WordPress · dmsnell · Feb 7, 2024 · Feb 7, 2024 · Feb 7, 2024 · Feb 7, 2024
diff --git a/highlight-html.php b/highlight-html.php
@@ -0,0 +1,176 @@
+<?php
+
+require_once __DIR__ . '/src/wp-load.php';
+
+define( 'C_TEXT', "\e[m" );
+define( 'C_REF', "\e[38;2;154;110;58m" );
+define( 'C_COMMENT', "\e[38;2;112;128;144m" );
+define( 'C_SYNTAX', "\e[38;2;153;153;153m" );
+define( 'C_TAGNAME', "\e[38;2;153;0;85m" );
+define( 'C_ANAME', "\e[38;2;102;153;0m" );
+define( 'C_AVALUE', "\e[38;2;0;119;170m" );
+
+$options = getopt( 'fu:' ); // How rude!
+$do_format = isset( $options['f'] );
+
+$uri = 'php://stdin';
+if ( isset( $options['u'] ) ) {
+	$uri = $options['u'];
+	if ( ! preg_match( '~^https?://~', $uri ) ) {
+		$uri = "https://{$uri}";
+	}
+}
+
+$html = file_get_contents( $uri );
+
+$p = new class( $html ) extends WP_HTML_Processor {
+	public function get_raw_token() {
+		$this->set_bookmark('here');
+		$here = $this->bookmarks['_here'];
+		return substr( $this->html, $here->start, $here->length );
+	}
+};
+
+$p = $p::create_Full_parser( $html );
+
+while ( $p->next_token() ) {
+	switch ( $p->get_token_type() ) {
+	case '#comment':
+		echo C_COMMENT . '<!--' . $p->get_modifiable_text() . '-->';
+		break;
+
+	case '#doctype':
+		echo C_SYNTAX . '<!DOCTYPE' . $p->get_modifiable_text() . '>';
+		break;
+
+	case '#tag':
+		print_tag( $p );
+		break;
+
+	case '#text':
+		print_text( $p );
+		break;
+
+	default:
+		die( "Unsupported syntax: {$p->get_token_type()}" );
+	}
+}
+
+echo "\e[m\n";
+
+function print_text( $p ) {
+	$token_name = $p->get_token_name();
+	if ( in_array( $token_name, [ 'IFRAME', 'NOEMBED', 'NOFRAMES', 'SCRIPT', 'STYLE', 'TEXTAREA', 'TITLE', 'XMP' ], true ) ) {
+		return C_TEXT . $p->get_modifiable_text();
+	}
+
+	$raw_token = $p->get_raw_token();
+	$text      = C_TEXT;
+	$at        = 0;
+	$was_at    = 0;
+	while ( $at < strlen( $raw_token ) ) {
+		$next_at = strpos( $raw_token, '&', $at );
+		if ( false === $next_at ) {
+			break;
+		}
+
+		$replacement = WP_HTML_Decoder::read_character_reference( 'data', $raw_token, $next_at, $skip_bytes );
+		if ( isset( $replacement ) ) {
+			$text  .= substr( $raw_token, $was_at, $next_at - $was_at ) . C_REF . substr( $raw_token, $next_at, $skip_bytes ) . C_TEXT;
+			$at     = $next_at + $skip_bytes;
+			$was_at = $at;
+			continue;
+		}
+
+		++$at;
+	}
+	if ( $was_at < strlen( $raw_token ) ) {
+		$text .= substr( $raw_token, $was_at );
+	}
+	echo C_TEXT . $text;
+}
+
+function print_tag( $p ) {
+	global $do_format;
+
+	static $depth = 0;
+
+	$tag_name  = $p->get_tag();
+	$is_closer = $p->is_tag_closer();
+	$closer    = $is_closer ? '/' : '';
+	$is_void   = WP_HTML_Processor::is_void( $tag_name );
+	$voider    = $is_void ? '/' : '';
+
+	if ( $is_closer && in_array( $tag_name, [ 'HEAD', 'BODY', 'OL', 'UL', 'DIV' ], true ) ) {
+		$depth--;
+	}
+
+	$indent = str_pad( '', $depth * 2, ' ' );
+
+	if ( $do_format && (
+		(
+			! $is_closer && in_array( $tag_name, [
+				'DIV', 'P', 'UL', 'OL', 'DETAILS', 'SVG', 'PATH', 'G',
+				'LINK', 'META', 'HTML', 'HEAD', 'BODY', 'TITLE', 'TEXTAREA',
+				'PRE', 'H1', 'H2', 'H3', 'H4', 'H5', 'H6', 'HGROUP',
