From: drh <drh@noemail.net>
Date: Sun, 28 Apr 2019 19:27:02 +0000 (+0000)
Subject: Take collating sequence into account when removing redundant columns from
X-Git-Tag: version-3.29.0~173^2~1
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=f78d0f426c2ce84904823a087e3950e343c8e6ec;p=thirdparty%2Fsqlite.git

Take collating sequence into account when removing redundant columns from
indexes on WITHOUT ROWID tables.  This is the first proof-of-concept fix
for ticket [3182d3879020ef3]. More testing needed.

FossilOrigin-Name: b34fa5bff40d3d364bd8c80e7de55c606ef3caac47b14b5265ebcb38857eb85e
---

diff --git a/manifest b/manifest
index ffbce1300f..dec5885b95 100644
--- a/manifest
+++ b/manifest
@@ -1,5 +1,5 @@
-C Add\sthe\s".recover"\scommand\sto\sthe\sshell\stool.\sFor\srecovering\sas\smuch\sdata\sas\spossible\sfrom\scorrupt\sdatabases.
-D 2019-04-27T20:30:19.423
+C Take\scollating\ssequence\sinto\saccount\swhen\sremoving\sredundant\scolumns\sfrom\nindexes\son\sWITHOUT\sROWID\stables.\s\sThis\sis\sthe\sfirst\sproof-of-concept\sfix\nfor\sticket\s[3182d3879020ef3].\sMore\stesting\sneeded.
+D 2019-04-28T19:27:02.781
 F .fossil-settings/empty-dirs dbb81e8fc0401ac46a1491ab34a7f2c7c0452f2f06b54ebb845d024ca8283ef1
 F .fossil-settings/ignore-glob 35175cdfcf539b2318cb04a9901442804be81cd677d8b889fcc9149c21f239ea
 F LICENSE.md df5091916dbb40e6e9686186587125e1b2ff51f022cc334e886c19a0e9982724
@@ -463,7 +463,7 @@ F src/btmutex.c 8acc2f464ee76324bf13310df5692a262b801808984c1b79defb2503bbafadb6
 F src/btree.c ffe7101006aee2ab9e9dec2fc001998e57a8e59419c6ea4072d6c3935d3d50fb
 F src/btree.h c11446f07ec0e9dc85af8041cb0855c52f5359c8b2a43e47e02a685282504d89
 F src/btreeInt.h 6111c15868b90669f79081039d19e7ea8674013f907710baa3c814dc3f8bfd3f
-F src/build.c e9d560fdc39e0f037b1ebd78fde0ff616963646c8d85a7afa18db064c52c5b75
+F src/build.c 3fedab95ff7a9b3ed51eceb12aaa61de6bbd063dabd276767494cc448a151b7b
 F src/callback.c 25dda5e1c2334a367b94a64077b1d06b2553369f616261ca6783c48bcb6bda73
 F src/complete.c a3634ab1e687055cd002e11b8f43eb75c17da23e
 F src/ctime.c 109e58d00f62e8e71ee1eb5944ac18b90171c928ab2e082e058056e1137cc20b
@@ -1821,8 +1821,10 @@ F vsixtest/vsixtest.tcl 6a9a6ab600c25a91a7acc6293828957a386a8a93
 F vsixtest/vsixtest.vcxproj.data 2ed517e100c66dc455b492e1a33350c1b20fbcdc
 F vsixtest/vsixtest.vcxproj.filters 37e51ffedcdb064aad6ff33b6148725226cd608e
 F vsixtest/vsixtest_TemporaryKey.pfx e5b1b036facdb453873e7084e1cae9102ccc67a0
-P 95209072176ff21a91e96d5bd014b35ef100da2b0b93958baf6df4294a8daa85 425d708c3908fe74f69b62e6dd1722a0018088977e12f14b312dad1df0fbb804
-R f359cf293ea45304e0ddadd16c7a1206
-T +closed 425d708c3908fe74f69b62e6dd1722a0018088977e12f14b312dad1df0fbb804
-U dan
-Z ee7f2d083d1fca64ba3bb33b40a19208
+P 50fe48458942fa7a6bcc76316c6321f95b23dc34f2f8e0a483826483b2fb16f6
+R fa5b9567c2bc0a92e3c8181fab56d7f8
+T *branch * tkt-3182d38790
+T *sym-tkt-3182d38790 *
+T -sym-trunk *
+U drh
+Z dd82847aa38c87860cc4ec05cc3ab230
diff --git a/manifest.uuid b/manifest.uuid
index 55700b6967..6c45322202 100644
--- a/manifest.uuid
+++ b/manifest.uuid
@@ -1 +1 @@
-50fe48458942fa7a6bcc76316c6321f95b23dc34f2f8e0a483826483b2fb16f6
\ No newline at end of file
+b34fa5bff40d3d364bd8c80e7de55c606ef3caac47b14b5265ebcb38857eb85e
\ No newline at end of file
diff --git a/src/build.c b/src/build.c
index 3e4101df92..fac3163eba 100644
--- a/src/build.c
+++ b/src/build.c
@@ -1726,10 +1726,46 @@ static void estimateIndexWidth(Index *pIdx){
   pIdx->szIdxRow = sqlite3LogEst(wIndex*4);
 }
 
