includes/Tokenizer.php

   1 <?php
   2 class Tokenizer {
   3         /* private */ var $mText,               # Text to be processed by the tokenizer
   4                           $mPos,                # current position of tokenizer in text
   5                           $mTextLength,         # Length of $mText
   6                           $mQueuedToken;        # Tokens that were already found, but not
   7                                                 # returned yet.
   8
   9         /* private */ function Tokenizer()
  10         {
  11                 global $wgLang;
  12
  13                 $this->mPos=0;
  14                 $this->mTokenQueue=array();
  15                 $this->linkPrefixExtension = $wgLang->linkPrefixExtension();
  16         }
  17
  18         # factory function
  19         function newFromString( $s )
  20         {
  21                 $fname = "Tokenizer::newFromString";
  22                 wfProfileIn( $fname );
  23
  24                 $t = new Tokenizer();
  25                 $t->mText = $s;
  26                 $t->mTextLength = strlen( $s );
  27
  28                 wfProfileOut( $fname );
  29                 return $t;
  30         }
  31
  32
  33         // Return the next token, but do not increase the pointer. The next call
  34         // to previewToken or nextToken will return the same token again.
  35         // Actually, the pointer is increased, but the token is queued. The next
  36         // call to previewToken or nextToken will check the queue and return
  37         // the stored token.
  38         function previewToken()
  39         {
  40                 $fname = "Tokenizer::previewToken";
  41                 wfProfileIn( $fname );
  42
  43                 if ( count( $this->mQueuedToken ) != 0 ) {
  44                         // still one token from the last round around. Return that one first.
  45                         $token = $this->mQueuedToken[0];
  46                 } else {
  47                         $token = $this->nextToken();
  48                         array_unshift( $this->mQueuedToken, $token );
  49                 }
  50
  51                 wfProfileOut( $fname );
  52                 return $token;
  53         }
  54
  55
  56         // get the next token
  57         // proceeds character by character through the text, looking for characters needing
  58         // special attention. Those are currently: I, R, ', [, ], newline
  59         //
  60         // TODO:  handling of French blanks not yet implemented
  61         function nextToken()
  62         {
  63                 $fname = "Tokenizer::nextToken";
  64                 wfProfileIn( $fname );
  65
  66                 if ( count( $this->mQueuedToken ) != 0 ) {
  67                         // still one token from the last round around. Return that one first.
  68                         $token = array_shift( $this->mQueuedToken );
  69                 } else if ( $this->mPos > $this->mTextLength ) {
  70                         // If no text is left, return "false".
  71                         $token = false;
  72                 } else {
  73
  74                         $token["text"]="";
  75                         $token["type"]="text";
  76
  77                         while ( $this->mPos <= $this->mTextLength ) {
  78                                 switch ( @$ch = $this->mText[$this->mPos] ) {
  79                                         case 'R': // for "RFC "
  80                                                 if ( $this->continues("FC ") ) {
  81                                                         $queueToken["type"] = $queueToken["text"] = "RFC ";
  82                                                         $this->mQueuedToken[] = $queueToken;
  83                                                         $this->mPos += 3;
  84                                                         break 2; // switch + while
  85                                                 }
  86                                                 break;
  87                                         case 'I': // for "ISBN "
  88                                                 if ( $this->continues("SBN ") ) {
  89                                                         $queueToken["type"] = $queueToken["text"] = "ISBN ";
  90                                                         $this->mQueuedToken[] = $queueToken;
  91                                                         $this->mPos += 4;
  92                                                         break 2; // switch + while
  93                                                 }
  94                                                 break;
  95                                         case "[": // for links "[["
  96                                                 if ( $this->continues("[[") ) {
  97                                                         $queueToken["type"] = "[[[";
  98                                                         $queueToken["text"] = "";
  99                                                         $this->mQueuedToken[] = $queueToken;
 100                                                         $this->mPos += 3;
 101                                                         break 2; // switch + while
 102                                                 } else if ( $this->continues("[") ) {
 103                                                         $queueToken["type"] = "[[";
