84 lines
2.5 KiB
JavaScript
84 lines
2.5 KiB
JavaScript
(function () {
|
|
|
|
if (typeof Prism === 'undefined' || !Prism.languages['diff']) {
|
|
return;
|
|
}
|
|
|
|
|
|
var LANGUAGE_REGEX = /diff-([\w-]+)/i;
|
|
var HTML_TAG = /<\/?(?!\d)[^\s>\/=$<%]+(?:\s(?:\s*[^\s>\/=]+(?:\s*=\s*(?:"[^"]*"|'[^']*'|[^\s'">=]+(?=[\s>]))|(?=[\s/>])))+)?\s*\/?>/gi;
|
|
//this will match a line plus the line break while ignoring the line breaks HTML tags may contain.
|
|
var HTML_LINE = RegExp(/(?:__|[^\r\n<])*(?:\r\n?|\n|(?:__|[^\r\n<])(?![^\r\n]))/.source.replace(/__/g, function () { return HTML_TAG.source; }), 'gi');
|
|
|
|
var PREFIXES = Prism.languages.diff.PREFIXES;
|
|
|
|
|
|
Prism.hooks.add('before-sanity-check', function (env) {
|
|
var lang = env.language;
|
|
if (LANGUAGE_REGEX.test(lang) && !env.grammar) {
|
|
env.grammar = Prism.languages[lang] = Prism.languages['diff'];
|
|
}
|
|
});
|
|
Prism.hooks.add('before-tokenize', function (env) {
|
|
var lang = env.language;
|
|
if (LANGUAGE_REGEX.test(lang) && !Prism.languages[lang]) {
|
|
Prism.languages[lang] = Prism.languages['diff'];
|
|
}
|
|
});
|
|
|
|
Prism.hooks.add('wrap', function (env) {
|
|
var diffLanguage, diffGrammar;
|
|
|
|
if (env.language !== 'diff') {
|
|
var langMatch = LANGUAGE_REGEX.exec(env.language);
|
|
if (!langMatch) {
|
|
return; // not a language specific diff
|
|
}
|
|
|
|
diffLanguage = langMatch[1];
|
|
diffGrammar = Prism.languages[diffLanguage];
|
|
}
|
|
|
|
// one of the diff tokens without any nested tokens
|
|
if (env.type in PREFIXES) {
|
|
/** @type {string} */
|
|
var content = env.content.replace(HTML_TAG, ''); // remove all HTML tags
|
|
|
|
/** @type {string} */
|
|
var decoded = content.replace(/</g, '<').replace(/&/g, '&');
|
|
|
|
// remove any one-character prefix
|
|
var code = decoded.replace(/(^|[\r\n])./g, '$1');
|
|
|
|
// highlight, if possible
|
|
var highlighted;
|
|
if (diffGrammar) {
|
|
highlighted = Prism.highlight(code, diffGrammar, diffLanguage);
|
|
} else {
|
|
highlighted = Prism.util.encode(code);
|
|
}
|
|
|
|
// get the HTML source of the prefix token
|
|
var prefixToken = new Prism.Token('prefix', PREFIXES[env.type], [/\w+/.exec(env.type)[0]]);
|
|
var prefix = Prism.Token.stringify(prefixToken, env.language);
|
|
|
|
// add prefix
|
|
var lines = [], m;
|
|
HTML_LINE.lastIndex = 0;
|
|
while (m = HTML_LINE.exec(highlighted)) {
|
|
lines.push(prefix + m[0]);
|
|
}
|
|
if (/(?:^|[\r\n]).$/.test(decoded)) {
|
|
// because both "+a\n+" and "+a\n" will map to "a\n" after the line prefixes are removed
|
|
lines.push(prefix);
|
|
}
|
|
env.content = lines.join('');
|
|
|
|
if (diffGrammar) {
|
|
env.classes.push('language-' + diffLanguage);
|
|
}
|
|
}
|
|
});
|
|
|
|
}());
|