@extractus/article-extractor

(forked version for use in workers that do not have window.DOMParser defined.)

Extract main article, main image and meta data from URL.

(This library is derived from article-parser renamed.)

Demo

• Give it a try!
• Example FaaS

Install & Usage

Node.js

1npm i @extractus/article-extractor
2
3# pnpm
4pnpm i @extractus/article-extractor
5
6# yarn
7yarn add @extractus/article-extractor

1// es6 module
2import { extract } from '@extractus/article-extractor'

Deno

1import { extract } from 'https://esm.sh/@extractus/article-extractor'
2
3// deno > 1.28
4import { extract } from 'npm:@extractus/article-extractor'

Browser

1import { extract } from 'https://esm.sh/@extractus/article-extractor'

Please check the examples for reference.

APIs

• extract()
• extractFromHtml()
• Transformations
  • transformation object
  • .addTransformations
  • .removeTransformations
  • Priority order
• sanitize-html's options

---

extract()

Load and extract article data. Return a Promise object.

Syntax

1extract(String input)
2extract(String input, Object parserOptions)
3extract(String input, Object parserOptions, Object fetchOptions)

Example:

1import { extract } from '@extractus/article-extractor'
2
3const input = 'https://www.cnbc.com/2022/09/21/what-another-major-rate-hike-by-the-federal-reserve-means-to-you.html'
4
5// here we use top-level await, assume current platform supports it
6try {
7  const article = await extract(input)
8  console.log(article)
9} catch (err) {
10  console.error(err)
11}

The result - article - can be null or an object with the following structure:

1{
2  url: String,
3  title: String,
4  description: String,
5  image: String,
6  author: String,
7  favicon: String,
8  content: String,
9  published: Date String,
10  type: String, // page type
11  source: String, // original publisher
12  links: Array, // list of alternative links
13  ttr: Number, // time to read in second, 0 = unknown
14}

Parameters

input required

URL string links to the article or HTML content of that web page.

parserOptions optional

Object with all or several of the following properties:

• wordsPerMinute: Number, to estimate time to read. Default 300.
• descriptionTruncateLen: Number, max num of chars generated for description. Default 210.
• descriptionLengthThreshold: Number, min num of chars required for description. Default 180.
• contentLengthThreshold: Number, min num of chars required for content. Default 200.

For example:

1import { extract } from '@extractus/article-extractor'
2
3const article = await extract('https://www.cnbc.com/2022/09/21/what-another-major-rate-hike-by-the-federal-reserve-means-to-you.html', {
4  descriptionLengthThreshold: 120,
5  contentLengthThreshold: 500
6})
7
8console.log(article)

fetchOptions optional

fetchOptions is an object that can have the following properties:

• headers: to set request headers
• proxy: another endpoint to forward the request to
• agent: a HTTP proxy agent
• signal: AbortController signal or AbortSignal timeout to terminate the request

For example, you can use this param to set request headers to fetch as below:

1import { extract } from '@extractus/article-extractor'
2
3const url = 'https://www.cnbc.com/2022/09/21/what-another-major-rate-hike-by-the-federal-reserve-means-to-you.html'
4const article = await extract(url, {}, {
5  headers: {
6    'user-agent': 'Opera/9.60 (Windows NT 6.0; U; en) Presto/2.1.1'
7  }
8})
9
10console.log(article)

You can also specify a proxy endpoint to load remote content, instead of fetching directly.

For example:

1import { extract } from '@extractus/article-extractor'
2
3const url = 'https://www.cnbc.com/2022/09/21/what-another-major-rate-hike-by-the-federal-reserve-means-to-you.html'
4
5await extract(url, {}, {
6  headers: {
7    'user-agent': 'Opera/9.60 (Windows NT 6.0; U; en) Presto/2.1.1'
8  },
9  proxy: {
10    target: 'https://your-secret-proxy.io/loadXml?url=',
11    headers: {
12      'Proxy-Authorization': 'Bearer YWxhZGRpbjpvcGVuc2VzYW1l...'
13    },
14  }
15})

Passing requests to proxy is useful while running @extractus/article-extractor on browser. View examples/browser-article-parser as reference example.

