summaryrefslogtreecommitdiff
path: root/includes/Cdb_PHP.php
diff options
context:
space:
mode:
authorPierre Schmitz <pierre@archlinux.de>2010-07-28 11:52:48 +0200
committerPierre Schmitz <pierre@archlinux.de>2010-07-28 11:52:48 +0200
commit222b01f5169f1c7e69762e0e8904c24f78f71882 (patch)
tree8e932e12546bb991357ec48eb1638d1770be7a35 /includes/Cdb_PHP.php
parent00ab76a6b686e98a914afc1975812d2b1aaa7016 (diff)
update to MediaWiki 1.16.0
Diffstat (limited to 'includes/Cdb_PHP.php')
-rw-r--r--includes/Cdb_PHP.php374
1 files changed, 374 insertions, 0 deletions
diff --git a/includes/Cdb_PHP.php b/includes/Cdb_PHP.php
new file mode 100644
index 00000000..49294f71
--- /dev/null
+++ b/includes/Cdb_PHP.php
@@ -0,0 +1,374 @@
+<?php
+
+/**
+ * This is a port of D.J. Bernstein's CDB to PHP. It's based on the copy that
+ * appears in PHP 5.3. Changes are:
+ * * Error returns replaced with exceptions
+ * * Exception thrown if sizes or offsets are between 2GB and 4GB
+ * * Some variables renamed
+ */
+
+/**
+ * Common functions for readers and writers
+ */
+class CdbFunctions {
+ /**
+ * Take a modulo of a signed integer as if it were an unsigned integer.
+ * $b must be less than 0x40000000 and greater than 0
+ */
+ public static function unsignedMod( $a, $b ) {
+ if ( $a & 0x80000000 ) {
+ $m = ( $a & 0x7fffffff ) % $b + 2 * ( 0x40000000 % $b );
+ return $m % $b;
+ } else {
+ return $a % $b;
+ }
+ }
+
+ /**
+ * Shift a signed integer right as if it were unsigned
+ */
+ public static function unsignedShiftRight( $a, $b ) {
+ if ( $b == 0 ) {
+ return $a;
+ }
+ if ( $a & 0x80000000 ) {
+ return ( ( $a & 0x7fffffff ) >> $b ) | ( 0x40000000 >> ( $b - 1 ) );
+ } else {
+ return $a >> $b;
+ }
+ }
+
+ /**
+ * The CDB hash function.
+ */
+ public static function hash( $s ) {
+ $h = 5381;
+ for ( $i = 0; $i < strlen( $s ); $i++ ) {
+ $h5 = ($h << 5) & 0xffffffff;
+ // Do a 32-bit sum
+ // Inlined here for speed
+ $sum = ($h & 0x3fffffff) + ($h5 & 0x3fffffff);
+ $h =
+ (
+ ( $sum & 0x40000000 ? 1 : 0 )
+ + ( $h & 0x80000000 ? 2 : 0 )
+ + ( $h & 0x40000000 ? 1 : 0 )
+ + ( $h5 & 0x80000000 ? 2 : 0 )
+ + ( $h5 & 0x40000000 ? 1 : 0 )
+ ) << 30
+ | ( $sum & 0x3fffffff );
+ $h ^= ord( $s[$i] );
+ $h &= 0xffffffff;
+ }
+ return $h;
+ }
+}
+
+/**
+ * CDB reader class
+ */
+class CdbReader_PHP extends CdbReader {
+ /** The file handle */
+ var $handle;
+
+ /* number of hash slots searched under this key */
+ var $loop;
+
+ /* initialized if loop is nonzero */
+ var $khash;
+
+ /* initialized if loop is nonzero */
+ var $kpos;
+
+ /* initialized if loop is nonzero */
+ var $hpos;
+
+ /* initialized if loop is nonzero */
+ var $hslots;
+
+ /* initialized if findNext() returns true */
+ var $dpos;
+
+ /* initialized if cdb_findnext() returns 1 */
+ var $dlen;
+
+ function __construct( $fileName ) {
+ $this->handle = fopen( $fileName, 'rb' );
+ if ( !$this->handle ) {
+ throw new MWException( 'Unable to open DB file "' . $fileName . '"' );
+ }
+ $this->findStart();
+ }
+
+ function close() {
+ if( isset($this->handle) )
+ fclose( $this->handle );
+ unset( $this->handle );
+ }
+
+ public function get( $key ) {
+ // strval is required
+ if ( $this->find( strval( $key ) ) ) {
+ return $this->read( $this->dlen, $this->dpos );
+ } else {
+ return false;
+ }
+ }
+
+ protected function match( $key, $pos ) {
+ $buf = $this->read( strlen( $key ), $pos );
+ return $buf === $key;