+				'PICTURE', 'SOURCE', 'FIGURE', 'FORM', 'TABLE', 'TR',
+				'FIGCAPTION', 'BLOCKQUOTE', 'OBJECT', 'EMBED', 'IFRAME',
+				'SCRIPT', 'STYLE', 'NOSCRIPT', 'NAV', 'LI'
+			], true )
+		) || (
+			$is_closer && in_array( $tag_name, [
+				'HEAD', 'HTML', 'BODY', 'PICTURE', 'FIGURE', 'TABLE'
+			], true  )
+		)
+	) ) {
+		echo "\n{$indent}";
+	}
+	echo C_SYNTAX . '<' . $closer;
+
+	echo C_TAGNAME . strtolower( $p->get_tag() );
+	$attributes = $p->get_attribute_names_with_prefix( '' ) ?? array();
+
+	foreach( $attributes as $name ) {
+		$value = $p->get_attribute( $name );
+
+		echo ' ' . C_ANAME . $name;
+		if ( true === $value ) {
+			continue;
+		}
+
+		echo C_SYNTAX . '="';
+		echo C_AVALUE . str_replace( '"', '&quot;', $value );
+		echo C_SYNTAX . '"';
+	}
+	echo C_SYNTAX . '>';
+
+	$text = $p->get_modifiable_text();
+	if ( ! empty( $text ) ) {
+		echo 'TITLE' === $p->get_tag() ? C_TEXT : C_COMMENT;
+
+		$add_newlines = (
+			$do_format &&
+			strlen( trim( $text ) ) > 0 &&
+			(
+				'SCRIPT' === $tag_name ||
+				'STYLE' === $tag_name ||
+				'TEXTAREA' === $tag_name ||
+				'PRE' === $tag_name
+			)
+		);
+
+		if ( $add_newlines ) {
+			echo "\n" . trim( $text, "\n" ) . "\n";
+		} else {
+			echo $text;
+		}
+
+		echo C_SYNTAX . '</' . C_TAGNAME . strtolower( $p->get_tag() ) . C_SYNTAX . '>';
+	} elseif ( in_array( $tag_name, [ 'SCRIPT', 'STYLE', 'TEXTAREA', 'PRE' ], true ) ) {
+		echo C_SYNTAX . '</' . C_TAGNAME . strtolower( $p->get_tag() ) . C_SYNTAX . '>';
+	}
+
+	if ( ! $is_closer && in_array( $tag_name, [ 'HEAD', 'BODY', 'OL', 'UL', 'DIV' ], true ) ) {
+		$depth++;
+	}
+}
diff --git a/html-grep.php b/html-grep.php
@@ -0,0 +1,187 @@
+<?php
+
+require_once( __DIR__ . '/src/wp-load.php' );
+
+function main() {
+	global $argv;
+
+	$opts = getopt( 'A:B:p:i:m:', [] );
+
+	if ( ! isset( $opts['p'] ) ) {
+		die( 'Please supply a search pattern with -p, e.g. `-p "[a-f0-9]+"`' );
+	}
+
+	if ( ! isset( $opts['i'] ) && ! in_array( '-', $argv, true ) ) {
+		die( 'Please specify input filename with -i or use stdin with -, e.g. `-i file.html`' );
+	}
+
+	$lines_before = ctype_digit( $opts['B'] ?? '' ) ? intval( $opts['B'] ) : 0;
+	$lines_after = ctype_digit( $opts['A'] ?? '' ) ? intval( $opts['A'] ) : 0;
+
+	$max = ( isset( $opts['m'] ) && ctype_digit( $opts['m'] ) && (int) $opts['m'] > 0 )
+		? (int) $opts['m']
+		: 1;
+
+	$input = in_array( '-', $argv, true ) ? 'php://stdin' : $opts['i'];
+	Grepper::scan( $input, $opts['p'], $lines_before, $lines_after, $max );
+}
+
+class Debugger extends WP_HTML_Tag_Processor {
+	public function h() {
+		return $this->html;
+	}
+
+	public function extend( $line ) {
+		$this->html .= $line;
+
+		if (
+			$this->parser_state === self::STATE_COMPLETE ||
+			$this->parser_state === self::STATE_INCOMPLETE_INPUT
+		) {
+			$this->parser_state = self::STATE_READY;
+		}
+	}
+
+	public function next_token() {
+		$r = parent::next_token();
+		$this->set_bookmark( 'here' );
+		return $r;
+	}
+
+	public function at() {
+		return $this->bookmarks['here'];
+	}
+}
+
+class Grepper {
+	public static function scan( $input, $pattern, $before, $after, $max ) {