 104                                                         $queueToken["text"] = "";
 105                                                         // Check for a "prefixed link", e.g. Al[[Khazar]]
 106                                                         // Mostly for arabic wikipedia
 107                                                         if ( $this->linkPrefixExtension ) {
 108                                                                 while (    $this->linkPrefixExtension
 109                                                                         && ($len = strlen( $token["text"] ) ) > 0
 110                                                                         && !ctype_space( $token["text"][$len-1] ) )
 111                                                                 {
 112                                                                         //prepend the character to the link's open tag
 113                                                                         $queueToken["text"] = $token["text"][$len-1] . $queueToken["text"];
 114                                                                         //remove character from the end of the text token
 115                                                                         $token["text"] = substr( $token["text"], 0, -1);
 116                                                                 }
 117                                                         }
 118                                                         $this->mQueuedToken[] = $queueToken;
 119                                                         $this->mPos += 2;
 120                                                         break 2; // switch + while
 121                                                 }
 122                                                 break;
 123                                         case "]": // for end of links "]]"
 124                                                 if ( $this->continues("]") ) {
 125                                                         $queueToken["type"] = "]]";
 126                                                         $queueToken["text"] = "";
 127                                                         $this->mQueuedToken[] = $queueToken;
 128                                                         $this->mPos += 2;
 129                                                         break 2; // switch + while
 130                                                 }
 131                                                 break;
 132                                         case "'": // for all kind of em's and strong's
 133                                                 if ( $this->continues("'") ) {
 134                                                         $queueToken["type"] = "'";
 135                                                         $queueToken["text"] = "";
 136                                                         while(   ($this->mPos+1 < $this->mTextLength)
 137                                                                && $this->mText[$this->mPos+1] == "'" )
 138                                                         {
 139                                                                 $queueToken["type"] .= "'";
 140                                                                 $queueToken["pos"] = $this->mPos;
 141                                                                 $this->mPos ++;
 142                                                         }
 143
 144                                                         $this->mQueuedToken[] = $queueToken;
 145                                                         $this->mPos ++;
 146                                                         break 2; // switch + while
 147                                                 }
 148                                                 break;
 149                                         case "\n": // for block levels, actually, only "----" is handled.
 150                                         case "\r":
 151                                                 if ( $this->continues( "----" ) )
 152                                                 {
 153                                                         $queueToken["type"] = "----";
 154                                                         $queueToken["text"] = "";
 155                                                         $this->mQueuedToken[] = $queueToken;
 156                                                         $this->mPos += 5;
 157                                                         while (     $this->mPos<$this->mTextLength
 158                                                                 and $this->mText[$this->mPos] == "-" )
 159                                                         {
 160                                                                 $this->mPos ++;
 161                                                         }
 162                                                         break 2;
 163                                                 }
 164                                                 break;
 165                                         case "!": // French spacing rules have a space before exclamation
 166                                         case "?": // and question marks. Those have to become &nbsp;
 167                                         case ":": // And colons, Hashar says ...
