luin / readability

📚 Turn any web page into a clean view
2.49k stars 312 forks source link
gbk instapaper jsdom readability

Readability

Turn any web page into a clean view. This module is based on arc90's readability project.

Features

  1. Optimized for more websites.
  2. Supporting HTML5 tags (article, section) and Microdata API.
  3. Focusing on both accuracy and performance. 4x times faster than arc90's version.
  4. Supporting encodings such as GBK and GB2312.
  5. Converting relative urls to absolute for images and links automatically (Thank Guillermo Baigorria & Tom Sutton).

Example

Before -> After

Install

$ npm install node-readability

Note that from v2.0.0, this module only works with Node.js >= 2.0. In the meantime you are still welcome to install a release in the 1.x series (by npm install node-readability@1) if you use an older Node.js version.

Usage

read(html [, options], callback)

Where

Example

var read = require('node-readability');

read('http://howtonode.org/really-simple-file-uploads', function(err, article, meta) {
  // Main Article
  console.log(article.content);
  // Title
  console.log(article.title);

  // HTML Source Code
  console.log(article.html);
  // DOM
  console.log(article.document);

  // Response Object from Request Lib
  console.log(meta);

  // Close article to clean up jsdom and prevent leaks
  article.close();
});

NB If the page has been marked with charset other than utf-8, it will be converted automatically. Charsets such as GBK, GB2312 is also supported.

Options

node-readability will pass the options to request directly. See request lib to view all available options.

node-readability has two additional options:

If true rule is valid, otherwise no. options.cleanRulers = [callback(obj, tagName)]

read(url, {
  cleanRulers: [
    function(obj, tag) {
      if(tag === 'object') {
        if(obj.getAttribute('class') === 'BrightcoveExperience') {
          return true;
        }
      }
    }
  ]}, function(err, article, response) {
    //...
  });

options.preprocess = callback(source, response, contentType, callback);

read(url, {
    preprocess: function(source, response, contentType, callback) {
      if (source.length > maxBodySize) {
        return callback(new Error('too big'));
      }
      callback(null, source);
    }
  }, function(err, article, response) {
    //...
  });

article object

content

The article content of the web page. Return false if failed.

title

The article title of the web page. It's may not same to the text in the <title> tag.

textBody

A string containing all the text found on the page

html

The original html of the web page.

document

The document of the web page generated by jsdom. You can use it to access the DOM directly (for example, article.document.getElementById('main')).

meta object

Response object from request lib. If you need to get current url after all redirect or get some headers it can be useful.

Why not Cheerio

This lib is using jsdom to parse HTML instead of cheerio because some data such as image size and element visibility isn't able to acquire when using cheerio, which will significantly affect the result.

Contributors

https://github.com/luin/node-readability/graphs/contributors

License

This code is under the Apache License 2.0. http://www.apache.org/licenses/LICENSE-2.0