+		$f         = fopen( $input, 'r' );
+		$c         = 0;
+		$n         = 0;
+		$lines     = [];
+		$lc        = 1 + $before + $after;
+		$o         = static function ( $s ) { return WP_HTML_Decoder::decode_text_node( $s ); };
+		$ws        = static function ( $s ) { return preg_replace( '~[ \r\f\t\n]+~', ' ', $s ); };
+		$pre_depth = 0;
+		$p         = new Debugger( '' );
+		$t         = '';
+
+		while ( false !== ( $line = fgets( $f ) ) ) {
+			$n++;
+
+			$p->extend( $line );
+			while ( $p->next_token() ) {
+				$at        = $p->at();
+				$type      = $p->get_token_type();
+				$node_text = $o( $p->get_modifiable_text() );
+				$node_text = $pre_depth > 0 ? $node_text : $ws( $node_text );
+
+				if ( '#tag' !== $type && '#text' !== $type ) {
+					continue;
+				}
+
+				switch ( $p->get_token_name() ) {
+					case 'PRE':
+						$pre_depth += $p->is_tag_closer() ? -1 : 1;
+						break;
+
+					case '#text':
+						$t .= $node_text;
+				}
+
+				if ( preg_match( $pattern, $t, $match, PREG_OFFSET_CAPTURE ) ) {
+					$h = (
+						"\e[32m" .
+						ltrim( mb_strcut( $t, 0, $match[0][1] ) ) .
+						"\e[33m" .
+						$match[0][0] .
+						"\e[32m" .
+						rtrim( mb_strcut( $t, $match[0][1] + strlen( $match[0][0] ) ) ) .
+						"\e[90m"
+					);
+
+					for ( $i = 0; $i < $after; $i++ ) {
+						$line = fgets( $f );
+						if ( false !== $line ) {
+							$p->extend( $line );
+						}
+					}
+
+					$cb = mb_strcut( $p->h(), 0, $at->start );
+					$cc = mb_strcut( $p->h(), $at->start, $at->length );
+					$ca = mb_strcut( $p->h(), $at->start + $at->length );
+
+					// Limit context to N lines preview
+					$cb = explode( "\n", $cb );
+					$cb = array_slice( $cb, -$before );
+					$cb = mb_strcut( implode( "\n", $cb ), -$before * 80 );
+
+					// Limit context to N lines preview
+					$ca = explode( "\n", $ca );
+					$ca = array_slice( $ca, 0, $after );
+					$ca = mb_strcut( implode( "\n", $ca ), 0, $after * 80 );
+
+					// If contained in last node.
+					$tt = $p->get_modifiable_text();
+					if ( preg_match( $pattern, $tt, $mm, PREG_OFFSET_CAPTURE ) ) {
+						$cc = (
+							"\e[90m" .
+							mb_strcut( $tt, 0, $mm[0][1] ) .
+							"\e[33m" .
+							$mm[0][0] .
+							"\e[90m" .
+							mb_strcut( $tt, $mm[0][1] + strlen( $mm[0][0] ) )
+						);
+					}
+
+					echo "\n\e[32m{$n}\e[90m: \e[31m{$p->get_token_name()} \e[90m{$h}\e[m\n";
+					echo "\e[90m{$cb}\e[33m{$cc}\e[90m{$ca}\e[m";
+
+					if ( ++$c >= $max ) {
+						fclose( $f );
+						exit;
+					}
+
+					$t = '';
+				}
+
+				$t = mb_strcut( $t, -100 );
+			}
+
+		}
+	}
+
+	public static function indent( $lines ) {
+		return implode( "\n", array_map(
+			static function ( $line ) { return '    ' . $line; },
+			explode( "\n", $lines )
+		) );
+	}
+}
+
+main();
+
+function is_line_breaker( $tag_name ) {
+	switch ( $tag_name ) {
+		case 'BLOCKQUOTE':
+		case 'BR':
+		case 'DD':
+		case 'DIV':
+		case 'DL':
+		case 'DT':
+		case 'H1':
+		case 'H2':
+		case 'H3':
+		case 'H4':
+		case 'H5':
+		case 'H6':
+		case 'HR':
+		case 'LI':
+		case 'OL':
+		case 'P':
+		case 'UL':
+			return true;
+	}
+
+	return false;
+}