Added simple WCSA

[SXSI/TextCollection.git] / TextCollectionBuilder.h

diff --git a/TextCollectionBuilder.h b/TextCollectionBuilder.h
index 13734d2..6b3819a 100644 (file)
--- a/TextCollectionBuilder.h
+++ b/TextCollectionBuilder.h
@@ -22,7 +22,10 @@
 #define _SXSI_TextCollectionBuilder_h_
 
 #include "TextCollection.h"
+#include "TextStorage.h"
 #include "Tools.h" // Defines ulong and uchar.
+
+#include <string>
 #include <vector>
 #include <utility> // Defines std::pair.
 #include <cstring> // Defines std::strlen, added by Kim
@@ -33,6 +36,8 @@
 // Default samplerate for suffix array samples
 #define TEXTCOLLECTION_DEFAULT_SAMPLERATE 64
 
+// Default input length, used to calculate the buffer size.
+#define TEXTCOLLECTION_DEFAULT_INPUT_LENGTH (150 * 1024 * 1024)
 
 
 namespace SXSI
@@ -45,7 +50,8 @@ namespace SXSI
     class TextCollectionBuilder
     {
     public:
-        explicit TextCollectionBuilder(unsigned samplerate = TEXTCOLLECTION_DEFAULT_SAMPLERATE);
+        explicit TextCollectionBuilder(unsigned samplerate = TEXTCOLLECTION_DEFAULT_SAMPLERATE, 
+                                       ulong estimatedInputLength =  TEXTCOLLECTION_DEFAULT_INPUT_LENGTH);
         ~TextCollectionBuilder();
         
         /** 
@@ -55,17 +61,25 @@ namespace SXSI
          * Can not be called after makeStatic().
          * The i'th text insertion gets an identifier value i-1.
          * In other words, document identifiers start from 0.
+         *
+         * Second parameter tells if the text will be added to the
+         * index also. If false, text is added only to the TextCollection
+         * and can not be searched for.
          */
-        void InsertText(uchar const *);
+        void InsertText(uchar const *, bool index = true);
         /**
          * Make static
          *
-         * Convert to a static collection; reduces space and time complexities.
+         * Convert to a static collection.
          * New texts can not be inserted after this operation.
+         *
+         * TextStorage type defaults to TYPE_PLAIN_TEXT, another
+         * possible type is TYPE_LZ_INDEX.
          */
-        TextCollection * InitTextCollection();
+        TextCollection * InitTextCollection(char type = TextStorage::TYPE_PLAIN_TEXT);
         
     private:
+        // Using Pimpl idiom to hide RLCSA implementation.
         struct TCBuilderRep * p_;
 
         // No copy constructor or assignment