BlockLevelPass.php

Go to the documentation of this file.

<?php
 
class BlockLevelPass {
    private $DTopen = false;
    private $inPre = false;
    private $lastSection = '';
    private $lineStart;
    private $text;
 
    # State constants for the definition list colon extraction
    const COLON_STATE_TEXT = 0;
    const COLON_STATE_TAG = 1;
    const COLON_STATE_TAGSTART = 2;
    const COLON_STATE_CLOSETAG = 3;
    const COLON_STATE_TAGSLASH = 4;
    const COLON_STATE_COMMENT = 5;
    const COLON_STATE_COMMENTDASH = 6;
    const COLON_STATE_COMMENTDASHDASH = 7;
    const COLON_STATE_LC = 8;
 
    public static function doBlockLevels( $text, $lineStart ) {
        $pass = new self( $text, $lineStart );
        return $pass->execute();
    }
 
    private function __construct( $text, $lineStart ) {
        $this->text = $text;
        $this->lineStart = $lineStart;
    }
 
    private function closeParagraph() {
        $result = '';
        if ( $this->lastSection !== '' ) {
            $result = '</' . $this->lastSection . ">\n";
        }
        $this->inPre = false;
        $this->lastSection = '';
        return $result;
    }
 
    private function getCommon( $st1, $st2 ) {
        $shorter = min( strlen( $st1 ), strlen( $st2 ) );
 
        for ( $i = 0; $i < $shorter; ++$i ) {
            if ( $st1[$i] !== $st2[$i] ) {
                break;
            }
        }
        return $i;
    }
 
    private function openList( $char ) {
        $result = $this->closeParagraph();
 
        if ( '*' === $char ) {
            $result .= "<ul><li>";
        } elseif ( '#' === $char ) {
            $result .= "<ol><li>";
        } elseif ( ':' === $char ) {
            $result .= "<dl><dd>";
        } elseif ( ';' === $char ) {
            $result .= "<dl><dt>";
            $this->DTopen = true;
        } else {
            $result = '<!-- ERR 1 -->';
        }
 
        return $result;
    }
 
    private function nextItem( $char ) {
        if ( '*' === $char || '#' === $char ) {
            return "</li>\n<li>";
        } elseif ( ':' === $char || ';' === $char ) {
            $close = "</dd>\n";
            if ( $this->DTopen ) {
                $close = "</dt>\n";
            }
            if ( ';' === $char ) {
                $this->DTopen = true;
                return $close . '<dt>';
            } else {
                $this->DTopen = false;
                return $close . '<dd>';
            }
        }
        return '<!-- ERR 2 -->';
    }
 
    private function closeList( $char ) {
        if ( '*' === $char ) {
            $text = "</li></ul>";
        } elseif ( '#' === $char ) {
            $text = "</li></ol>";
        } elseif ( ':' === $char ) {
            if ( $this->DTopen ) {
                $this->DTopen = false;
                $text = "</dt></dl>";
            } else {
                $text = "</dd></dl>";
            }
        } else {
            return '<!-- ERR 3 -->';
        }
        return $text;
    }
 
    private function execute() {
        $text = $this->text;
        # Parsing through the text line by line.  The main thing
        # happening here is handling of block-level elements p, pre,
        # and making lists from lines starting with * # : etc.
        $textLines = StringUtils::explode( "\n", $text );
 
        $lastPrefix = $output = '';
        $this->DTopen = $inBlockElem = false;
        $prefixLength = 0;
        $pendingPTag = false;
        $inBlockquote = false;
 
        foreach ( $textLines as $inputLine ) {
            # Fix up $lineStart
            if ( !$this->lineStart ) {
                $output .= $inputLine;
                $this->lineStart = true;
                continue;
            }
            # * = ul
            # # = ol
            # ; = dt
            # : = dd
 
            $lastPrefixLength = strlen( $lastPrefix );
            $preCloseMatch = preg_match( '/<\\/pre/i', $inputLine );
            $preOpenMatch = preg_match( '/<pre/i', $inputLine );
            # If not in a <pre> element, scan for and figure out what prefixes are there.
            if ( !$this->inPre ) {
                # Multiple prefixes may abut each other for nested lists.
                $prefixLength = strspn( $inputLine, '*#:;' );
                $prefix = substr( $inputLine, 0, $prefixLength );
 
