\ No newline at end of file