-/* Return true if value x is found any of the first nCol entries of aiCol[]
+/* Return true if column number x is any of the first nCol entries of aiCol[].
+** This is used to determine if the column number x appears in any of the
+** first nCol entries of an index.
 */
 static int hasColumn(const i16 *aiCol, int nCol, int x){
-  while( nCol-- > 0 ) if( x==*(aiCol++) ) return 1;
+  while( nCol-- > 0 ){
+    assert( aiCol[0]>=0 );
+    if( x==*(aiCol++) ){
+      return 1;
+    }
+  }
+  return 0;
+}
+
+/*
+** Return true if any of the first nKey entries of index pIdx1 exactly
+** match the iCol-th entry of pIdx2.
+**
+** The first nKey entries of pIdx1 are guaranteed to be ordinary columns,
+** not a rowid or expression.
+**
+** This routine differs from hasColumn() in that both the column and the
+** collating sequence must match for this routine, but for hasColumn() only
+** the column name must match.
+*/
+static int isDupColumn(Index *pIdx1, int nKey, Index *pIdx2, int iCol){
+  int i, j;
+  assert( nKey<=pIdx1->nColumn );
+  assert( iCol<MAX(pIdx2->nColumn,pIdx2->nKeyCol) );
+  j = pIdx2->aiColumn[iCol];
+  testcase( j==XN_EXPR );
+  assert( j!=XN_ROWID );
+  for(i=0; i<nKey; i++){
+    assert( pIdx1->aiColumn[i]>=0 || j>=0 );
+    if( pIdx1->aiColumn[i]==j 
+     && sqlite3StrICmp(pIdx1->azColl[i],pIdx2->azColl[iCol])==0
+    ){
+      return 1;
+    }
+  }
   return 0;
 }
 
@@ -1835,9 +1871,10 @@ static void convertToWithoutRowidTable(Parse *pParse, Table *pTab){
     ** code assumes the PRIMARY KEY contains no repeated columns.
     */
     for(i=j=1; i<pPk->nKeyCol; i++){
-      if( hasColumn(pPk->aiColumn, j, pPk->aiColumn[i]) ){
+      if( isDupColumn(pPk, j, pPk, i) ){
         pPk->nColumn--;
       }else{
+        testcase( hasColumn(pPk->aiColumn, j, pPk->aiColumn[i]) );
         pPk->aiColumn[j++] = pPk->aiColumn[i];
       }
     }
@@ -1867,7 +1904,10 @@ static void convertToWithoutRowidTable(Parse *pParse, Table *pTab){
     int n;
     if( IsPrimaryKeyIndex(pIdx) ) continue;
     for(i=n=0; i<nPk; i++){
-      if( !hasColumn(pIdx->aiColumn, pIdx->nKeyCol, pPk->aiColumn[i]) ) n++;
+      if( !isDupColumn(pIdx, pIdx->nKeyCol, pPk, i) ){
+        testcase( hasColumn(pIdx->aiColumn, pIdx->nKeyCol, pPk->aiColumn[i]) );
+        n++;
+      }
     }
     if( n==0 ){
       /* This index is a superset of the primary key */
@@ -1876,7 +1916,8 @@ static void convertToWithoutRowidTable(Parse *pParse, Table *pTab){
     }
     if( resizeIndexObject(db, pIdx, pIdx->nKeyCol+n) ) return;
     for(i=0, j=pIdx->nKeyCol; i<nPk; i++){
-      if( !hasColumn(pIdx->aiColumn, pIdx->nKeyCol, pPk->aiColumn[i]) ){
+      if( !isDupColumn(pIdx, pIdx->nKeyCol, pPk, i) ){
+        testcase( hasColumn(pIdx->aiColumn, pIdx->nKeyCol, pPk->aiColumn[i]) );
         pIdx->aiColumn[j] = pPk->aiColumn[i];
         pIdx->azColl[j] = pPk->azColl[i];
         j++;
@@ -3392,9 +3433,10 @@ void sqlite3CreateIndex(
     for(j=0; j<pPk->nKeyCol; j++){
       int x = pPk->aiColumn[j];
       assert( x>=0 );
-      if( hasColumn(pIndex->aiColumn, pIndex->nKeyCol, x) ){
+      if( isDupColumn(pIndex, pIndex->nKeyCol, pPk, j) ){
         pIndex->nColumn--; 
       }else{
+        testcase( hasColumn(pIndex->aiColumn,pIndex->nKeyCol,x) );
         pIndex->aiColumn[i] = x;
         pIndex->azColl[i] = pPk->azColl[j];
         pIndex->aSortOrder[i] = pPk->aSortOrder[j];