Timeout and results saving

config.go

@@ -13,6 +13,7 @@ var config struct {
 	Token      string
 	Retries    int
 	Workers    int
+	Timeout    time.Duration
 	Tasks      int32
 	CrawlStats time.Duration
 	AllocStats time.Duration
@@ -25,6 +26,7 @@ const (
 	ConfTasks      = "crawl.tasks"
 	ConfRetries    = "crawl.retries"
 	ConfWorkers    = "crawl.connections"
+	ConfTimeout    = "crawl.timeout"
 	ConfCrawlStats = "output.crawl_stats"
 	ConfAllocStats = "output.resource_stats"
 	ConfVerbose    = "output.verbose"
@@ -34,6 +36,7 @@ func prepareConfig() {
 	viper.SetDefault(ConfRetries, 5)
 	viper.SetDefault(ConfWorkers, 2)
 	viper.SetDefault(ConfTasks, 3)
+	viper.SetDefault(ConfTimeout, 10 * time.Second)
 	viper.SetDefault(ConfCrawlStats, 3 * time.Second)
 	viper.SetDefault(ConfAllocStats, 0)
 	viper.SetDefault(ConfVerbose, false)
@@ -73,6 +76,8 @@ func readConfig() {
 		configOOB(ConfTasks, int(config.Tasks))
 	}
 
+	config.Timeout = viper.GetDuration(ConfTimeout)
+
 	config.CrawlStats = viper.GetDuration(ConfCrawlStats)
 
 	config.AllocStats = viper.GetDuration(ConfAllocStats)

config.yml

@@ -24,3 +24,5 @@ crawl:
   # How often to retry getting data
   # from the site before giving up
   retries: 5
+  # Time before discarding a network request
+  timeout: 10s

crawl.go

@@ -3,7 +3,6 @@ package main
 import (
 	"bytes"
 	"fmt"
-	"github.com/sirupsen/logrus"
 	"github.com/terorie/oddb-go/ds/redblackhash"
 	"github.com/terorie/oddb-go/fasturl"
 	"github.com/valyala/fasthttp"
@@ -28,13 +27,10 @@ func GetDir(j *Job, f *File) (links []fasturl.URL, err error) {
 	res := fasthttp.AcquireResponse()
 	defer fasthttp.ReleaseResponse(res)
 
-	err = client.Do(req, res)
+	err = client.DoTimeout(req, res, config.Timeout)
 	fasthttp.ReleaseRequest(req)
 
-	if err != nil {
+	if err != nil { return }
-		logrus.Error(err)
-		return
-	}
 
 	err = checkStatusCode(res.StatusCode())
 	if err != nil { return }
@@ -129,7 +125,7 @@ func GetFile(u fasturl.URL, f *File) (err error) {
 	res.SkipBody = true
 	defer fasthttp.ReleaseResponse(res)
 
-	err = client.Do(req, res)
+	err = client.DoTimeout(req, res, config.Timeout)
 	fasthttp.ReleaseRequest(req)
 
 	if err != nil { return }

main.go

@@ -61,7 +61,13 @@ func cmdCrawler(clic *cli.Context) error {
 			u.Path += "/"
 		}
 		if err != nil { return err }
-		remotes[i] = &OD{ BaseUri: u }
+		remotes[i] = &OD {
+			Task: &Task{
+				WebsiteId: 0,
+				Url: u.String(),
+			},
+			BaseUri: u,
+		}
 	}
 
 	c := context.Background()

model.go

@@ -16,9 +16,9 @@ type Job struct {
 }
 
 type OD struct {
+	Task    *Task
 	Wait    sync.WaitGroup
 	BaseUri fasturl.URL
-	Files   []File
 	WCtx    WorkerContext
 	Scanned redblackhash.Tree
 

scheduler.go

@@ -2,7 +2,11 @@ package main
 
 import (
 	"context"
+	"encoding/json"
+	"fmt"
 	"github.com/sirupsen/logrus"
+	"os"
+	"path"
 	"sync/atomic"
 )
 
@@ -12,19 +16,17 @@ var totalBuffered int64
 func Schedule(c context.Context, remotes <-chan *OD) {
 	go Stats(c)
 
-	for {
+	for remote := range remotes {
-		select {
-		case <-c.Done():
-			return
-
-		case remote := <-remotes:
 		logrus.WithField("url", remote.BaseUri.String()).
 			Info("Starting crawler")
 
+		// Collect results
+		results := make(chan File)
+
 		// Spawn workers
 		remote.WCtx.in, remote.WCtx.out = makeJobBuffer(c)
 		for i := 0; i < config.Workers; i++ {
-				go remote.WCtx.Worker()
+			go remote.WCtx.Worker(results)
 		}
 
 		// Enqueue initial job
@@ -37,12 +39,13 @@ func Schedule(c context.Context, remotes <-chan *OD) {
 		})
 
 		// Upload result when ready
-			go remote.Watch()
+		go remote.Watch(results)
-		}
 	}
 }
 
-func (r *OD) Watch() {
+func (r *OD) Watch(results chan File) {
+	go r.Task.Collect(results)
+
 	// Wait for all jobs on remote to finish
 	r.Wait.Wait()
 	close(r.WCtx.in)
@@ -52,6 +55,40 @@ func (r *OD) Watch() {
 		Info("Crawler finished")
 
 	globalWait.Done()
+
+	close(results)
+}
+
+func (t *Task) Collect(results chan File) {
+	err := t.collect(results)
+	if err != nil {
+		logrus.WithError(err).
+			Error("Failed saving crawl results")
+	}
+}
+
+func (t *Task) collect(results chan File) error {
+	err := os.MkdirAll("crawled", 0755)
+	if err != nil { return err }
+
+	f, err := os.OpenFile(
+		path.Join("crawled", fmt.Sprintf("%d.json", t.WebsiteId)),
+		os.O_CREATE | os.O_WRONLY | os.O_TRUNC,
+		0755,
+	)
+	if err != nil { return err }
+	defer f.Close()
+
+	for result := range results {
+		resJson, err := json.Marshal(result)
+		if err != nil { panic(err) }
+		_, err = f.Write(resJson)
+		if err != nil { return err }
+		_, err = f.Write([]byte{'\n'})
+		if err != nil { return err }
+	}
+
+	return nil
 }
 
 func makeJobBuffer(c context.Context) (chan<- Job, <-chan Job) {

worker.go

@@ -19,13 +19,13 @@ type WorkerContext struct {
 	numRateLimits int
 }
 
-func (w WorkerContext) Worker() {
+func (w WorkerContext) Worker(results chan<- File) {
 	for job := range w.out {
-		w.step(job)
+		w.step(results, job)
 	}
 }
 
-func (w WorkerContext) step(job Job) {
+func (w WorkerContext) step(results chan<- File, job Job) {
 	defer w.finishJob(&job)
 
 	var f File
@@ -64,7 +64,9 @@ func (w WorkerContext) step(job Job) {
 		w.queueJob(job)
 	}
 
-	job.OD.Files = append(job.OD.Files, f)
+	if !f.IsDir {
+		results <- f
+	}
 }
 
 func DoJob(job *Job, f *File) (newJobs []Job, err error) {