Lift docListMerge() call out of loadSegmentLeavesInt() for prefix

search.  Doclists from multiple prefix matches will need a union merge
function, which will have to logically happen across a segment before
doclists are merged between segments. (CVS 3887)

FossilOrigin-Name: 7ddb82668906e33e2d6a796f2da1795032e036d5

diff --git a/ext/fts2/fts2.c b/ext/fts2/fts2.c
index b13f252d627d66f534536820dd2197f473cf91ef..f6ae290badf96dc71dad01d5c70cfb74f183bf2a 100644 (file)
--- a/ext/fts2/fts2.c
+++ b/ext/fts2/fts2.c
@@ -5116,21 +5116,9 @@ static int loadSegmentLeavesInt(fulltext_vtab *v, LeavesReader *pReader,
     if( c==0 ){
       const char *pData = leavesReaderData(pReader);
       int nData = leavesReaderDataBytes(pReader);
-      if( out->nData==0 ){
-        dataBufferReplace(out, pData, nData);
-      }else{
-        DLReader readers[2];
-        DataBuffer result;
-        dlrInit(&readers[0], DL_DEFAULT, out->pData, out->nData);
-        dlrInit(&readers[1], DL_DEFAULT, pData, nData);
-        dataBufferInit(&result, out->nData+nData);
-        docListMerge(&result, readers, 2);
-        dataBufferDestroy(out);
-        *out = result;
-        dlrDestroy(&readers[0]);
-        dlrDestroy(&readers[1]);
-      }
-     }
+      assert( out->nData==0 );
+      dataBufferReplace(out, pData, nData);
+    }
     if( c>=0 ) break;      /* Past any possible matches. */
 
     rc = leavesReaderStep(v, pReader);
@@ -5236,18 +5224,12 @@ static int loadAndGetChildContaining(fulltext_vtab *v, sqlite_int64 iBlockid,
 }
 
 /* Traverse the tree represented by pData[nData] looking for
-** pTerm[nTerm], merging its doclist over *out if found (any duplicate
-** doclists read from the segment rooted at pData will overwrite those
-** in *out).
+** pTerm[nTerm], placing its doclist into *out.  This is internal to
+** loadSegment() to make error-handling cleaner.
 */
-static int loadSegment(fulltext_vtab *v, const char *pData, int nData,
-                       sqlite_int64 iLeavesEnd,
-                       const char *pTerm, int nTerm, DataBuffer *out){
-  assert( nData>1 );
-
-  /* This code should never be called with buffered updates. */
-  assert( v->nPendingData<0 );
-
+static int loadSegmentInt(fulltext_vtab *v, const char *pData, int nData,
+                          sqlite_int64 iLeavesEnd,
+                          const char *pTerm, int nTerm, DataBuffer *out){
   /* Special case where root is a leaf. */
   if( *pData=='\0' ){
     return loadSegmentLeaf(v, pData, nData, pTerm, nTerm, out);
@@ -5268,6 +5250,50 @@ static int loadSegment(fulltext_vtab *v, const char *pData, int nData,
   }
 }
 
+/* Call loadSegmentInt() to collect the doclist for pTerm/nTerm, then
+** merge its doclist over *out (any duplicate doclists read from the
+** segment rooted at pData will overwrite those in *out).
+*/
+/* NOTE(shess) Previous code passed out down to sub-routines for use
+** in docListMerge().  This version deoptimizes things slightly, but
+** prefix searches require a different merge function entirely.
+*/
+static int loadSegment(fulltext_vtab *v, const char *pData, int nData,
+                       sqlite_int64 iLeavesEnd,
+                       const char *pTerm, int nTerm, DataBuffer *out){
+  DataBuffer result;
+  int rc;
+
+  assert( nData>1 );
+
+  /* This code should never be called with buffered updates. */
+  assert( v->nPendingData<0 );
+
+  dataBufferInit(&result, 0);
+  rc = loadSegmentInt(v, pData, nData, iLeavesEnd, pTerm, nTerm, &result);
+  if( rc==SQLITE_OK && result.nData>0 ){
+    if( out->nData==0 ){
+      DataBuffer tmp = *out;
+      *out = result;
+      result = tmp;
+    }else{
+      DataBuffer merged;
+      DLReader readers[2];
+
+      dlrInit(&readers[0], DL_DEFAULT, out->pData, out->nData);
+      dlrInit(&readers[1], DL_DEFAULT, result.pData, result.nData);
+      dataBufferInit(&merged, out->nData+result.nData);
+      docListMerge(&merged, readers, 2);
+      dataBufferDestroy(out);
+      *out = merged;
+      dlrDestroy(&readers[0]);
+      dlrDestroy(&readers[1]);
+    }
+  }
+  dataBufferDestroy(&result);
+  return rc;
+}
+
 /* Scan the database and merge together the posting lists for the term
 ** into *out.
 */

