Allow configuration of links to follow

3 files changed, 148 insertions, 16 deletions

diff --git a/README.md b/README.md
index 7d9e249..1802a56 100644
--- a/README.md
+++ b/README.md
@@ -140,6 +140,7 @@ Below is an example scraping script that showcases the capabilities of flyscrape
 export const config = {
     url: "https://example.com/", // Specify the URL to start scraping from.
     depth: 0,                    // Specify how deep links should be followed.  (default = 0, no follow)
+    follow: [],                  // Speficy the css selectors to follow         (default = ["a[href]"])
     allowedDomains: [],          // Specify the allowed domains. ['*'] for all. (default = domain from url)
     blockedDomains: [],          // Specify the blocked domains.                (default = none)
     allowedURLs: [],             // Specify the allowed URLs as regex.          (default = all allowed)
diff --git a/modules/followlinks/followlinks.go b/modules/followlinks/followlinks.go
index c53f167..c1448be 100644
--- a/modules/followlinks/followlinks.go
+++ b/modules/followlinks/followlinks.go
@@ -6,6 +6,7 @@ package followlinks
 
 import (
 	"net/url"
+	"regexp"
 	"strings"
 
 	"github.com/PuerkitoBio/goquery"
@@ -16,7 +17,9 @@ func init() {
 	flyscrape.RegisterModule(Module{})
 }
 
-type Module struct{}
+type Module struct {
+	Follow []string `json:"follow"`
+}
 
 func (Module) ModuleInfo() flyscrape.ModuleInfo {
 	return flyscrape.ModuleInfo{
@@ -25,13 +28,19 @@ func (Module) ModuleInfo() flyscrape.ModuleInfo {
 	}
 }
 
+func (m *Module) Provision(ctx flyscrape.Context) {
+	if len(m.Follow) == 0 {
+		m.Follow = []string{"a[href]"}
+	}
+}
+
 func (m *Module) ReceiveResponse(resp *flyscrape.Response) {
-	for _, link := range parseLinks(string(resp.Body), resp.Request.URL) {
+	for _, link := range m.parseLinks(string(resp.Body), resp.Request.URL) {
 		resp.Visit(link)
 	}
 }
 
-func parseLinks(html string, origin string) []string {
+func (m *Module) parseLinks(html string, origin string) []string {
 	var links []string
 	doc, err := goquery.NewDocumentFromReader(strings.NewReader(html))
 	if err != nil {
@@ -44,22 +53,26 @@ func parseLinks(html string, origin string) []string {
 	}
 
 	uniqueLinks := make(map[string]bool)
-	doc.Find("a").Each(func(i int, s *goquery.Selection) {
-		link, _ := s.Attr("href")
 
-		parsedLink, err := originurl.Parse(link)
+	for _, selector := range m.Follow {
+		attr := parseSelectorAttr(selector)
+		doc.Find(selector).Each(func(i int, s *goquery.Selection) {
+			link, _ := s.Attr(attr)
 
-		if err != nil || !isValidLink(parsedLink) {
-			return
-		}
+			parsedLink, err := originurl.Parse(link)
 
-		absLink := parsedLink.String()
+			if err != nil || !isValidLink(parsedLink) {
+				return
+			}
 
-		if !uniqueLinks[absLink] {
-			links = append(links, absLink)
-			uniqueLinks[absLink] = true
-		}
-	})
+			absLink := parsedLink.String()
+
+			if !uniqueLinks[absLink] {
+				links = append(links, absLink)
+				uniqueLinks[absLink] = true
+			}
+		})
+	}
 
 	return links
 }
@@ -72,4 +85,26 @@ func isValidLink(link *url.URL) bool {
 	return true
 }
 