 168                                                 if ( $this->preceeded( " " ) )
 169                                                 {
 170                                                         // strip blank from Token
 171                                                         $token["text"] = substr( $token["text"], 0, -1 );
 172                                                         $queueToken["type"] = "blank";
 173                                                         $queueToken["text"] = " {$ch}";
 174                                                         $this->mQueuedToken[] = $queueToken;
 175                                                         $this->mPos ++;
 176                                                         break 2; // switch + while
 177                                                 }
 178                                                 break;
 179                                         case "0": // A space between two numbers is used to ease reading
 180                                         case "1": // of big numbers, e.g. 1 000 000. Those spaces need
 181                                         case "2": // to be unbreakable
 182                                         case "3":
 183                                         case "4":
 184                                         case "5":
 185                                         case "6":
 186                                         case "7":
 187                                         case "8":
 188                                         case "9":
 189                                                 if (    ($this->mTextLength >= $this->mPos +2)
 190                                                      && ($this->mText[$this->mPos+1] == " ")
 191                                                      && ctype_digit( $this->mText[$this->mPos+2] ) )
 192                                                 {
 193                                                         $queueToken["type"] = "blank";
 194                                                         $queueToken["text"] = $ch . " ";
 195                                                         $this->mQueuedToken[] = $queueToken;
 196                                                         $this->mPos += 2;
 197                                                         break 2; // switch + while
 198                                                 }
 199                                                 break;
 200                                         case "\302": // first byte of UTF-8 Character Guillemet-left
 201                                                 if ( $this->continues( "\253 ") ) // second byte and a blank
 202                                                 {
 203                                                         $queueToken["type"] = "blank";
 204                                                         $queueToken["text"] = "\302\253 ";
 205                                                         $this->mQueuedToken[] = $queueToken;
 206                                                         $this->mPos += 3;
 207                                                         break 2; // switch + while
 208                                                 }
 209                                                 break;
 210                                         case "\273": //last byte of UTF-8 Character Guillemet-right
 211                                                 if ( $this->preceeded( " \302" ) )
 212                                                 {
 213                                                         $queueToken["type"] = "blank";
 214                                                         $queueToken["text"] = " \302\273";
 215                                                         $token["text"] = substr( $token["text"], 0, -2 );
 216                                                         $this->mQueuedToken[] = $queueToken;
 217                                                         $this->mPos ++;
 218                                                         break 2; // switch + while
 219                                                 }
 220                                                 break;
 221                                         case "&": //extensions like <timeline>, since HTML stripping has already been done,
 222                                                   //those look like &lt;timeline&gt;
 223                                                 if ( $this->continues( "lt;timeline&gt;" ) )
 224                                                 {
 225                                                         $queueToken["type"] = "<timeline>";
 226                                                         $queueToken["text"] = "&lt;timeline&gt;";
 227                                                         $this->mQueuedToken[] = $queueToken;
 228                                                         $this->mPos += 16;
 229                                                         break 2; // switch + while
 230                                                 }
 231                                                 break;
 232
 233                                 } /* switch */
 234                                 $token["text"].=$ch;
 235                                 $this->mPos ++;
 236                                 // echo $this->mPos . "<br>\n";
 237                         } /* while */
 238                 } /* if (nothing left in queue) */
 239
 240                 wfProfileOut( $fname );
 241                 return $token;
 242         }
 243
 244         // function continues
 245         // checks whether the mText continues with $cont from mPos+1
 246         /* private */ function continues( $cont )
 247         {
 248                 // If string is not long enough to contain $cont, return false
 249                 if ( $this->mTextLength < $this->mPos + strlen( $cont ) )
 250                         return false;
 251                 for ( $i=0; $i < strlen( $cont ); $i++ )
 252                 {
 253                         if ( $this->mText[$this->mPos+1+$i] != $cont[$i] )
 254                                 return false;
 255                 }
 256                 return true;
 257         }
 258
 259         // function preceeded
 260         // checks whether the mText is preceeded by $prec at position mPos
 261         /* private */ function preceeded( $prec )
 262         {
 263                 $len = strlen( $prec );
 264                 // if $prec is longer than the text up to mPos, return false
 265                 if ( $this->mPos < $len )
 266                         return false;
 267                 return ( 0 == strcmp( $prec, substr($this->mText, $this->mPos-$len, $len) ) );
 268         }
 269
 270         function readAllUntil( $border )
 271         {
 272                 $n = strpos( $this->mText, $border, $this->mPos );
 273                 if ( $n === false )
 274                         return "";
 275                 $ret = substr( $this->mText, $this->mPos, $n - $this->mPos );
 276                 $this->mPos = $n + strlen( $border ) + 1;
 277                 return $ret;
 278         }
 279
 280 }
 281