SQLite

**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains code for implementations of the r-tree and r*-tree
** algorithms packaged as an SQLite virtual table module.
*/

/*
** Database Format of R-Tree Tables
** --------------------------------
**
** The data structure for a single virtual r-tree table is stored in three 
** native SQLite tables declared as follows. In each case, the '%' character
** in the table name is replaced with the user-supplied name of the r-tree
** table.
**
**   CREATE TABLE %_node(nodeno INTEGER PRIMARY KEY, data BLOB)
**   CREATE TABLE %_parent(nodeno INTEGER PRIMARY KEY, parentnode INTEGER)
**   CREATE TABLE %_rowid(rowid INTEGER PRIMARY KEY, nodeno INTEGER)
**
** The data for each node of the r-tree structure is stored in the %_node
** table. For each node that is not the root node of the r-tree, there is
** an entry in the %_parent table associating the node with its parent.
** And for each row of data in the table, there is an entry in the %_rowid
** table that maps from the entries rowid to the id of the node that it
** is stored on.
**
** The root node of an r-tree always exists, even if the r-tree table is
** empty. The nodeno of the root node is always 1. All other nodes in the
** table must be the same size as the root node. The content of each node
** is formatted as follows:
**
**   1. If the node is the root node (node 1), then the first 2 bytes
**      of the node contain the tree depth as a big-endian integer.
**      For non-root nodes, the first 2 bytes are left unused.
**
**   2. The next 2 bytes contain the number of entries currently 
**      stored in the node.
**
**   3. The remainder of the node contains the node entries. Each entry
**      consists of a single 8-byte integer followed by an even number
**      of 4-byte coordinates. For leaf nodes the integer is the rowid
**      of a record. For internal nodes it is the node number of a
**      child page.
*/

#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_RTREE)

/*
** This file contains an implementation of a couple of different variants
** of the r-tree algorithm. See the README file for further details. The 
** same data-structure is used for all, but the algorithms for insert and

** If an R*-tree "Reinsert" operation is required, the same number of
** cells are removed from the overfull node and reinserted into the tree.
*/
#define RTREE_MINCELLS(p) ((((p)->iNodeSize-4)/(p)->nBytesPerCell)/3)
#define RTREE_REINSERT(p) RTREE_MINCELLS(p)
#define RTREE_MAXCELLS 51

/*
** The smallest possible node-size is (512-64)==448 bytes. And the largest
** supported cell size is 48 bytes (8 byte rowid + ten 4 byte coordinates).
** Therefore all non-root nodes must contain at least 3 entries. Since 
** 2^40 is greater than 2^64, an r-tree structure always has a depth of
** 40 or less.
*/
#define RTREE_MAX_DEPTH 40

/* 
** An rtree cursor object.
*/
struct RtreeCursor {
  sqlite3_vtab_cursor base;
  RtreeNode *pNode;                 /* Node cursor is currently pointing at */
  int iCell;                        /* Index of current cell in pNode */

#define RTREE_LT    0x43
#define RTREE_GE    0x44
#define RTREE_GT    0x45
#define RTREE_MATCH 0x46

/* 
** An rtree structure node.















*/
struct RtreeNode {
  RtreeNode *pParent;               /* Parent node */
  i64 iNode;
  int nRef;
  int isDirty;
  u8 *zData;

/*
** Search the node hash table for node iNode. If found, return a pointer
** to it. Otherwise, return 0.
*/
static RtreeNode *nodeHashLookup(Rtree *pRtree, i64 iNode){
  RtreeNode *p;

  for(p=pRtree->aHash[nodeHash(iNode)]; p && p->iNode!=iNode; p=p->pNext);
  return p;
}

/*
** Add node pNode to the node hash table.
*/

      pNode->pParent = pParent;
    }
    pNode->nRef++;
    *ppNode = pNode;
    return SQLITE_OK;
  }

  sqlite3_bind_int64(pRtree->pReadNode, 1, iNode);
  rc = sqlite3_step(pRtree->pReadNode);
  if( rc==SQLITE_ROW ){
    const u8 *zBlob = sqlite3_column_blob(pRtree->pReadNode, 0);
    if( pRtree->iNodeSize==sqlite3_column_bytes(pRtree->pReadNode, 0) ){
      pNode = (RtreeNode *)sqlite3_malloc(sizeof(RtreeNode)+pRtree->iNodeSize);
      if( !pNode ){

        rc = SQLITE_NOMEM;
      }else{
        pNode->pParent = pParent;
        pNode->zData = (u8 *)&pNode[1];
        pNode->nRef = 1;
        pNode->iNode = iNode;
        pNode->isDirty = 0;
        pNode->pNext = 0;






        memcpy(pNode->zData, zBlob, pRtree->iNodeSize);
        nodeReference(pParent);



      }
    }

  }
  rc = sqlite3_reset(pRtree->pReadNode);