                # eh?
                # ; and : are both from definition-lists, so they're equivalent
                #  for the purposes of determining whether or not we need to open/close
                #  elements.
                $prefix2 = str_replace( ';', ':', $prefix );
                $t = substr( $inputLine, $prefixLength );
                $this->inPre = (bool)$preOpenMatch;
            } else {
                # Don't interpret any other prefixes in preformatted text
                $prefixLength = 0;
                $prefix = $prefix2 = '';
                $t = $inputLine;
            }
 
            # List generation
            if ( $prefixLength && $lastPrefix === $prefix2 ) {
                # Same as the last item, so no need to deal with nesting or opening stuff
                $output .= $this->nextItem( substr( $prefix, -1 ) );
                $pendingPTag = false;
 
                if ( substr( $prefix, -1 ) === ';' ) {
                    # The one nasty exception: definition lists work like this:
                    # ; title : definition text
                    # So we check for : in the remainder text to split up the
                    # title and definition, without b0rking links.
                    $term = $t2 = '';
                    if ( $this->findColonNoLinks( $t, $term, $t2 ) !== false ) {
                        $t = $t2;
                        // Trim whitespace in list items
                        $output .= trim( $term ) . $this->nextItem( ':' );
                    }
                }
            } elseif ( $prefixLength || $lastPrefixLength ) {
                # We need to open or close prefixes, or both.
 
                # Either open or close a level...
                $commonPrefixLength = $this->getCommon( $prefix, $lastPrefix );
                $pendingPTag = false;
 
                # Close all the prefixes which aren't shared.
                while ( $commonPrefixLength < $lastPrefixLength ) {
                    $output .= $this->closeList( $lastPrefix[$lastPrefixLength - 1] );
                    --$lastPrefixLength;
                }
 
                # Continue the current prefix if appropriate.
                if ( $prefixLength <= $commonPrefixLength && $commonPrefixLength > 0 ) {
                    $output .= $this->nextItem( $prefix[$commonPrefixLength - 1] );
                }
 
                # Close an open <dt> if we have a <dd> (":") starting on this line
                if ( $this->DTopen && $commonPrefixLength > 0 && $prefix[$commonPrefixLength - 1] === ':' ) {
                    $output .= $this->nextItem( ':' );
                }
 
                # Open prefixes where appropriate.
                if ( $lastPrefix && $prefixLength > $commonPrefixLength ) {
                    $output .= "\n";
                }
                while ( $prefixLength > $commonPrefixLength ) {
                    $char = $prefix[$commonPrefixLength];
                    $output .= $this->openList( $char );
 
                    if ( ';' === $char ) {
                        # @todo FIXME: This is dupe of code above
                        if ( $this->findColonNoLinks( $t, $term, $t2 ) !== false ) {
                            $t = $t2;
                            // Trim whitespace in list items
                            $output .= trim( $term ) . $this->nextItem( ':' );
                        }
                    }
                    ++$commonPrefixLength;
                }
                if ( !$prefixLength && $lastPrefix ) {
                    $output .= "\n";
                }
                $lastPrefix = $prefix2;
            }
 
            # If we have no prefixes, go to paragraph mode.
            if ( 0 == $prefixLength ) {
                # No prefix (not in list)--go to paragraph mode
                # @todo consider using a stack for nestable elements like span, table and div
 
                // P-wrapping and indent-pre are suppressed inside, not outside
                $blockElems = 'table|h1|h2|h3|h4|h5|h6|pre|p|ul|ol|dl';
                // P-wrapping and indent-pre are suppressed outside, not inside
                $antiBlockElems = 'td|th';
 
                $openMatch = preg_match(
                    '/<('
                        . "({$blockElems})|\\/({$antiBlockElems})|"
                        // Always suppresses
                        . '\\/?(tr|dt|dd|li)'
                        . ')\\b/iS',
                    $t
                );
                $closeMatch = preg_match(
                    '/<('
                        . "\\/({$blockElems})|({$antiBlockElems})|"
                        // Never suppresses
                        . '\\/?(center|blockquote|div|hr|mw:)'
                        . ')\\b/iS',
                    $t
                );
 
                // Any match closes the paragraph, but only when `!$closeMatch`
                // do we enter block mode.  The oddities with table rows and
                // cells are to avoid paragraph wrapping in interstitial spaces
                // leading to fostered content.
 
