X-Git-Url: http://git.nguyen.vg/gitweb/?a=blobdiff_plain;f=TextCollectionBuilder.h;h=422bd224a5b7d4922f7c619918864a5a484143fb;hb=e4d0188a098767c4950ba8936b754922343cbe8b;hp=bcb3672515b613179f7cb58b580837720ff0b0cc;hpb=ee8a3e526fe7f39cdc075263824faf6c17389297;p=SXSI%2FTextCollection.git

diff --git a/TextCollectionBuilder.h b/TextCollectionBuilder.h
index bcb3672..422bd22 100644
--- a/TextCollectionBuilder.h
+++ b/TextCollectionBuilder.h
@@ -22,23 +22,35 @@
 #define _SXSI_TextCollectionBuilder_h_
 
 #include "TextCollection.h"
-#include "Tools.h" // Defines ulong and uchar.
-#include <vector>
-#include <utility> // Defines std::pair.
-#include <cstring> // Defines std::strlen, added by Kim
+#include "TextStorage.h"
+#include "Tools.h"
+
+// Default samplerate for suffix array samples
+#define TEXTCOLLECTION_DEFAULT_SAMPLERATE 32
+
+// Default input length, used to calculate the buffer size.
+#define TEXTCOLLECTION_DEFAULT_INPUT_LENGTH (150 * 1024 * 1024)
+
 
 namespace SXSI
 {
-    struct TCBuilderRep; // Pimpl
-    
     /**
-     * Build an instance of the TextCollection class.
+     * Builder for an instance of the TextCollection class.
      */
     class TextCollectionBuilder
     {
     public:
-        explicit TextCollectionBuilder(unsigned);
-        ~TextCollectionBuilder();
+        // Index type defaults to FM-index.
+        // SWCSA can be used for natural language inputs.
+        // NB: Current SWCSA uses a lot of memory during construction!
+        enum index_type_t { index_type_default, index_type_swcsa, index_type_rlcsa }; 
+
+        static TextCollectionBuilder* create(unsigned samplerate = TEXTCOLLECTION_DEFAULT_SAMPLERATE, 
+                                             index_type_t type = index_type_default,
+                                             ulong estimatedInputLength =  TEXTCOLLECTION_DEFAULT_INPUT_LENGTH);
+        
+
+        virtual ~TextCollectionBuilder() { };
         
         /** 
          * Insert text
@@ -47,23 +59,33 @@ 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 *);
+        virtual void InsertText(uchar const *, bool index = true) = 0;
         /**
          * 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();
+        virtual TextCollection * InitTextCollection(char type = TextStorage::TYPE_PLAIN_TEXT) = 0;
         
-    private:
-        struct TCBuilderRep * p_;
+    protected:
+        // Protected constructor; use the static method TextCollectionBuilder::create()
+        TextCollectionBuilder() { };
 
+    private:
         // No copy constructor or assignment
-        TextCollectionBuilder();
         TextCollectionBuilder(TextCollectionBuilder const&);
         TextCollectionBuilder& operator = (TextCollectionBuilder const&);
     };
+
 }
+
 #endif