- Restored the DatabaseConfiguration.getDBnameSeparator() method.

- Changed the RandomGenerator.getThreadLocalRandomUnseeded() to ensure we get different random numbers across threads.
- CERT no longer received any dbName parameter as we don't store anything on disk.
- Removed underscores from all temporary names in the framework.
- When we close() a Trainer that has not be loaded or saved the knowledgeBase will be deleted to remove any temporary files.
- The models of a specific dbName are added in a directory structure.
- Created an other level of abstraction for File-based Database Connectors and Configurations.
- Rename Folder to Directory on comments, methods, vars and config files.
- Empty parent directories of the algorithm output are automatically cleaned up.

CHANGELOG.md

@@ -1,17 +1,17 @@
 CHANGELOG
 =========
 
-Version 0.8.0-SNAPSHOT - Build 20161221
+Version 0.8.0-SNAPSHOT - Build 20161222
 ---------------------------------------
 
 - Improved Validation:
     - Removed the ValidationMetrics from the Algorithms. Now it is a separate object called Metrics.
     - Removed the kFold validation from Algorithms. Now we offer a new validator mechanism.
     - A single KnowledgeBase implementation is now used.
 - Removed the unnecessary n & d model parameters from all models.
-- Random unseeded filenames are now produced using RandomGenerator.getThreadLocalRandomUnseeded().
+- Random unseeded filenames are now produced using RandomGenerator.getRandomUnseeded().
 - Removing the need to call KnowledgeBase.init() in any predict/transform method.
-- Improved DatabaseConnector: existsObject method, InMemory now stores objects independently, MapDB stores all files in folder.
+- Improved DatabaseConnector: existsObject method, InMemory now stores objects independently, MapDB stores all files in directory.
 - The training parameters are now provided on the constructor of the algorithms not with a setter.
 - TextClassifier inherits from Modeler.
 - Removed all unnecessary passing of class objects from Stepwise Regression, Wrappers and Ensumble learning classes.
@@ -21,8 +21,16 @@ Version 0.8.0-SNAPSHOT - Build 20161221
 - Created a TrainableBundle to keep track of the Trainables of Modeler, AbstractBoostingBagging and StepwiseRegression.
 - Removed automatic save after fit, now save() must be called.
 - AbstractTrainer no longer stores a local copy of dbName. The save method accepts a dbName.
-- The DatabaseConfiguration.getDBnameSeparator() method was removed.
 - The rename() is created in DatabaseConnectors and it's used by KnowledgeBase to saveAs the models.
+- Restored the DatabaseConfiguration.getDBnameSeparator() method.
+- Changed the RandomGenerator.getThreadLocalRandomUnseeded() to ensure we get different random numbers across threads.
+- CERT no longer received any dbName parameter as we don't store anything on disk.
+- Removed underscores from all temporary names in the framework.
+- When we close() a Trainer that has not be loaded or saved the knowledgeBase will be deleted to remove any temporary files. 
+- The models of a specific dbName are added in a directory structure.
+- Created an other level of abstraction for File-based Database Connectors and Configurations.
+- Rename Folder to Directory on comments, methods, vars and config files.
+- Empty parent directories of the algorithm output are automatically cleaned up.
 
 Version 0.7.1-SNAPSHOT - Build 20161217
 ---------------------------------------

TODO.txt

@@ -1,8 +1,7 @@
 CODE IMPROVEMENTS
 =================
 
-- Can we add all the files of a model in a single folder.
-- Can we make the two constructors of the Trainers to call a common constructor to eliminate duplicate code?
+- Save and load method for Dataset.
 
 - Support of better Transformers (Zscore, decouple boolean transforming from numeric etc).
 - Write a ShuffleSplitValidator class similar to KFold. Perhaps we need a single Validator class and separate Splitters.

datumbox-framework-applications/src/main/java/com/datumbox/framework/applications/datamodeling/Modeler.java

@@ -39,7 +39,7 @@ public class Modeler extends AbstractTrainer<Modeler.ModelParameters, Modeler.Tr
     private static final String FS_KEY = "fs";
     private static final String ML_KEY = "ml";
 
-    private TrainableBundle bundle = new TrainableBundle();
+    private final TrainableBundle bundle = new TrainableBundle();
 