+ }
+
+ protected function findStart() {
+ $this->loop = 0;
+ }
+
+ protected function read( $length, $pos ) {
+ if ( fseek( $this->handle, $pos ) == -1 ) {
+ // This can easily happen if the internal pointers are incorrect
+ throw new MWException( __METHOD__.': seek failed, file may be corrupted.' );
+ }
+
+ if ( $length == 0 ) {
+ return '';
+ }
+
+ $buf = fread( $this->handle, $length );
+ if ( $buf === false || strlen( $buf ) !== $length ) {
+ throw new MWException( __METHOD__.': read from cdb file failed, file may be corrupted' );
+ }
+ return $buf;
+ }
+
+ /**
+ * Unpack an unsigned integer and throw an exception if it needs more than 31 bits
+ */
+ protected function unpack31( $s ) {
+ $data = unpack( 'V', $s );
+ if ( $data[1] > 0x7fffffff ) {
+ throw new MWException( __METHOD__.': error in CDB file, integer too big' );
+ }
+ return $data[1];
+ }
+
+ /**
+ * Unpack a 32-bit signed integer
+ */
+ protected function unpackSigned( $s ) {
+ $data = unpack( 'va/vb', $s );
+ return $data['a'] | ( $data['b'] << 16 );
+ }
+
+ protected function findNext( $key ) {
+ if ( !$this->loop ) {
+ $u = CdbFunctions::hash( $key );
+ $buf = $this->read( 8, ( $u << 3 ) & 2047 );
+ $this->hslots = $this->unpack31( substr( $buf, 4 ) );
+ if ( !$this->hslots ) {
+ return false;
+ }
+ $this->hpos = $this->unpack31( substr( $buf, 0, 4 ) );
+ $this->khash = $u;
+ $u = CdbFunctions::unsignedShiftRight( $u, 8 );
+ $u = CdbFunctions::unsignedMod( $u, $this->hslots );
+ $u <<= 3;
+ $this->kpos = $this->hpos + $u;
+ }
+
+ while ( $this->loop < $this->hslots ) {
+ $buf = $this->read( 8, $this->kpos );
+ $pos = $this->unpack31( substr( $buf, 4 ) );
+ if ( !$pos ) {
+ return false;
+ }
+ $this->loop += 1;
+ $this->kpos += 8;
+ if ( $this->kpos == $this->hpos + ( $this->hslots << 3 ) ) {
+ $this->kpos = $this->hpos;
+ }
+ $u = $this->unpackSigned( substr( $buf, 0, 4 ) );
+ if ( $u === $this->khash ) {
+ $buf = $this->read( 8, $pos );
+ $keyLen = $this->unpack31( substr( $buf, 0, 4 ) );
+ if ( $keyLen == strlen( $key ) && $this->match( $key, $pos + 8 ) ) {
+ // Found
+ $this->dlen = $this->unpack31( substr( $buf, 4 ) );
+ $this->dpos = $pos + 8 + $keyLen;
+ return true;
+ }
+ }
+ }
+ return false;
+ }
+
+ protected function find( $key ) {
+ $this->findStart();
+ return $this->findNext( $key );
+ }
+}
+
+/**
+ * CDB writer class
+ */
+class CdbWriter_PHP extends CdbWriter {
+ var $handle, $realFileName, $tmpFileName;
+
+ var $hplist;
+ var $numEntries, $pos;
+
+ function __construct( $fileName ) {
+ $this->realFileName = $fileName;
+ $this->tmpFileName = $fileName . '.tmp.' . mt_rand( 0, 0x7fffffff );
+ $this->handle = fopen( $this->tmpFileName, 'wb' );
+ if ( !$this->handle ) {
+ throw new MWException( 'Unable to open DB file for write "' . $fileName . '"' );
+ }
+ $this->hplist = array();
+ $this->numentries = 0;
+ $this->pos = 2048; // leaving space for the pointer array, 256 * 8
+ if ( fseek( $this->handle, $this->pos ) == -1 ) {
+ throw new MWException( __METHOD__.': fseek failed' );
+ }
+ }
+
+ function __destruct() {
+ if ( isset( $this->handle ) ) {
+ $this->close();
+ }
+ }
+
+ public function set( $key, $value ) {
+ if ( strval( $key ) === '' ) {
+ // DBA cross-check hack
+ return;
+ }
+ $this->addbegin( strlen( $key ), strlen( $value ) );
+ $this->write( $key );
+ $this->write( $value );
+ $this->addend( strlen( $key ), strlen( $value ), CdbFunctions::hash( $key ) );
