wiki.techinc.nl/includes/Tokenizer.php
2004-06-17 09:49:28 +00:00

295 lines
8.9 KiB
PHP

<?php
class Tokenizer {
/* private */ var $mText, # Text to be processed by the tokenizer
$mPos, # current position of tokenizer in text
$mTextLength, # Length of $mText
$mQueuedToken; # Tokens that were already found, but not
# returned yet.
/* private */ function Tokenizer()
{
global $wgLang;
$this->mPos=0;
$this->mTokenQueue=array();
$this->linkPrefixExtension = $wgLang->linkPrefixExtension();
}
# factory function
function newFromString( $s )
{
$fname = 'Tokenizer::newFromString';
wfProfileIn( $fname );
$t = new Tokenizer();
$t->mText = $s;
$t->mTextLength = strlen( $s );
wfProfileOut( $fname );
return $t;
}
// Return the next token, but do not increase the pointer. The next call
// to previewToken or nextToken will return the same token again.
// Actually, the pointer is increased, but the token is queued. The next
// call to previewToken or nextToken will check the queue and return
// the stored token.
function previewToken()
{
$fname = 'Tokenizer::previewToken';
wfProfileIn( $fname );
if ( count( $this->mQueuedToken ) != 0 ) {
// still one token from the last round around. Return that one first.
$token = $this->mQueuedToken[0];
} else {
$token = $this->nextToken();
array_unshift( $this->mQueuedToken, $token );
}
wfProfileOut( $fname );
return $token;
}
// get the next token
// proceeds character by character through the text, looking for characters needing
// special attention. Those are currently: I, R, ', [, ], newline
//
// TODO: handling of French blanks not yet implemented
function nextToken()
{
$fname = 'Tokenizer::nextToken';
wfProfileIn( $fname );
if ( count( $this->mQueuedToken ) != 0 ) {
// still one token from the last round around. Return that one first.
$token = array_shift( $this->mQueuedToken );
} else if ( $this->mPos > $this->mTextLength ) {
// If no text is left, return "false".
$token = false;
} else {
$token['text']='';
$token['type']='text';
while ( $this->mPos <= $this->mTextLength ) {
switch ( @$ch = $this->mText[$this->mPos] ) {
case 'R': // for "RFC "
if ( $this->continues('FC ') ) {
$queueToken['type'] = $queueToken['text'] = 'RFC ';
$this->mQueuedToken[] = $queueToken;
$this->mPos += 3;
break 2; // switch + while
}
break;
case 'I': // for "ISBN "
if ( $this->continues('SBN ') ) {
$queueToken['type'] = $queueToken['text'] = 'ISBN ';
$this->mQueuedToken[] = $queueToken;
$this->mPos += 4;
break 2; // switch + while
}
break;
case '[': // for links "[["
if ( $this->continues('[[') ) {
$queueToken['type'] = '[[[';
$queueToken['text'] = '';
$this->mQueuedToken[] = $queueToken;
$this->mPos += 3;
break 2; // switch + while
} else if ( $this->continues('[') ) {
$queueToken['type'] = '[[';
$queueToken['text'] = '';
// Check for a "prefixed link", e.g. Al[[Khazar]]
// Mostly for arabic wikipedia
if ( $this->linkPrefixExtension ) {
while ( $this->linkPrefixExtension
&& ($len = strlen( $token['text'] ) ) > 0
&& !ctype_space( $token['text'][$len-1] ) )
{
//prepend the character to the link's open tag
$queueToken['text'] = $token['text'][$len-1] . $queueToken['text'];
//remove character from the end of the text token
$token['text'] = substr( $token['text'], 0, -1);
}
}
$this->mQueuedToken[] = $queueToken;
$this->mPos += 2;
break 2; // switch + while
}
break;
case ']': // for end of links "]]"
if ( $this->continues(']') ) {
$queueToken['type'] = ']]';
$queueToken['text'] = '';
$this->mQueuedToken[] = $queueToken;
$this->mPos += 2;
break 2; // switch + while
}
break;
case "'": // for all kind of em's and strong's
if ( $this->continues("'") ) {
$queueToken['type'] = "'";
$queueToken['text'] = '';
while( ($this->mPos+1 < $this->mTextLength)
&& $this->mText[$this->mPos+1] == "'" )
{
$queueToken['type'] .= "'";
$queueToken['pos'] = $this->mPos;
$this->mPos ++;
}
$this->mQueuedToken[] = $queueToken;
$this->mPos ++;
break 2; // switch + while
}
break;
case "\n": // for block levels, actually, only "----" is handled.
