From bcbed10c547780b6e5b2028d936eae337ecebac5 Mon Sep 17 00:00:00 2001
From: nvalimak <nvalimak@3cdefd35-fc62-479d-8e8d-bae585ffb9ca>
Date: Fri, 8 May 2009 12:03:31 +0000
Subject: [PATCH] Added TextStorage class

git-svn-id: svn+ssh://idea.nguyen.vg/svn/sxsi/trunk/TextCollection@375 3cdefd35-fc62-479d-8e8d-bae585ffb9ca
---
 TCImplementation.cpp | 21 ++++++++++++++++-----
 TCImplementation.h   | 39 +++++++++++++++++++++++++++++++++------
 2 files changed, 49 insertions(+), 11 deletions(-)

diff --git a/TCImplementation.cpp b/TCImplementation.cpp
index a8303d8..5959e6a 100644
--- a/TCImplementation.cpp
+++ b/TCImplementation.cpp
@@ -19,6 +19,7 @@
  *****************************************************************************/
 #include "TCImplementation.h"
 
+//#define DEBUG_MEMUSAGE
 #ifdef DEBUG_MEMUSAGE
 #include "HeapProfiler.h" // FIXME remove
 #endif
@@ -40,7 +41,7 @@ namespace SXSI
 {
 
 // Save file version info
-const uchar TCImplementation::versionFlag = 4;
+const uchar TCImplementation::versionFlag = 6;
 
 /**
  * Constructor inits an empty dynamic FM-index.
@@ -63,10 +64,12 @@ bool TCImplementation::EmptyText(DocId k) const
     return false; // Empty texts are not indexed
 }
 
-uchar* TCImplementation::GetText(DocId k) const
+uchar * TCImplementation::GetText(DocId k) const
 {
     assert(k < (DocId)numberOfTexts);
-    TextPosition i = k;
+
+    return textStorage->GetText(k);
+/*    TextPosition i = k;
 
     string result;
     // Reserve average string length to avoid reallocs
@@ -87,7 +90,7 @@ uchar* TCImplementation::GetText(DocId k) const
     res[i] = '\0';   
     for (ulong j = 0; j < i; ++j)
         res[i-j-1] = result[j];
-    return res;
+        return res;*/
 }
 /*
  * Not supported
@@ -769,6 +772,7 @@ void TCImplementation::Save(FILE *file) const
         throw std::runtime_error("TCImplementation::Save(): file write error (maxTextLength).");
 
     Doc->save(file);
+    textStorage->Save(file);
     fflush(file);
 }
 
@@ -815,6 +819,7 @@ TCImplementation::TCImplementation(FILE *file, unsigned samplerate_)
         throw std::runtime_error("TCImplementation::Load(): file read error (maxTextLength).");
 
     Doc = static_sequence::load(file);
+    textStorage = new TextStorage(file);
 
     // FIXME Construct data structures with new samplerate
     //maketables(); 
@@ -906,6 +911,7 @@ TCImplementation::~TCImplementation() {
     delete suffixes;
     delete suffixDocId;
     delete Doc;
+    delete textStorage;
 }
 
 void TCImplementation::makewavelet(uchar *bwt)
@@ -1036,6 +1042,8 @@ void TCImplementation::maketables()
     p=bwtEndPos;
     textId = numberOfTexts;
 
+    TextStorageBuilder tsbuilder(n);
+
     /**
      * Second pass: populate tables suffixes and suffixDocId.
      */
@@ -1052,6 +1060,8 @@ void TCImplementation::maketables()
         }
 
         uchar c = alphabetrank->access(p, alphabetrank_i_tmp);
+        tsbuilder[i] = c;
+
         if (c == '\0')
         {
             --textId;
@@ -1062,9 +1072,10 @@ void TCImplementation::maketables()
             p = C[c]+alphabetrank_i_tmp-1;
     }
     assert(textId == 0);
-
     delete textStartPos;
 
+    textStorage = tsbuilder.InitTextStorage();
+
 #ifdef DEBUG_MEMUSAGE
     std::cerr << "max heap usage before Doc: " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes" << std::endl;
     HeapProfiler::ResetMaxHeapConsumption();
diff --git a/TCImplementation.h b/TCImplementation.h
index 5430f21..0439978 100644
--- a/TCImplementation.h
+++ b/TCImplementation.h
@@ -40,6 +40,9 @@
 #endif
 #undef bitset
 
+#include "TextStorage.h"
+
+
 namespace SXSI 
 {
 
@@ -54,11 +57,32 @@ public:
     ~TCImplementation();
 
     bool EmptyText(DocId) const;
-    uchar* GetText(DocId) const;
+
+    /**
+     * Returns a pointer to the original text.
+     *
+     * Do *not* try to free the array. 
+     * (However, this implementation is suspect to change.)
+     *
+     * See TextStorage.h for details.
+     */
+    uchar * GetText(DocId) const;
+
     /**
-     * Next method is not supported:
-     * Supporting GetText for some substring [i,j]
-     * would require more space.
+     * Returns a pointer to the beginning of texts i, i+1, ..., j.
+     * Texts are separated by a '\0' byte.
+     *
+     * Do *not* try to free the array. 
+     * (However, this implementation is suspect to change.)
+     *
+     * See TextStorage.h for details.
+     */
+    uchar * GetText(DocId i, DocId j) const;
+
+    /**
+     * Returns a substring of given text ID.
+     * 
+     * FIXME This may be reimplemented via TextStorage.
      */
 //    uchar* GetText(DocId, TextPosition, TextPosition) const;
 
@@ -117,7 +141,7 @@ private:
     static_sequence * alphabetrank;
 
     // Sample structures for texts longer than samplerate
-    static_bitsequence * sampled;  // FIXME Replace with RRR02
+    static_bitsequence * sampled;
     BlockArray *suffixes;
     BlockArray *suffixDocId;
 
@@ -129,7 +153,10 @@ private:
     // Array of document id's in the order of end-markers in BWT
     static_sequence *Doc;
 
-    // Following are not part of the public API
+    // Text storage for fast extraction
+    TextStorage * textStorage;
+
+    // Following methods are not part of the public API
     uchar * BWT(uchar *);
     void makewavelet(uchar *);
     void maketables();
-- 
2.17.1