Travis Maintainability Codecov npm version npm installs dependencies
- Generator based: It will only scrape as fast as you can consume the results
- Powerful HTML parsing: Uses the popular cheerio library under the hood
- Easy to test: Uses Axios to make network requests, which can be easily mocked
using npm
npm install @epegzz/node-scraper --save
using yarn
yarn add @epegzz/node-scraper
node-scraper is very minimalistic: You provide the URL of the website you want to scrape and a parser function that converts HTML into Javascript objects.
Parser functions are implemented as generators, which means they will yield results
instead of returning them. That guarantees that network requests are made only
as fast/frequent as we can consume them.
Stopping consuming the results will stop further network requests ✨
const scrape = require('@epegzz/node-scraper') // Start scraping our made-up website `https://car-list.com` and console log the results // // This will print: // { brand: 'Ford', model: 'Focus', ratings: [{ value: 5, comment: 'Excellent car!'}]} // { brand: 'Audi', model: 'A8', ratings: [{ value: 4.5, comment: 'I like it'}, {value: 5, comment: 'Best car I ever owned'}]} // ... ;(async function() { const scrapeResults = scrape('https://car-list.com', parseCars) for await (const carListing of scrapeResults) { console.log(JSON.stringify(carListing)) } })() /** * https://car-list.com * * <body> * <ul> * <li class="car"> * <span class="brand">Ford</span> * <span class="model">Focus</span> * <a class="ratings" href="/ratings/ford-focus">show ratings</a> * </li> * ... * </ul> * </body> */ async function* parseCars({ find, follow, capture }) { const cars = find('.car') for (const car of cars) { yield { brand: car.find('.brand').text(), model: car.find('.model').text(), ratings: await capture(car.find('a.ratings').attr('href'), parseCarRatings) } } follow(find('.next-page')) } /** * https://car-list.com/ratings/ford-focus * * <body> * <ul> * <li class="rating"> * <span class="value">5</span> * <span class="comment">Excellent car!</span> * </li> * ... * </ul> * </body> */ function* parseCarRatings({ find }) { const ratings = find('.rating') for (const rating of ratings) { yield { value: rating.find('.value').text(), comment: rating.find('.comment').text(), } } }
Here's the basic usage:
// import scraper const scrape = require('@epegzz/node-scraper') // define a parser function function* parser() { // ... } // call scraper with URL and parser const scrapeResults = scrape('https://some-website.com', parser) // consume scrape results for await (const scrapedItem of scrapeResults) { console.log(JSON.stringify(scrapedItem)) }
Instead of calling the scraper with a URL, you can also call it with an Axios request config object to gain more control over the requests:
const scrapeResults = scrape({ url: 'https://some-website.com', timeout: 5000, }, parser)
A parser function is a synchronous or asynchronous generator function which receives three utility functions as argument: find, follow and capture.
A fourth parser function argument is the context variable, which can be passed using the scrape, follow or capture function.
Whatever is yielded by the generator function, can be consumed as scrape result.
async function* parseCars({ find, follow, capture }) { const cars = find('.car') for (const car of cars) { yield { brand: car.find('.brand').text(), model: car.find('.model').text(), ratings: await capture(car.find('a.ratings').attr('href'), parseCarRatings) } } follow(find('a.next-page').href) } ;(async function() { const scrapeResults = scrape('https://car-list.com', parseCars) for await (const car of scrapeResults) { // whatever is yielded by the parser, ends up here console.log(JSON.stringify(car)) } })()
The find function allows you to extract data from the website.
It's basically just performing a Cheerio query, so check out their
documentation for details on how to use it.
Think of find as the $ in their documentation, loaded with the HTML contents of the
scraped website.
Example:
// yields the href and text of all links from the webpage for (const link of find('a')) { yield { linkHref: link.attr('href'), linkText: link.text(), }; }
The major difference between cheerio's $ and node-scraper's find is, that the results of find
are iterable. So you can do for (element of find(selector)) { ... } instead of having
to use a .each callback, which is important if we want to yield results.
The other difference is, that you can pass an optional node argument to find. This
will not search the whole document, but instead limits the search to that particular node's
inner HTML.
The main use-case for the follow function scraping paginated websites.
In that case you would use the href of the "next" button to let the scraper follow to the next page:
async function* parser({ find, follow }) { ... follow(find('a.next-page').attr('href')) }
The follow function will by default use the current parser to parse the
results of the new URL. You can, however, provide a different parser if you like.
The capture function is somewhat similar to the follow function: It takes
a new URL and a parser function as argument to scrape data. But instead of yielding the data as scrape results
it instead returns them as an array.
This is useful if you want add more details to a scraped object, where getting those details requires an additional network request:
async function* parseCars({ find, follow, capture }) { const cars = find('.car') for (const car of cars) { yield { brand: car.find('.brand').text(), model: car.find('.model').text(), ratings: await capture(car.find('a.ratings').attr('href'), parseCarRatings) } } }
In the example above the comments for each car are located on a nested car
details page. We are therefore making a capture call. All yields from the
parseCarRatings parser will be added to the resulting array that we're
assigning to the ratings property.
Note that we have to use await, because network requests are always asynchronous.