-var _ flyscrape.ResponseReceiver = (*Module)(nil)
+func parseSelectorAttr(sel string) string {
+	matches := selectorExpr.FindAllString(sel, -1)
+	if len(matches) == 0 {
+		return "href"
+	}
+
+	attr := attrExpr.FindString(matches[len(matches)-1])
+	if attr == "" {
+		return "href"
+	}
+
+	return attr
+}
+
+var (
+	_ flyscrape.Provisioner      = (*Module)(nil)
+	_ flyscrape.ResponseReceiver = (*Module)(nil)
+)
+
+var (
+	selectorExpr = regexp.MustCompile(`\[(.*?)\]`)
+	attrExpr     = regexp.MustCompile(`[\w-]+`)
+)
diff --git a/modules/followlinks/followlinks_test.go b/modules/followlinks/followlinks_test.go
index 0a628c3..f3eb4fe 100644
--- a/modules/followlinks/followlinks_test.go
+++ b/modules/followlinks/followlinks_test.go
@@ -47,3 +47,99 @@ func TestFollowLinks(t *testing.T) {
 	require.Contains(t, urls, "http://www.google.com")
 	require.Contains(t, urls, "http://www.google.com/baz")
 }
+
+func TestFollowSelector(t *testing.T) {
+	var urls []string
+	var mu sync.Mutex
+
+	scraper := flyscrape.NewScraper()
+	scraper.LoadModule(&starturl.Module{URL: "http://www.example.com/foo/bar"})
+	scraper.LoadModule(&followlinks.Module{
+		Follow: []string{".next a[href]"},
+	})
+
+	scraper.LoadModule(hook.Module{
+		AdaptTransportFn: func(rt http.RoundTripper) http.RoundTripper {
+			return flyscrape.MockTransport(200, `
+				<a href="/baz">Baz</a>
+				<a href="baz">Baz</a>
+                <div class="next">
+				    <a href="http://www.google.com">Google</a>
+                </div>`)
+		},
+		ReceiveResponseFn: func(r *flyscrape.Response) {
+			mu.Lock()
+			urls = append(urls, r.Request.URL)
+			mu.Unlock()
+		},
+	})
+
+	scraper.Run()
+
+	require.Len(t, urls, 2)
+	require.Contains(t, urls, "http://www.example.com/foo/bar")
+	require.Contains(t, urls, "http://www.google.com")
+}
+
+func TestFollowDataAttr(t *testing.T) {
+	var urls []string
+	var mu sync.Mutex
+
+	scraper := flyscrape.NewScraper()
+	scraper.LoadModule(&starturl.Module{URL: "http://www.example.com/foo/bar"})
+	scraper.LoadModule(&followlinks.Module{
+		Follow: []string{"[data-url]"},
+	})
+
+	scraper.LoadModule(hook.Module{
+		AdaptTransportFn: func(rt http.RoundTripper) http.RoundTripper {
+			return flyscrape.MockTransport(200, `
+				<a href="/baz">Baz</a>
+				<a href="baz">Baz</a>
+				<div data-url="http://www.google.com">Google</div>`)
+		},
+		ReceiveResponseFn: func(r *flyscrape.Response) {
+			mu.Lock()
+			urls = append(urls, r.Request.URL)
+			mu.Unlock()
+		},
+	})
+
+	scraper.Run()
+
+	require.Len(t, urls, 2)
+	require.Contains(t, urls, "http://www.example.com/foo/bar")
+	require.Contains(t, urls, "http://www.google.com")
+}
+
+func TestFollowMultiple(t *testing.T) {
+	var urls []string
+	var mu sync.Mutex
+
+	scraper := flyscrape.NewScraper()
+	scraper.LoadModule(&starturl.Module{URL: "http://www.example.com/foo/bar"})
+	scraper.LoadModule(&followlinks.Module{
+		Follow: []string{"a.prev", "a.next"},
+	})
+
+	scraper.LoadModule(hook.Module{
+		AdaptTransportFn: func(rt http.RoundTripper) http.RoundTripper {
+			return flyscrape.MockTransport(200, `
+				<a href="/baz">Baz</a>
+				<a class="prev" href="a">a</a>
+				<a class="next" href="b">b</a>`)
+		},
+		ReceiveResponseFn: func(r *flyscrape.Response) {
+			mu.Lock()
+			urls = append(urls, r.Request.URL)
+			mu.Unlock()
+		},
+	})
+
+	scraper.Run()
+
+	require.Len(t, urls, 3)
+	require.Contains(t, urls, "http://www.example.com/foo/bar")
+	require.Contains(t, urls, "http://www.example.com/foo/a")
+	require.Contains(t, urls, "http://www.example.com/foo/b")
+}