  /* If the root node was just loaded, set pRtree->iDepth to the height
  ** of the r-tree structure. A height of zero means all data is stored on
  ** the root node. A height of one means the children of the root node
  ** are the leaves, and so on. If the depth as specified on the root node
  ** is greater than RTREE_MAX_DEPTH, the r-tree structure must be corrupt.
  */
  if( pNode && iNode==1 ){
    pRtree->iDepth = readInt16(pNode->zData);
    if( pRtree->iDepth>RTREE_MAX_DEPTH ){
      rc = SQLITE_CORRUPT;
    }
  }

  /* If no error has occurred so far, check if the "number of entries"
  ** field on the node is too large. If so, set the return code to 
  ** SQLITE_CORRUPT.
  */
  if( pNode && rc==SQLITE_OK ){
    if( NCELL(pNode)>((pRtree->iNodeSize-4)/pRtree->nBytesPerCell) ){
      rc = SQLITE_CORRUPT;
    }
  }

  if( rc==SQLITE_OK ){
    if( pNode!=0 ){
      nodeHashInsert(pRtree, pNode);
    }else{
      rc = SQLITE_CORRUPT;
    }
    *ppNode = pNode;
  }else{
    sqlite3_free(pNode);
    *ppNode = 0;
  }

  return rc;
}

/*
** Overwrite cell iCell of node pNode with the contents of pCell.

){
  int nCell;                    /* Current number of cells in pNode */
  int nMaxCell;                 /* Maximum number of cells for pNode */

  nMaxCell = (pRtree->iNodeSize-4)/pRtree->nBytesPerCell;
  nCell = NCELL(pNode);

  assert( nCell<=nMaxCell );

  if( nCell<nMaxCell ){
    nodeOverwriteCell(pRtree, pNode, pCell, nCell);
    writeInt16(&pNode->zData[2], nCell+1);
    pNode->isDirty = 1;
  }

  return (nCell==nMaxCell);

){
  Rtree *pRtree = (Rtree *)pVtab;
  int rc = SQLITE_OK;

  rtreeReference(pRtree);

  assert(nData>=1);




  /* If azData[0] is not an SQL NULL value, it is the rowid of a
  ** record to delete from the r-tree table. The following block does
  ** just that.
  */
  if( sqlite3_value_type(azData[0])!=SQLITE_NULL ){
    i64 iDelete;                /* The rowid to delete */

# 2010 September 22
#
# The author disclaims copyright to this source code.  In place of
# a legal notice, here is a blessing:
#
#    May you do good and not evil.
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
# This file contains tests for the r-tree module. Specifically, it tests
# that corrupt or inconsistent databases do not cause crashes in the r-tree
# module.
# 

if {![info exists testdir]} {
  set testdir [file join [file dirname [info script]] .. .. test]
} 
source $testdir/tester.tcl
ifcapable !rtree { finish_test ; return }

proc create_t1 {} {
  db close
  forcedelete test.db
  sqlite3 db test.db
  execsql {
    CREATE VIRTUAL TABLE t1 USING rtree(id, x1, x2, y1, y2);
  }
}
proc populate_t1 {} {
  execsql BEGIN
  for {set i 0} {$i < 500} {incr i} {
    set x2 [expr $i+5]
    set y2 [expr $i+5]
    execsql { INSERT INTO t1 VALUES($i, $i, $x2, $i, $y2) }
  }
  execsql COMMIT
}

proc truncate_node {nodeno nTrunc} {
  set blob [db one {SELECT data FROM t1_node WHERE nodeno=$nodeno}]
  if {$nTrunc<0} {set nTrunc "end-$nTrunc"}
  set blob [string range $blob 0 $nTrunc]
  db eval { UPDATE t1_node SET data = $blob WHERE nodeno=$nodeno }
}

proc set_tree_depth {tbl {newvalue ""}} {
  set blob [db one "SELECT data FROM ${tbl}_node WHERE nodeno=1"]

  if {$newvalue == ""} {
    binary scan $blob Su oldvalue
    return $oldvalue
  }

  set blob [binary format Sua* $newvalue [string range $blob 2 end]]
  db eval "UPDATE ${tbl}_node SET data = \$blob WHERE nodeno=1"
  return [set_tree_depth $tbl]
}

proc set_entry_count {tbl nodeno {newvalue ""}} {
  set blob [db one "SELECT data FROM ${tbl}_node WHERE nodeno=$nodeno"]

  if {$newvalue == ""} {
    binary scan [string range $blob 2 end] Su oldvalue
    return $oldvalue
  }

  set blob [binary format a*Sua* \
    [string range $blob 0 1] $newvalue [string range $blob 4 end]
  ]
  db eval "UPDATE ${tbl}_node SET data = \$blob WHERE nodeno=$nodeno"
  return [set_entry_count $tbl $nodeno]
}



proc do_corruption_tests {prefix args} {
  set testarray [lindex $args end]
  set errormsg {database disk image is malformed}

  foreach {z value} [lrange $args 0 end-1] {
    set n [string length $z]
    if {$n>=2 && [string equal -length $n $z "-error"]} {
      set errormsg $value
    }
  }

  foreach {tn sql} $testarray {
    do_catchsql_test $prefix.$tn $sql [list 1 $errormsg]
  }
}