case "\r": // headings are detected to close any unbalanced em or strong tags in a section
if ( $this->continues( '----' ) )
{
$queueToken['type'] = '----';
$queueToken['text'] = '';
$this->mQueuedToken[] = $queueToken;
$this->mPos += 5;
while ( $this->mPos<$this->mTextLength
and $this->mText[$this->mPos] == '-' )
{
$this->mPos ++;
}
break 2;
} else if (
$this->continues( '<h' ) and (
$this->continues( '<h1' ) or
$this->continues( '<h2' ) or
$this->continues( '<h3' ) or
$this->continues( '<h4' ) or
$this->continues( '<h5' ) or
$this->continues( '<h6' )
)
) { // heading
$queueToken['type'] = 'h';
$queueToken['text'] = '';
$this->mQueuedToken[] = $queueToken;
$this->mPos ++;
break 2; // switch + while
}
break;
case '!': // French spacing rules have a space before exclamation
case '?': // and question marks. Those have to become &nbsp;
case ':': // And colons, Hashar says ...
if ( $this->preceeded( ' ' ) )
{
// strip blank from Token
$token['text'] = substr( $token['text'], 0, -1 );
$queueToken['type'] = 'blank';
$queueToken['text'] = " {$ch}";
$this->mQueuedToken[] = $queueToken;
$this->mPos ++;
break 2; // switch + while
}
break;
case '0': // A space between two numbers is used to ease reading
case '1': // of big numbers, e.g. 1 000 000. Those spaces need
case '2': // to be unbreakable
case '3':
case '4':
case '5':
case '6':
case '7':
case '8':
case '9':
if ( ($this->mTextLength >= $this->mPos +2)
&& ($this->mText[$this->mPos+1] == " ")
&& ctype_digit( $this->mText[$this->mPos+2] ) )
{
$queueToken['type'] = 'blank';
$queueToken['text'] = $ch . ' ';
$this->mQueuedToken[] = $queueToken;
$this->mPos += 2;
break 2; // switch + while
}
break;
case "\302": // first byte of UTF-8 Character Guillemet-left
if ( $this->continues( "\253 ") ) // second byte and a blank
{
$queueToken['type'] = 'blank';
$queueToken['text'] = "\302\253 ";
$this->mQueuedToken[] = $queueToken;
$this->mPos += 3;
break 2; // switch + while
}
break;
case "\273": //last byte of UTF-8 Character Guillemet-right
if ( $this->preceeded( " \302" ) )
{
$queueToken['type'] = 'blank';
$queueToken['text'] = " \302\273";
$token['text'] = substr( $token['text'], 0, -2 );
$this->mQueuedToken[] = $queueToken;
$this->mPos ++;
break 2; // switch + while
}
break;
case '&': //extensions like <timeline>, since HTML stripping has already been done,
//those look like &lt;timeline&gt;
if ( $this->continues( "lt;timeline&gt;" ) )
{
$queueToken['type'] = "<timeline>";
$queueToken['text'] = "&lt;timeline&gt;";
$this->mQueuedToken[] = $queueToken;
$this->mPos += 16;
break 2; // switch + while
}
break;
} /* switch */
$token['text'].=$ch;
$this->mPos ++;
// echo $this->mPos . "<br>\n";
} /* while */
} /* if (nothing left in queue) */
wfProfileOut( $fname );
return $token;
}
// function continues
// checks whether the mText continues with $cont from mPos+1
/* private */ function continues( $cont )
{
// If string is not long enough to contain $cont, return false
if ( $this->mTextLength < $this->mPos + strlen( $cont ) )
return false;
for ( $i=0; $i < strlen( $cont ); $i++ )
{
if ( $this->mText[$this->mPos+1+$i] != $cont[$i] )
return false;
}
return true;
}
// function preceeded
// checks whether the mText is preceeded by $prec at position mPos
/* private */ function preceeded( $prec )
{
$len = strlen( $prec );
// if $prec is longer than the text up to mPos, return false
if ( $this->mPos < $len )
return false;
return ( 0 == strcmp( $prec, substr($this->mText, $this->mPos-$len, $len) ) );
}
function readAllUntil( $border )
{
$n = strpos( $this->mText, $border, $this->mPos );
if ( $n === false )
return '';
$ret = substr( $this->mText, $this->mPos, $n - $this->mPos );
$this->mPos = $n + strlen( $border ) + 1;
return $ret;
}
}