                if ( $openMatch || $closeMatch ) {
                    $pendingPTag = false;
                    // Only close the paragraph if we're not inside a <pre> tag, or if
                    // that <pre> tag has just been opened
                    if ( !$this->inPre || $preOpenMatch ) {
                        // @todo T7718: paragraph closed
                        $output .= $this->closeParagraph();
                    }
                    if ( $preOpenMatch && !$preCloseMatch ) {
                        $this->inPre = true;
                    }
                    $bqOffset = 0;
                    while ( preg_match( '/<(\\/?)blockquote[\s>]/i', $t,
                        $bqMatch, PREG_OFFSET_CAPTURE, $bqOffset )
                    ) {
                        $inBlockquote = !$bqMatch[1][0]; // is this a close tag?
                        $bqOffset = $bqMatch[0][1] + strlen( $bqMatch[0][0] );
                    }
                    $inBlockElem = !$closeMatch;
                } elseif ( !$inBlockElem && !$this->inPre ) {
                    if ( ' ' == substr( $t, 0, 1 )
                        && ( $this->lastSection === 'pre' || trim( $t ) != '' )
                        && !$inBlockquote
                    ) {
                        # pre
                        if ( $this->lastSection !== 'pre' ) {
                            $pendingPTag = false;
                            $output .= $this->closeParagraph() . '<pre>';
                            $this->lastSection = 'pre';
                        }
                        $t = substr( $t, 1 );
                    } elseif ( preg_match( '/^(?:<style\\b[^>]*>.*?<\\/style>\s*|<link\\b[^>]*>\s*)+$/iS', $t ) ) {
                        # T186965: <style> or <link> by itself on a line shouldn't open or close paragraphs.
                        # But it should clear $pendingPTag.
                        if ( $pendingPTag ) {
                            $output .= $this->closeParagraph();
                            $pendingPTag = false;
                            $this->lastSection = '';
                        }
                    } else {
                        # paragraph
                        if ( trim( $t ) === '' ) {
                            if ( $pendingPTag ) {
                                $output .= $pendingPTag . '<br />';
                                $pendingPTag = false;
                                $this->lastSection = 'p';
                            } else {
                                if ( $this->lastSection !== 'p' ) {
                                    $output .= $this->closeParagraph();
                                    $this->lastSection = '';
                                    $pendingPTag = '<p>';
                                } else {
                                    $pendingPTag = '</p><p>';
                                }
                            }
                        } else {
                            if ( $pendingPTag ) {
                                $output .= $pendingPTag;
                                $pendingPTag = false;
                                $this->lastSection = 'p';
                            } elseif ( $this->lastSection !== 'p' ) {
                                $output .= $this->closeParagraph() . '<p>';
                                $this->lastSection = 'p';
                            }
                        }
                    }
                }
            }
            # somewhere above we forget to get out of pre block (T2785)
            if ( $preCloseMatch && $this->inPre ) {
                $this->inPre = false;
            }
            if ( $pendingPTag === false ) {
                if ( $prefixLength === 0 ) {
                    $output .= $t;
                    $output .= "\n";
                } else {
                    // Trim whitespace in list items
                    $output .= trim( $t );
                }
            }
        }
        while ( $prefixLength ) {
            $output .= $this->closeList( $prefix2[$prefixLength - 1] );
            --$prefixLength;
            if ( !$prefixLength ) {
                $output .= "\n";
            }
        }
        if ( $this->lastSection !== '' ) {
            $output .= '</' . $this->lastSection . '>';
            $this->lastSection = '';
        }
 
        return $output;
    }
 
    private function findColonNoLinks( $str, &$before, &$after ) {
        if ( !preg_match( '/:|<|-\{/', $str, $m, PREG_OFFSET_CAPTURE ) ) {
            # Nothing to find!
            return false;
        }
 
        if ( $m[0][0] === ':' ) {
            # Easy; no tag nesting to worry about
            $colonPos = $m[0][1];
            $before = substr( $str, 0, $colonPos );
            $after = substr( $str, $colonPos + 1 );
            return $colonPos;
        }
 