     /**
      * It contains all the Model Parameters which are learned during the training.
@@ -240,9 +240,11 @@ protected void _fit(Dataframe trainingData) {
     @Override
     public void save(String dbName) {
         initBundle();
-        String knowledgeBaseName = createKnowledgeBaseName(dbName);
-        bundle.save(knowledgeBaseName);
         super.save(dbName);
+
+        String separator = knowledgeBase.getConf().getDbConfig().getDBnameSeparator();
+        String knowledgeBaseName = createKnowledgeBaseName(dbName, separator);
+        bundle.save(knowledgeBaseName, separator);
     }
 
     /** {@inheritDoc} */
@@ -269,13 +271,14 @@ private void initBundle() {
         TrainingParameters trainingParameters = knowledgeBase.getTrainingParameters();
         Configuration conf = knowledgeBase.getConf();
         String dbName = knowledgeBase.getDbc().getDatabaseName();
+        String separator = conf.getDbConfig().getDBnameSeparator();
 
         if(!bundle.containsKey(DT_KEY)) {
             AbstractTransformer.AbstractTrainingParameters dtParams = trainingParameters.getDataTransformerTrainingParameters();
 
             AbstractTransformer dataTransformer = null;
             if(dtParams != null) {
-                dataTransformer = MLBuilder.load(dtParams.getTClass(), dbName + "_" + DT_KEY, conf);
+                dataTransformer = MLBuilder.load(dtParams.getTClass(), dbName + separator + DT_KEY, conf);
             }
             bundle.put(DT_KEY, dataTransformer);
         }
@@ -285,15 +288,15 @@ private void initBundle() {
 
             AbstractFeatureSelector featureSelector = null;
             if(fsParams != null) {
-                featureSelector = MLBuilder.load(fsParams.getTClass(), dbName + "_" + FS_KEY, conf);
+                featureSelector = MLBuilder.load(fsParams.getTClass(), dbName + separator + FS_KEY, conf);
             }
             bundle.put(FS_KEY, featureSelector);
         }
 
         if(!bundle.containsKey(ML_KEY)) {
             AbstractModeler.AbstractTrainingParameters mlParams = trainingParameters.getModelerTrainingParameters();
 
-            bundle.put(ML_KEY, MLBuilder.load(mlParams.getTClass(), dbName + "_" + ML_KEY, conf));
+            bundle.put(ML_KEY, MLBuilder.load(mlParams.getTClass(), dbName + separator + ML_KEY, conf));
         }
     }
 

datumbox-framework-applications/src/main/java/com/datumbox/framework/applications/nlp/CETR.java

@@ -115,19 +115,16 @@ public void setSmoothingAverageRadius(int smoothingAverageRadius) {
             this.smoothingAverageRadius = smoothingAverageRadius;
         }
     }
-
-    private final String dbName;
+
     private final Configuration conf;
 
     /**
      * Constructor for the CETR class. It accepts as arguments the name of the
      * database were the temporary results are stored and the Database Configuration.
-     * 
-     * @param dbName
+     *
      * @param conf 
      */
-    public CETR(String dbName, Configuration conf) {
-        this.dbName = dbName;
+    public CETR(Configuration conf) {
         this.conf = conf;
     }
 

datumbox-framework-applications/src/test/java/com/datumbox/framework/applications/nlp/CETRTest.java

@@ -44,8 +44,6 @@ public void testExtract() {
 
         Configuration conf = Configuration.getConfiguration();
 
-        String dbName = this.getClass().getSimpleName();
-
         String text;        
         try {
             List<String> lines = Files.readAllLines(Paths.get(this.getClass().getClassLoader().getResource("datasets/example.com.html").toURI()), StandardCharsets.UTF_8);
@@ -64,7 +62,7 @@ public void testExtract() {
         parameters.setNumberOfClusters(2);
         parameters.setAlphaWindowSizeFor2DModel(3);
         parameters.setSmoothingAverageRadius(2);
-        CETR instance = new CETR(dbName, conf);
+        CETR instance = new CETR(conf);
         String expResult = "This domain is established to be used for illustrative examples in documents. You may use this domain in examples without prior coordination or asking for permission.";
         String result = instance.extract(text, parameters);
         assertEquals(expResult, result);