summaryrefslogtreecommitdiff
path: root/modules/urlfilter
diff options
context:
space:
mode:
authorPhilipp Tanlak <philipp.tanlak@gmail.com>2023-10-05 14:53:37 +0200
committerPhilipp Tanlak <philipp.tanlak@gmail.com>2023-10-05 14:53:37 +0200
commit1fc497fbdc79a43c62ac2e8eaf4827752dbeef8e (patch)
tree67738e213ef97f249bdfa0f1bddda0839192cb77 /modules/urlfilter
parentbd9e7f7acfd855d4685aa4544169c0e29cdbf205 (diff)
Refactor codebase into modules
Diffstat (limited to 'modules/urlfilter')
-rw-r--r--modules/urlfilter/urlfilter.go35
-rw-r--r--modules/urlfilter/urlfilter_test.go55
2 files changed, 62 insertions, 28 deletions
diff --git a/modules/urlfilter/urlfilter.go b/modules/urlfilter/urlfilter.go
index 00a4bd2..1297c35 100644
--- a/modules/urlfilter/urlfilter.go
+++ b/modules/urlfilter/urlfilter.go
@@ -11,7 +11,7 @@ import (
)
func init() {
- flyscrape.RegisterModule(new(Module))
+ flyscrape.RegisterModule(Module{})
}
type Module struct {
@@ -23,7 +23,18 @@ type Module struct {
blockedURLsRE []*regexp.Regexp
}
-func (m *Module) OnLoad(v flyscrape.Visitor) {
+func (Module) ModuleInfo() flyscrape.ModuleInfo {
+ return flyscrape.ModuleInfo{
+ ID: "urlfilter",
+ New: func() flyscrape.Module { return new(Module) },
+ }
+}
+
+func (m *Module) Provision(v flyscrape.Context) {
+ if m.disabled() {
+ return
+ }
+
for _, pat := range m.AllowedURLs {
re, err := regexp.Compile(pat)
if err != nil {
@@ -41,9 +52,13 @@ func (m *Module) OnLoad(v flyscrape.Visitor) {
}
}
-func (m *Module) CanRequest(rawurl string, depth int) bool {
+func (m *Module) ValidateRequest(r *flyscrape.Request) bool {
+ if m.disabled() {
+ return true
+ }
+
// allow root url
- if rawurl == m.URL {
+ if r.URL == m.URL {
return true
}
@@ -58,14 +73,14 @@ func (m *Module) CanRequest(rawurl string, depth int) bool {
}
for _, re := range m.allowedURLsRE {
- if re.MatchString(rawurl) {
+ if re.MatchString(r.URL) {
ok = true
break
}
}
for _, re := range m.blockedURLsRE {
- if re.MatchString(rawurl) {
+ if re.MatchString(r.URL) {
ok = false
break
}
@@ -74,7 +89,11 @@ func (m *Module) CanRequest(rawurl string, depth int) bool {
return ok
}
+func (m *Module) disabled() bool {
+ return len(m.AllowedURLs) == 0 && len(m.BlockedURLs) == 0
+}
+
var (
- _ flyscrape.CanRequest = (*Module)(nil)
- _ flyscrape.OnLoad = (*Module)(nil)
+ _ flyscrape.RequestValidator = (*Module)(nil)
+ _ flyscrape.Provisioner = (*Module)(nil)
)
diff --git a/modules/urlfilter/urlfilter_test.go b/modules/urlfilter/urlfilter_test.go
index e383a32..9ebb8a5 100644
--- a/modules/urlfilter/urlfilter_test.go
+++ b/modules/urlfilter/urlfilter_test.go
@@ -5,16 +5,22 @@
package urlfilter_test
import (
+ "net/http"
+ "sync"
"testing"
"github.com/philippta/flyscrape"
"github.com/philippta/flyscrape/modules/followlinks"
+ "github.com/philippta/flyscrape/modules/hook"
"github.com/philippta/flyscrape/modules/starturl"
"github.com/philippta/flyscrape/modules/urlfilter"
"github.com/stretchr/testify/require"
)
func TestURLFilterAllowed(t *testing.T) {
+ var urls []string
+ var mu sync.Mutex
+
scraper := flyscrape.NewScraper()
scraper.LoadModule(&starturl.Module{URL: "http://www.example.com/"})
scraper.LoadModule(&followlinks.Module{})
@@ -22,16 +28,19 @@ func TestURLFilterAllowed(t *testing.T) {
URL: "http://www.example.com/",
AllowedURLs: []string{`/foo\?id=\d+`, `/bar$`},
})
-
- scraper.SetTransport(flyscrape.MockTransport(200, `
- <a href="foo?id=123">123</a>
- <a href="foo?id=ABC">ABC</a>
- <a href="/bar">bar</a>
- <a href="/barz">barz</a>`))
-
- var urls []string
- scraper.OnRequest(func(req *flyscrape.Request) {
- urls = append(urls, req.URL)
+ scraper.LoadModule(hook.Module{
+ AdaptTransportFn: func(rt http.RoundTripper) http.RoundTripper {
+ return flyscrape.MockTransport(200, `
+ <a href="foo?id=123">123</a>
+ <a href="foo?id=ABC">ABC</a>
+ <a href="/bar">bar</a>
+ <a href="/barz">barz</a>`)
+ },
+ ReceiveResponseFn: func(r *flyscrape.Response) {
+ mu.Lock()
+ urls = append(urls, r.Request.URL)
+ mu.Unlock()
+ },
})
scraper.Run()
@@ -43,6 +52,9 @@ func TestURLFilterAllowed(t *testing.T) {
}
func TestURLFilterBlocked(t *testing.T) {
+ var urls []string
+ var mu sync.Mutex
+
scraper := flyscrape.NewScraper()
scraper.LoadModule(&starturl.Module{URL: "http://www.example.com/"})
scraper.LoadModule(&followlinks.Module{})
@@ -50,16 +62,19 @@ func TestURLFilterBlocked(t *testing.T) {
URL: "http://www.example.com/",
BlockedURLs: []string{`/foo\?id=\d+`, `/bar$`},
})
-
- scraper.SetTransport(flyscrape.MockTransport(200, `
- <a href="foo?id=123">123</a>
- <a href="foo?id=ABC">ABC</a>
- <a href="/bar">bar</a>
- <a href="/barz">barz</a>`))
-
- var urls []string
- scraper.OnRequest(func(req *flyscrape.Request) {
- urls = append(urls, req.URL)
+ scraper.LoadModule(hook.Module{
+ AdaptTransportFn: func(rt http.RoundTripper) http.RoundTripper {
+ return flyscrape.MockTransport(200, `
+ <a href="foo?id=123">123</a>
+ <a href="foo?id=ABC">ABC</a>
+ <a href="/bar">bar</a>
+ <a href="/barz">barz</a>`)
+ },
+ ReceiveResponseFn: func(r *flyscrape.Response) {
+ mu.Lock()
+ urls = append(urls, r.Request.URL)
+ mu.Unlock()
+ },
})
scraper.Run()