For more info about proxy authentication, please refer HTTP authentication

For a deeper customization, you can consider using Proxy to replace fetch behaviors with your own handlers.

Another way to work with proxy is use agent option instead of proxy as below:

1import { extract } from '@extractus/article-extractor'
2
3import { HttpsProxyAgent } from 'https-proxy-agent'
4
5const proxy = 'http://abc:RaNdoMpasswORd_country-France@proxy.packetstream.io:31113'
6
7const url = 'https://www.cnbc.com/2022/09/21/what-another-major-rate-hike-by-the-federal-reserve-means-to-you.html'
8
9const article = await extract(url, {}, {
10  agent: new HttpsProxyAgent(proxy),
11})
12console.log('Run article-extractor with proxy:', proxy)
13console.log(article)

For more info about https-proxy-agent, check its repo.

By default, there is no request timeout. You can use the option signal to cancel request at the right time.

The common way is to use AbortControler:

1const controller = new AbortController()
2
3// stop after 5 seconds
4setTimeout(() => {
5  controller.abort()
6}, 5000)
7
8const data = await extract(url, null, {
9  signal: controller.signal,
10})

A newer solution is AbortSignal's timeout() static method:

1// stop after 5 seconds
2const data = await extract(url, null, {
3  signal: AbortSignal.timeout(5000),
4})

For more info:

• AbortController constructor
• AbortSignal: timeout() static method

extractFromHtml()

Extract article data from HTML string. Return a Promise object as same as extract() method above.

Syntax

1extractFromHtml(String html)
2extractFromHtml(String html, String url)
3extractFromHtml(String html, String url, Object parserOptions)

Example:

1import { extractFromHtml } from '@extractus/article-extractor'
2
3const url = 'https://www.cnbc.com/2022/09/21/what-another-major-rate-hike-by-the-federal-reserve-means-to-you.html'
4
5const res = await fetch(url)
6const html = await res.text()
7
8// you can do whatever with this raw html here: clean up, remove ads banner, etc
9// just ensure a html string returned
10
11const article = await extractFromHtml(html, url)
12console.log(article)

Parameters

html required

HTML string which contains the article you want to extract.

url optional

URL string that indicates the source of that HTML content. article-extractor may use this info to handle internal/relative links.

parserOptions optional

See parserOptions above.

---

Transformations

Sometimes the default extraction algorithm may not work well. That is the time when we need transformations.

By adding some functions before and after the main extraction step, we aim to come up with a better result as much as possible.

There are 2 methods to play with transformations:

• addTransformations(Object transformation | Array transformations)
• removeTransformations(Array patterns)

At first, let's talk about transformation object.

transformation object

In @extractus/article-extractor, transformation is an object with the following properties:

• patterns: required, a list of regexps to match the URLs
• pre: optional, a function to process raw HTML
• post: optional, a function to process extracted article

Basically, the meaning of transformation can be interpreted like this:

with the urls which match these patterns
let's run pre function to normalize HTML content
then extract main article content with normalized HTML, and if success
let's run post function to normalize extracted article content

Here is an example transformation:

