Fix the PHP language

This patch is an attempt to fix the PHP language, when it is combined
with markup.

The problem is, that markup has a higher priority than all other tokens.
This leads to weird errors, where HTML tags are highlighted inside of
comments PrismJS#197. One solution to this was to set the `greedy` flag for the
comment token, but this leads to far worse errors like PrismJS#1097.

This patch should fix both issues PrismJS#197 and PrismJS#1097, by switching the
grammar to markup on the fly. One potential problem is, that it relies
on the `<?php` tag to detect if markup is present. So if a PHP file
contains only markup and no PHP code at all, the result will look
broken.

components/prism-php.js

@@ -16,8 +16,7 @@ Prism.languages.php = Prism.languages.extend('clike', {
 	'constant': /\b[A-Z0-9_]{2,}\b/,
 	'comment': {
 		pattern: /(^|[^\\])(?:\/\*[\w\W]*?\*\/|\/\/.*)/,
-		lookbehind: true,
-		greedy: true
+		lookbehind: true
 	}
 });
 
@@ -32,7 +31,10 @@ Prism.languages.insertBefore('php', 'class-name', {
 });
 
 Prism.languages.insertBefore('php', 'keyword', {
-	'delimiter': /\?>|<\?(?:php)?/i,
+	'delimiter': {
+		pattern: /\?>|<\?(?:php)?/i,
+		alias: 'important'
+	},
 	'variable': /\$\w+\b/i,
 	'package': {
 		pattern: /(\\|namespace\s+|use\s+)[\w\\]+/,
@@ -51,61 +53,54 @@ Prism.languages.insertBefore('php', 'operator', {
 	}
 });
 
-// Add HTML support of the markup language exists
+// Add HTML support if the markup language exists
 if (Prism.languages.markup) {
 
 	// Tokenize all inline PHP blocks that are wrapped in <?php ?>
 	// This allows for easy PHP + markup highlighting
 	Prism.hooks.add('before-highlight', function(env) {
-		if (env.language !== 'php') {
+		if (env.language !== 'php' || !/(?:<\?php|<\?)/ig.test(env.code)) {
 			return;
 		}
 
 		env.tokenStack = [];
 
 		env.backupCode = env.code;
-		env.code = env.code.replace(/(?:<\?php|<\?)[\w\W]*?(?:\?>)/ig, function(match) {
+		env.code = env.code.replace(/(?:<\?php|<\?)[\w\W]*?(?:\?>|$)/ig, function(match) {
 			env.tokenStack.push(match);
 
 			return '{{{PHP' + env.tokenStack.length + '}}}';
 		});
+
+		// Switch the grammar to markup
+		env.grammar = Prism.languages.markup;
 	});
 
 	// Restore env.code for other plugins (e.g. line-numbers)
 	Prism.hooks.add('before-insert', function(env) {
-		if (env.language === 'php') {
+		if (env.language === 'php' && env.backupCode) {
 			env.code = env.backupCode;
 			delete env.backupCode;
 		}
 	});
 
 	// Re-insert the tokens after highlighting
 	Prism.hooks.add('after-highlight', function(env) {
-		if (env.language !== 'php') {
+		if (env.language !== 'php' || !env.tokenStack) {
 			return;
 		}
 
+		// Switch the grammar back
+		env.grammar = Prism.languages.php;
+
 		for (var i = 0, t; t = env.tokenStack[i]; i++) {
 			// The replace prevents $$, $&, $`, $', $n, $nn from being interpreted as special patterns
-			env.highlightedCode = env.highlightedCode.replace('{{{PHP' + (i + 1) + '}}}', Prism.highlight(t, env.grammar, 'php').replace(/\$/g, '$$$$'));
+			env.highlightedCode = env.highlightedCode.replace('{{{PHP' + (i + 1) + '}}}',
+					"<span class=\"token php\">" +
+					Prism.highlight(t, env.grammar, 'php').replace(/\$/g, '$$$$') +
+					"</span>");
 		}
 
 		env.element.innerHTML = env.highlightedCode;
 	});
-
-	// Wrap tokens in classes that are missing them
-	Prism.hooks.add('wrap', function(env) {
-		if (env.language === 'php' && env.type === 'markup') {
-			env.content = env.content.replace(/(\{\{\{PHP[0-9]+\}\}\})/g, "<span class=\"token php\">$1</span>");
-		}
-	});
-
-	// Add the rules before all others
-	Prism.languages.insertBefore('php', 'comment', {
-		'markup': {
-			pattern: /<[^?]\/?(.*?)>/,
-			inside: Prism.languages.markup
-		},
-		'php': /\{\{\{PHP[0-9]+\}\}\}/
-	});
 }

tests/helper/test-case.js

@@ -175,7 +175,7 @@ module.exports = {
 					code: code
 				};
 				Prism.hooks.run('before-highlight', env);
-				env.highlightedCode = Prism.highlight(env.code, Prism.languages[usedLanguages.mainLanguage], usedLanguages.mainLanguage);
+				env.highlightedCode = Prism.highlight(env.code, env.grammar, env.language);
 				Prism.hooks.run('before-insert', env);
 				env.element.innerHTML = env.highlightedCode;
 				Prism.hooks.run('after-highlight', env);

tests/languages/php/comment_feature.test

@@ -3,14 +3,16 @@
 /**/
 /* foo
 bar */
+/* <me@example.com> */
 
 ----------------------------------------------------
 
 [
 	["comment", "//"],
 	["comment", "// foobar"],
 	["comment", "/**/"],
-	["comment", "/* foo\r\nbar */"]
+	["comment", "/* foo\r\nbar */"],
+	["comment", "/* <me@example.com> */"]
 ]
 
 ----------------------------------------------------

tests/languages/php/string_feature.test

@@ -0,0 +1,13 @@
+"https://example.com"
+" /* not a comment */ "
+
+----------------------------------------------------
+
+[
+	["string", "\"https://example.com\""],
+	["string", "\" /* not a comment */ \""]
+]
+
+----------------------------------------------------
+
+Checks for strings with comments.