This is my fork of indix/web-auto-extractor where I've merged some of the PRs from the original repo and added some of my own fixes. Published to npm as @rane/web-auto-extractor.

Web Auto Extractor

Parse semantically structured information from any HTML webpage.

Supported formats:-

• Encodings that support Schema.org vocabularies:-
  • Microdata
  • RDFa-lite
  • JSON-LD
• Random Meta tags

Popularly, many websites mark up their webpages with Schema.org vocabularies for better SEO. This library helps you parse that information to JSON.

Demo it on tonicdev

Installation

npm install @rane/web-auto-extractor

Usage

// IF CommonJS
var WAE = require('@rane/web-auto-extractor').default
// IF ES6
import WAE from '@rane/web-auto-extractor'

var parsed = WAE().parse(sampleHTML)

Let's use the following text as the sampleHTML in our example. It uses Schema.org vocabularies to structure a Product information and is encoded in microdata format.

Input

<div itemscope itemtype="http://schema.org/Product">
  <span itemprop="brand">ACME</span>
  <span itemprop="name">Executive Anvil</span>
  <img itemprop="image" src="anvil_executive.jpg" alt="Executive Anvil logo" />
  <span itemprop="description">Sleeker than ACME's Classic Anvil, the
    Executive Anvil is perfect for the business traveler
    looking for something to drop from a height.
  </span>
  Product #: <span itemprop="mpn">925872</span>
  <span itemprop="aggregateRating" itemscope itemtype="http://schema.org/AggregateRating">
    <span itemprop="ratingValue">4.4</span> stars, based on <span itemprop="reviewCount">89
      </span> reviews
  </span>

  <span itemprop="offers" itemscope itemtype="http://schema.org/Offer">
    Regular price: $179.99
    <meta itemprop="priceCurrency" content="USD" />
    $<span itemprop="price">119.99</span>
    (Sale ends <time itemprop="priceValidUntil" datetime="2020-11-05">
      5 November!</time>)
    Available from: <span itemprop="seller" itemscope itemtype="http://schema.org/Organization">
                      <span itemprop="name">Executive Objects</span>
                    </span>
    Condition: <link itemprop="itemCondition" href="http://schema.org/UsedCondition"/>Previously owned,
      in excellent condition
    <link itemprop="availability" href="http://schema.org/InStock"/>In stock! Order now!</span>
  </span>
</div>

Output

Our parsed object should look like -

{
  "microdata": {
    "Product": [
      {
        "@context": "http://schema.org/",
        "@type": "Product",
        "brand": "ACME",
        "name": "Executive Anvil",
        "image": "anvil_executive.jpg",
        "description": "Sleeker than ACME's Classic Anvil, the\n    Executive Anvil is perfect for the business traveler\n    looking for something to drop from a height.",
        "mpn": "925872",
        "aggregateRating": {
          "@context": "http://schema.org/",
          "@type": "AggregateRating",
          "ratingValue": "4.4",
          "reviewCount": "89"
        },
        "offers": {
          "@context": "http://schema.org/",
          "@type": "Offer",
          "priceCurrency": "USD",
          "price": "119.99",
          "priceValidUntil": "5 November!",
          "seller": {
            "@context": "http://schema.org/",
            "@type": "Organization",
            "name": "Executive Objects"
          },
          "itemCondition": "http://schema.org/UsedCondition",
          "availability": "http://schema.org/InStock"
        }
      }
    ]
  },
  "rdfa": {},
  "jsonld": {},
  "metatags": {
    "priceCurrency": [
      "USD",
      "USD"
    ]
  }
}

The parsed object includes four objects - microdata, rdfa, jsonld and metatags. Since the above HTML does not have any information encoded in rdfa and jsonld, those two objects are empty.

Errors

There is no guarantee against malformed content when working with live pages on the Internet. This is especially true when webmasters attempt to work with JSON-LD. You can expect to parse a page with malformed content at some point.

In the case of JSON-LD, the parser will accumulate any parse errors encountered in an errors array if at least one was encountered. Note, they will be full Node.js error objects, so be prepared to pull off the data you need.

const parsed = WAE().parse(whatWereTheyThinkingHTML)
const parseErrors = parsed.jsonld.errors(err => err.message)

Clients can take advantage of checking for the presence of the errors property and respond accordingly.

Caveat

I wouldn't call it a caveat but rather the parser is strict by design. It might not parse like expected if the HTML isn't encoded correctly, so one might assume the parser is broken.

For example, take the following HTML snippet.

<div itemscope itemtype="http://schema.org/Movie">
  <h1 itemprop="name">Ghostbusters</h1>
  <div itemprop="productionCompany" itemscope itemtype="http://schema.org/Organization">Black Rhino</div>
  <div itemprop="countryOfOrigin" itemscope itemtype="http://schema.org/Country">
    Country: <span itemprop="name" content="USA">United States</span><p>
  </div>
</div>

The problem here is the itemprop - productionCompany which is of itemtype - Organization doesn't have any itemprop as its children, in this case - name.

The parser assumes every itemtype contains an itemprop, or every typeof contains a property in case of rdfa. So the "Black Rhino" information is lost.

It'll be nice to fix this by having a non-strict mode for parsing this information. PRs are welcome.

License

MIT