+ }
+
+ public function close() {
+ $this->finish();
+ if( isset($this->handle) )
+ fclose( $this->handle );
+ if ( wfIsWindows() && file_exists($this->realFileName) ) {
+ unlink( $this->realFileName );
+ }
+ if ( !rename( $this->tmpFileName, $this->realFileName ) ) {
+ throw new MWException( 'Unable to move the new CDB file into place.' );
+ }
+ unset( $this->handle );
+ }
+
+ protected function write( $buf ) {
+ $len = fwrite( $this->handle, $buf );
+ if ( $len !== strlen( $buf ) ) {
+ throw new MWException( 'Error writing to CDB file.' );
+ }
+ }
+
+ protected function posplus( $len ) {
+ $newpos = $this->pos + $len;
+ if ( $newpos > 0x7fffffff ) {
+ throw new MWException( 'A value in the CDB file is too large' );
+ }
+ $this->pos = $newpos;
+ }
+
+ protected function addend( $keylen, $datalen, $h ) {
+ $this->hplist[] = array(
+ 'h' => $h,
+ 'p' => $this->pos
+ );
+
+ $this->numentries++;
+ $this->posplus( 8 );
+ $this->posplus( $keylen );
+ $this->posplus( $datalen );
+ }
+
+ protected function addbegin( $keylen, $datalen ) {
+ if ( $keylen > 0x7fffffff ) {
+ throw new MWException( __METHOD__.': key length too long' );
+ }
+ if ( $datalen > 0x7fffffff ) {
+ throw new MWException( __METHOD__.': data length too long' );
+ }
+ $buf = pack( 'VV', $keylen, $datalen );
+ $this->write( $buf );
+ }
+
+ protected function finish() {
+ // Hack for DBA cross-check
+ $this->hplist = array_reverse( $this->hplist );
+
+ // Calculate the number of items that will be in each hashtable
+ $counts = array_fill( 0, 256, 0 );
+ foreach ( $this->hplist as $item ) {
+ ++ $counts[ 255 & $item['h'] ];
+ }
+
+ // Fill in $starts with the *end* indexes
+ $starts = array();
+ $pos = 0;
+ for ( $i = 0; $i < 256; ++$i ) {
+ $pos += $counts[$i];
+ $starts[$i] = $pos;
+ }
+
+ // Excessively clever and indulgent code to simultaneously fill $packedTables
+ // with the packed hashtables, and adjust the elements of $starts
+ // to actually point to the starts instead of the ends.
+ $packedTables = array_fill( 0, $this->numentries, false );
+ foreach ( $this->hplist as $item ) {
+ $packedTables[--$starts[255 & $item['h']]] = $item;
+ }
+
+ $final = '';
+ for ( $i = 0; $i < 256; ++$i ) {
+ $count = $counts[$i];
+
+ // The size of the hashtable will be double the item count.
+ // The rest of the slots will be empty.
+ $len = $count + $count;
+ $final .= pack( 'VV', $this->pos, $len );
+
+ $hashtable = array();
+ for ( $u = 0; $u < $len; ++$u ) {
+ $hashtable[$u] = array( 'h' => 0, 'p' => 0 );
+ }
+
+ // Fill the hashtable, using the next empty slot if the hashed slot
+ // is taken.
+ for ( $u = 0; $u < $count; ++$u ) {
+ $hp = $packedTables[$starts[$i] + $u];
+ $where = CdbFunctions::unsignedMod(
+ CdbFunctions::unsignedShiftRight( $hp['h'], 8 ), $len );
+ while ( $hashtable[$where]['p'] )
+ if ( ++$where == $len )
+ $where = 0;
+ $hashtable[$where] = $hp;
+ }
+
+ // Write the hashtable
+ for ( $u = 0; $u < $len; ++$u ) {
+ $buf = pack( 'vvV',
+ $hashtable[$u]['h'] & 0xffff,
+ CdbFunctions::unsignedShiftRight( $hashtable[$u]['h'], 16 ),
+ $hashtable[$u]['p'] );
+ $this->write( $buf );
+ $this->posplus( 8 );
+ }
+ }
+
+ // Write the pointer array at the start of the file
+ rewind( $this->handle );
+ if ( ftell( $this->handle ) != 0 ) {
+ throw new MWException( __METHOD__.': Error rewinding to start of file' );
+ }
+ $this->write( $final );
+ }
+}