        # Ugly state machine to walk through avoiding tags.
        $state = self::COLON_STATE_TEXT;
        $ltLevel = 0;
        $lcLevel = 0;
        $len = strlen( $str );
        for ( $i = $m[0][1]; $i < $len; $i++ ) {
            $c = $str[$i];
 
            switch ( $state ) {
                case self::COLON_STATE_TEXT:
                    switch ( $c ) {
                        case "<":
                            # Could be either a <start> tag or an </end> tag
                            $state = self::COLON_STATE_TAGSTART;
                            break;
                        case ":":
                            if ( $ltLevel === 0 ) {
                                # We found it!
                                $before = substr( $str, 0, $i );
                                $after = substr( $str, $i + 1 );
                                return $i;
                            }
                            # Embedded in a tag; don't break it.
                            break;
                        default:
                            # Skip ahead looking for something interesting
                            if ( !preg_match( '/:|<|-\{/', $str, $m, PREG_OFFSET_CAPTURE, $i ) ) {
                                # Nothing else interesting
                                return false;
                            }
                            if ( $m[0][0] === '-{' ) {
                                $state = self::COLON_STATE_LC;
                                $lcLevel++;
                                $i = $m[0][1] + 1;
                            } else {
                                # Skip ahead to next interesting character.
                                $i = $m[0][1] - 1;
                            }
                            break;
                    }
                    break;
                case self::COLON_STATE_LC:
                    # In language converter markup -{ ... }-
                    if ( !preg_match( '/-\{|\}-/', $str, $m, PREG_OFFSET_CAPTURE, $i ) ) {
                        # Nothing else interesting to find; abort!
                        # We're nested in language converter markup, but there
                        # are no close tags left.  Abort!
                        break 2;
                    } elseif ( $m[0][0] === '-{' ) {
                        $i = $m[0][1] + 1;
                        $lcLevel++;
                    } elseif ( $m[0][0] === '}-' ) {
                        $i = $m[0][1] + 1;
                        $lcLevel--;
                        if ( $lcLevel === 0 ) {
                            $state = self::COLON_STATE_TEXT;
                        }
                    }
                    break;
                case self::COLON_STATE_TAG:
                    # In a <tag>
                    switch ( $c ) {
                        case ">":
                            $ltLevel++;
                            $state = self::COLON_STATE_TEXT;
                            break;
                        case "/":
                            # Slash may be followed by >?
                            $state = self::COLON_STATE_TAGSLASH;
                            break;
                        default:
                            # ignore
                    }
                    break;
                case self::COLON_STATE_TAGSTART:
                    switch ( $c ) {
                        case "/":
                            $state = self::COLON_STATE_CLOSETAG;
                            break;
                        case "!":
                            $state = self::COLON_STATE_COMMENT;
                            break;
                        case ">":
                            # Illegal early close? This shouldn't happen D:
                            $state = self::COLON_STATE_TEXT;
                            break;
                        default:
                            $state = self::COLON_STATE_TAG;
                    }
                    break;
                case self::COLON_STATE_CLOSETAG:
                    # In a </tag>
                    if ( $c === ">" ) {
                        if ( $ltLevel > 0 ) {
                            $ltLevel--;
                        } else {
                            # ignore the excess close tag, but keep looking for
                            # colons. (This matches Parsoid behavior.)
                            wfDebug( __METHOD__ . ": Invalid input; too many close tags\n" );
                        }
                        $state = self::COLON_STATE_TEXT;
                    }
                    break;
                case self::COLON_STATE_TAGSLASH:
                    if ( $c === ">" ) {
                        # Yes, a self-closed tag <blah/>
                        $state = self::COLON_STATE_TEXT;
                    } else {
                        # Probably we're jumping the gun, and this is an attribute
                        $state = self::COLON_STATE_TAG;
                    }
                    break;
                case self::COLON_STATE_COMMENT:
                    if ( $c === "-" ) {
                        $state = self::COLON_STATE_COMMENTDASH;
                    }
                    break;
                case self::COLON_STATE_COMMENTDASH:
                    if ( $c === "-" ) {
                        $state = self::COLON_STATE_COMMENTDASHDASH;
                    } else {
                        $state = self::COLON_STATE_COMMENT;
                    }
                    break;
                case self::COLON_STATE_COMMENTDASHDASH:
                    if ( $c === ">" ) {
                        $state = self::COLON_STATE_TEXT;
                    } else {
                        $state = self::COLON_STATE_COMMENT;
                    }
                    break;
                default:
                    throw new MWException( "State machine error in " . __METHOD__ );
            }
        }
        if ( $ltLevel > 0 || $lcLevel > 0 ) {
            wfDebug(
                __METHOD__ . ": Invalid input; not enough close tags " .
                "(level $ltLevel/$lcLevel, state $state)\n"
            );
            return false;
        }
        return false;
    }
}