shears
v0.0.0-alpha.92
Published
Functional web scraping in typescript
Downloads
2
Maintainers
pingidReadme
Shears
A Declarative web scraping library that aims to provide an extendable set of tools for building complex typesafe queries and web crawlers.
import * as sh from 'shears'
const article = sh.select({
title: sh.select('h1', sh.text),
content: sh.select('h1', sh.text)
image: sh.select('img', sh.atr('src'))
})
const article_list = sh.select('#content', ['ul > li'], article)
await sh.runP(article_list, '<html><...')
// [{ title: '...' content: '...' },{...]The library works best when used in combination with fp-ts however the runP returns a Promise instead of TaskEither returned by the run function, this allows for standalone use but still requires fp-ts as a peer dependency.
Usage
Selectors
The select function is the main building block and can be used for chaining selectors of type string, [string], Shear<Node, Node> where the final argument can be of type Shear<Node, T>.
sh.select('body > h1') // Shear<Node | Node[], Node<h1>>
sh.select('body > h1', 'span') // Shear<Node | Node[], Node<span>>
sh.select('body > ul', ['li']) // Shear<Node | Node[], Node<li>[]>
sh.select(['body > ul'], sh.text) // Shear<Node | Node[], string[]>
sh.select(['body > ul'], ['li'], sh.text) // Shear<Node | Node[], string[][]>
sh.select({ foo: sh.text }) // Shear<Node | Node[], { foo: string }>
sh.select([sh.text, sh.text]) // Shear<Node | Node[], [string, string]>string: Accepts a css query and returns the first matching DOM node, likedocument.querySelector.[string]: Accepts a css query and returns all matching DOM nodes likedocument.querySelectorAll.
Each query in the list of arguments operates on the part of the DOM returned by the previous query where parameters after [string] queries operate on each item in the return list.
select can also be used for creating structured queries where it takes a single argument of type { [x: string]: Shear<Node, T> } or Shear<Node, Node>[].
sh.select({ title: sh.select('title') }) // Shear<Node | Node[], { title: Node<title> }>
sh.select([sh.select('title'), sh.select('h1')] as const) // Shear<Node | Node[], [Node<title>, Node<title>]>customizing Shears
A "Shear" extends the ReaderTaskEither type class so you can easily build your selectors.
import { map } from 'fp-ts/ReaderTaskEither'
import { pipe } from 'fp-ts/function'
import sh, { run } from 'shears'
const trimText = pipe(
sh.text,
map((y) => y.trim())
)
run(sh('body > h1', trimText), `<h1> foo </h1>`) // TaskEither<never, 'foo'>Crawling
The library provides a few shears to help with queries across multiple pages.
import { map } from 'fp-ts/ReaderTaskEither'
import { pipe } from 'fp-ts/function'
import axios from 'axios'
import * as sh from 'shears'
const connection = sh.connect((url) => axios.get(url).then((x) => x.data))
sh.run(
sh.select({
posts: sh.select(
'[#post]',
sh.goTo(
sh.select('a', sh.atr('href')),
sh.select({
title: sh.select('h1', sh.text)
}),
{ connection }
)
)
})
)Often it is the case you want to follow relative links on a website where our connection would need to know the current hostname. The "Shear" provides a mechanism for passing context we just need to change the connection implementation.
const connection = sh.connect(
(url, ctx) =>
(/^http/.test(url) ? axios.get(url) : axios.get(ctx.hostname + url)).then((x) => [
x.data,
{ hostname: x.response.hostname }
]), // We return a tuple [{ html string }, { new context }]
{ hostname: '' } // Initial context state
)We can pass our connection into the run function which provides it in the shear context so we don't need to pass it to goTo.
sh.run(
sh.goTo(
'http://foo.bar',
sh.select({
posts: sh(
'[#post]',
sh.goTo(
sh.select('a', sh.atr('href')),
sh.select({
title: sh.select('h1', sh.text)
})
)
)
})
),
{ connection }
)