#-------------------------------------------------------------------------
# Test the libraries response if the %_node table is completely empty
# (i.e. the root node is missing), or has been removed from the database
# entirely.
#
create_t1
populate_t1
do_execsql_test rtreeA-1.0 {
  DELETE FROM t1_node;
} {}

do_corruption_tests rtreeA-1.1 {
  1   "SELECT * FROM t1"
  2   "SELECT * FROM t1 WHERE rowid=5"
  3   "INSERT INTO t1 VALUES(1000, 1, 2, 3, 4)"
  4   "SELECT * FROM t1 WHERE x1<10 AND x2>12"
}

do_execsql_test  rtreeA-1.2.0 { DROP TABLE t1_node } {}
do_corruption_tests rtreeA-1.2 -error "SQL logic error or missing database" {
  1   "SELECT * FROM t1"
  2   "SELECT * FROM t1 WHERE rowid=5"
  3   "INSERT INTO t1 VALUES(1000, 1, 2, 3, 4)"
  4   "SELECT * FROM t1 WHERE x1<10 AND x2>12"
}

#-------------------------------------------------------------------------
# Test the libraries response if some of the entries in the %_node table 
# are the wrong size.
#
create_t1
populate_t1
do_test rtreeA-2.1.0 {
  set nodes [db eval {select nodeno FROM t1_node}]
  foreach {a b c} $nodes { truncate_node $c 200 }
} {}
do_corruption_tests rtreeA-2.1 {
  1   "SELECT * FROM t1"
  2   "SELECT * FROM t1 WHERE rowid=5"
  3   "INSERT INTO t1 VALUES(1000, 1, 2, 3, 4)"
  4   "SELECT * FROM t1 WHERE x1<10 AND x2>12"
}

create_t1
populate_t1
do_test rtreeA-2.2.0 { truncate_node 1 200 } {}
do_corruption_tests rtreeA-2.2 {
  1   "SELECT * FROM t1"
  2   "SELECT * FROM t1 WHERE rowid=5"
  3   "INSERT INTO t1 VALUES(1000, 1, 2, 3, 4)"
  4   "SELECT * FROM t1 WHERE x1<10 AND x2>12"
}

#-------------------------------------------------------------------------
# Set the "depth" of the tree stored on the root node incorrectly. Test
# that this does not cause any problems.
#
create_t1
populate_t1
do_test rtreeA-3.1.0.1 { set_tree_depth t1 } {1}
do_test rtreeA-3.1.0.2 { set_tree_depth t1 3 } {3}
do_corruption_tests rtreeA-3.1 {
  1   "SELECT * FROM t1"
  2   "SELECT * FROM t1 WHERE rowid=5"
  3   "INSERT INTO t1 VALUES(1000, 1, 2, 3, 4)"
}

do_test rtreeA-3.2.0 { set_tree_depth t1 1000 } {1000}
do_corruption_tests rtreeA-3.2 {
  1   "SELECT * FROM t1"
  2   "SELECT * FROM t1 WHERE rowid=5"
  3   "INSERT INTO t1 VALUES(1000, 1, 2, 3, 4)"
}

create_t1
populate_t1
do_test rtreeA-3.3.0 { 
  execsql { DELETE FROM t1 WHERE rowid = 0 }
  set_tree_depth t1 65535
} {65535}
do_corruption_tests rtreeA-3.3 {
  1   "SELECT * FROM t1"
  2   "SELECT * FROM t1 WHERE rowid=5"
  3   "INSERT INTO t1 VALUES(1000, 1, 2, 3, 4)"
}

#-------------------------------------------------------------------------
# Set the "number of entries" field on some nodes incorrectly.
#
create_t1
populate_t1
do_test rtreeA-4.1.0 { 
  set_entry_count t1 1 4000
} {4000}
breakpoint
do_corruption_tests rtreeA-4.1 {
  1   "SELECT * FROM t1"
  2   "SELECT * FROM t1 WHERE rowid=5"
  3   "INSERT INTO t1 VALUES(1000, 1, 2, 3, 4)"
  4   "SELECT * FROM t1 WHERE x1<10 AND x2>12"
}

#-------------------------------------------------------------------------
# Remove entries from the %_parent table and check that this does not
# cause a crash.
#
create_t1
populate_t1
do_execsql_test rtreeA-5.1.0 { DELETE FROM t1_parent } {}
do_corruption_tests rtreeA-5.1 {
  1   "DELETE FROM t1 WHERE rowid = 5"
  2   "DELETE FROM t1"
}

finish_test