Improving file size of write output when integer values are encounter…

…ed by dropping the ".0" that is normally written from a double
EdwardRaff · Dec 27, 2015 · 3c0d235 · 3c0d235
1 parent 84685b5
commit 3c0d235
Show file tree

Hide file tree

Showing 3 changed files with 25 additions and 4 deletions.
diff --git a/JSAT/src/jsat/ARFFLoader.java b/JSAT/src/jsat/ARFFLoader.java
@@ -252,7 +252,11 @@ public static void writeArffFile(DataSet data, OutputStream os, String relation)
                 if(!firstFeature)
                     writer.write(",");
                 firstFeature = false;
-                writer.write(Double.toString(v.get(i)));
+                double val = v.get(i);
+                if(Math.rint(val) == val)//cast to long before writting to save space
+                    writer.write(Long.toString((long) val));
+                else
+                    writer.write(Double.toString(val));
             }
             if (data instanceof ClassificationDataSet)//also write out class variable
             {

diff --git a/JSAT/src/jsat/io/CSV.java b/JSAT/src/jsat/io/CSV.java
@@ -687,7 +687,12 @@ else if(data instanceof RegressionDataSet)
             {
                 if(!nothingWrittenYet)
                     writer.write(delimiter);
-                writer.write(Double.toString(v.get(j)));
+
+                double val = v.get(j);
+                if(Math.rint(val) == val)//cast to long before writting to save space
+                    writer.write(Long.toString((long) val));
+                else
+                    writer.write(Double.toString(val));
                 nothingWrittenYet = false;
             }
             //then categorical features, useing the safe names we constructed earlier

diff --git a/JSAT/src/jsat/io/LIBSVMLoader.java b/JSAT/src/jsat/io/LIBSVMLoader.java
@@ -479,7 +479,13 @@ public static void write(ClassificationDataSet data, OutputStream os)
             Vec vals = data.getDataPoint(i).getNumericalValues();
             writer.write(pred + " ");
             for(IndexValue iv : vals)
-                writer.write((iv.getIndex()+1) + ":" + iv.getValue() + " ");//+1 b/c 1 based indexing
+            {
+                double val = iv.getValue();
+                if(Math.rint(val) == val)//cast to long before writting to save space
+                    writer.write((iv.getIndex()+1) + ":" + (long)val + " ");//+1 b/c 1 based indexing
+                else
+                    writer.write((iv.getIndex()+1) + ":" + val + " ");//+1 b/c 1 based indexing
+            }
             writer.write("\n");
         }
         writer.flush();
@@ -501,7 +507,13 @@ public static void write(RegressionDataSet data, OutputStream os)
             Vec vals = data.getDataPoint(i).getNumericalValues();
             writer.write(pred + " ");
             for(IndexValue iv : vals)
-                writer.write((iv.getIndex()+1) + ":" + iv.getValue() + " ");//+1 b/c 1 based indexing
+            {
+                double val = iv.getValue();
+                if(Math.rint(val) == val)//cast to long before writting to save space
+                    writer.write((iv.getIndex()+1) + ":" + (long)val + " ");//+1 b/c 1 based indexing
+                else
+                    writer.write((iv.getIndex()+1) + ":" + val + " ");//+1 b/c 1 based indexing
+            }
             writer.write("\n");
         }
         writer.flush();