diff --git a/manifest b/manifest
index 4ee4f13320524909fa0956e34bf15b8640a6d09e..b54e84e36c93f1f8ef6abc08545ce1bbff0fb010 100644 (file)
--- a/manifest
+++ b/manifest
@@ -1,5 +1,5 @@
-C Try\sto\savoid\sreading\spages\swhen\smoving\soverflow\schains\sto\sthe\sfree-list.\s(CVS\s3886)
-D 2007-04-30T16:55:01
+C Lift\sdocListMerge()\scall\sout\sof\sloadSegmentLeavesInt()\sfor\sprefix\nsearch.\s\sDoclists\sfrom\smultiple\sprefix\smatches\swill\sneed\sa\sunion\smerge\nfunction,\swhich\swill\shave\sto\slogically\shappen\sacross\sa\ssegment\sbefore\ndoclists\sare\smerged\sbetween\ssegments.\s(CVS\s3887)
+D 2007-04-30T17:52:52
 F Makefile.in 8cab54f7c9f5af8f22fd97ddf1ecfd1e1860de62
 F Makefile.linux-gcc 2d8574d1ba75f129aba2019f0b959db380a90935
 F README 9c4e2d6706bdcc3efdd773ce752a8cdab4f90028
@@ -34,7 +34,7 @@ F ext/fts1/fulltext.h 08525a47852d1d62a0be81d3fc3fe2d23b094efd
 F ext/fts1/simple_tokenizer.c 1844d72f7194c3fd3d7e4173053911bf0661b70d
 F ext/fts1/tokenizer.h 0c53421b832366d20d720d21ea3e1f6e66a36ef9
 F ext/fts2/README.txt 8c18f41574404623b76917b9da66fcb0ab38328d
-F ext/fts2/fts2.c b45e07a23646e82525657af9bab063e10f9acffc
+F ext/fts2/fts2.c cb7ca4e320f9aa99dad1d160288944c13778a678
 F ext/fts2/fts2.h 591916a822cfb6426518fdbf6069359119bc46eb
 F ext/fts2/fts2_hash.c b3f22116d4ef0bc8f2da6e3fdc435c86d0951a9b
 F ext/fts2/fts2_hash.h e283308156018329f042816eb09334df714e105e
@@ -465,7 +465,7 @@ F www/tclsqlite.tcl bb0d1357328a42b1993d78573e587c6dcbc964b9
 F www/vdbe.tcl 87a31ace769f20d3627a64fa1fade7fed47b90d0
 F www/version3.tcl 890248cf7b70e60c383b0e84d77d5132b3ead42b
 F www/whentouse.tcl fc46eae081251c3c181bd79c5faef8195d7991a5
-P 89b1b3f897bda1fffceb9cf72fa4d42b809ccb8e
-R 25da30ea21a3fd3e2f878c36734a5bb9
-U danielk1977
-Z f3a814360f6db57fb203458a6b31fb7b
+P 8cccec68bd9073b2b19d3d31cf0b77b0ce76172e
+R 7069672da6b54cde9af80d1ef9e46049
+U shess
+Z 8a2908873a8cd6c0a742806cb3215c42

diff --git a/manifest.uuid b/manifest.uuid
index a39bd2cabe0e83b5df802fac05a756f7de531212..c9a7edcd4a767eeaba3b0662c3f7736e17fac8f1 100644 (file)
--- a/manifest.uuid
+++ b/manifest.uuid
@@ -1 +1 @@
-8cccec68bd9073b2b19d3d31cf0b77b0ce76172e
\ No newline at end of file
+7ddb82668906e33e2d6a796f2da1795032e036d5
\ No newline at end of file