1{
2  patterns: [
3    /([\w]+.)?domain.tld\/*/,
4    /domain.tld\/articles\/*/
5  ],
6  pre: (document) => {
7    // remove all .advertise-area and its siblings from raw HTML content
8    document.querySelectorAll('.advertise-area').forEach((element) => {
9      if (element.nodeName === 'DIV') {
10        while (element.nextSibling) {
11          element.parentNode.removeChild(element.nextSibling)
12        }
13        element.parentNode.removeChild(element)
14      }
15    })
16    return document
17  },
18  post: (document) => {
19    // with extracted article, replace all h4 tags with h2
20    document.querySelectorAll('h4').forEach((element) => {
21      const h2Element = document.createElement('h2')
22      h2Element.innerHTML = element.innerHTML
23      element.parentNode.replaceChild(h2Element, element)
24    })
25    // change small sized images to original version
26    document.querySelectorAll('img').forEach((element) => {
27      const src = element.getAttribute('src')
28      if (src.includes('domain.tld/pics/150x120/')) {
29        const fullSrc = src.replace('/pics/150x120/', '/pics/original/')
30        element.setAttribute('src', fullSrc)
31      }
32    })
33    return document
34  }
35}

• To write better transformation logic, please refer linkedom and Document Object.

addTransformations(Object transformation | Array transformations)

Add a single transformation or a list of transformations. For example:

1import { addTransformations } from '@extractus/article-extractor'
2
3addTransformations({
4  patterns: [
5    /([\w]+.)?abc.tld\/*/
6  ],
7  pre: (document) => {
8    // do something with document
9    return document
10  },
11  post: (document) => {
12    // do something with document
13    return document
14  }
15})
16
17addTransformations([
18  {
19    patterns: [
20      /([\w]+.)?def.tld\/*/
21    ],
22    pre: (document) => {
23      // do something with document
24      return document
25    },
26    post: (document) => {
27      // do something with document
28      return document
29    }
30  },
31  {
32    patterns: [
33      /([\w]+.)?xyz.tld\/*/
34    ],
35    pre: (document) => {
36      // do something with document
37      return document
38    },
39    post: (document) => {
40      // do something with document
41      return document
42    }
43  }
44])

The transformations without patterns will be ignored.

removeTransformations(Array patterns)

To remove transformations that match the specific patterns.

For example, we can remove all added transformations above:

1import { removeTransformations } from '@extractus/article-extractor'
2
3removeTransformations([
4  /([\w]+.)?abc.tld\/*/,
5  /([\w]+.)?def.tld\/*/,
6  /([\w]+.)?xyz.tld\/*/
7])

Calling removeTransformations() without parameter will remove all current transformations.

Priority order

While processing an article, more than one transformation can be applied.

Suppose that we have the following transformations:

1[
2  {
3    patterns: [
4      /http(s?):\/\/google.com\/*/,
5      /http(s?):\/\/goo.gl\/*/
6    ],
7    pre: function_one,
8    post: function_two
9  },
10  {
11    patterns: [
12      /http(s?):\/\/goo.gl\/*/,
13      /http(s?):\/\/google.inc\/*/
14    ],
15    pre: function_three,
16    post: function_four
17  }
18]

As you can see, an article from goo.gl certainly matches both them.

In this scenario, @extractus/article-extractor will execute both transformations, one by one:

function_one -> function_three -> extraction -> function_two -> function_four

---

sanitize-html's options

@extractus/article-extractor uses sanitize-html to make a clean sweep of HTML content.

Here is the default options

Depending on the needs of your content system, you might want to gather some HTML tags/attributes, while ignoring others.

There are 2 methods to access and modify these options in @extractus/article-extractor.

• getSanitizeHtmlOptions()
• setSanitizeHtmlOptions(Object sanitizeHtmlOptions)

Read sanitize-html docs for more info.

---

Test

1git clone https://github.com/extractus/article-extractor.git
2cd article-extractor
3pnpm i
4pnpm test

Quick evaluation

1git clone https://github.com/extractus/article-extractor.git
2cd article-extractor
3pnpm i
4pnpm eval {URL_TO_PARSE_ARTICLE}

License

The MIT License (MIT)

Support the project

If you find value from this open source project, you can support in the following ways:

• Give it a star ⭐
• Buy me a coffee: https://paypal.me/ndaidong 🍵
• Subscribe Article Extractor service on RapidAPI 😉

Thank you.

---