kokizzu / ant Goto Github PK

ant (alpha) is a web crawler for Go.

The package includes functions that can scan data from the page into your structs or slice of structs, this allows you to reduce the noise and complexity in your source-code.

You can also use a jQuery-like API that allows you to scrape complex HTML pages if needed.

var data struct { Title string `css:"title"` }
page, _ := ant.Fetch(ctx, "https://apple.com")
page.Scan(&data)
data.Title // => Apple

By default the crawler uses http.Client, however if you're crawling SPAs youc an use the antcdp.Client implementation which allows you to use chrome headless browser to crawl pages.

eng, err := ant.Engine(ant.EngineConfig{
  Fetcher: &ant.Fetcher{
    Client: antcdp.Client{},
  },
})

The crawler automatically fetches and caches robots.txt, making sure that it never causes issues to small website owners. Of-course you can disable this behavior.

eng, err := ant.NewEngine(ant.EngineConfig{
  Impolite: true,
})
eng.Run(ctx)

The crawler maintains a configurable amount of "worker" goroutines that read URLs off the queue, and spawn a goroutine for each URL.

Depending on your configuration, you may want to increase the number of workers to speed up URL reads, of-course if you don't have enough resources you can reduce the number of workers too.

eng, err := ant.NewEngine(ant.EngineConfig{
  // Spawn 5 worker goroutines that dequeue
  // URLs and spawn a new goroutine for each URL.
  Workers: 5,
})
eng.Run(ctx)

The package includes a powerful ant.Limiter interface that allows you to define rate limits per URL. There are some built-in limiters as well.

ant.Limit(1) // 1 rps on all URLs.
ant.LimitHostname(5, "amazon.com") // 5 rps on amazon.com hostname.
ant.LimitPattern(5, "amazon.com.*") // 5 rps on URLs starting with `amazon.co.`.
ant.LimitRegexp(5, "^apple.com\/iphone\/*") // 5 rps on URLs that match the regex.

Note that LimitPattern and LimitRegexp only match on the host and path of the URL.

Another powerful interface is ant.Matcher which allows you to define URL matchers, the matchers are called before URLs are queued.

ant.MatchHostname("amazon.com") // scrape amazon.com URLs only.
ant.MatchPattern("amazon.com/help/*")
ant.MatchRegexp("amazon\.com\/help/.+")

The crawl engine automatically retries any errors that implement Temporary() error that returns true.

Becuase the standard library returns errors that implement that interface the engine will retry most temporary network and HTTP errors.

eng, err := ant.NewEngine(ant.EngineConfig{
  Scraper: myscraper{},
  MaxAttempts: 5,
})

// Blocks until one of the following is true:
//
// 1. No more URLs to crawl (the scraper stops returning URLs)
// 2. A non-temporary error occured.
// 3. MaxAttempts was reached.
//
err = eng.Run(ctx)

The whole point of scraping is to extract data from websites into a machine readable format such as CSV or JSON, ant comes with built-in scrapers to make this ridiculously easy, here's a full cralwer that extracts quotes into stdout.

func main() {
	var url = "http://quotes.toscrape.com"
	var ctx = context.Background()
	var start = time.Now()

	type quote struct {
		Text string   `css:".text"   json:"text"`
		By   string   `css:".author" json:"by"`
		Tags []string `css:".tag"    json:"tags"`
	}

	type page struct {
		Quotes []quote `css:".quote" json:"quotes"`
	}

	eng, err := ant.NewEngine(ant.EngineConfig{
		Scraper: ant.JSON(os.Stdout, page{}, `li.next > a`),
		Matcher: ant.MatchHostname("quotes.toscrape.com"),
	})
	if err != nil {
		log.Fatalf("new engine: %s", err)
	}

	if err := eng.Run(ctx, url); err != nil {
		log.Fatal(err)
	}

	log.Printf("scraped in %s :)", time.Since(start))
}

anttest package makes it easy to test your scraper implementation it fetches a page by a URL, caches it in the OS's temporary directory and re-uses it.

The func depends on the file's modtime, the file expires daily, you can adjust the TTL by setting antttest.FetchTTL.

// Fetch calls `t.Fatal` on errors.
page := anttest.Fetch(t, "https://apple.com")
_, err := myscraper.Scrape(ctx, page)
assert.NoError(err)