// URLs implements a breadth-first search webcrawler based on the // example given in section 8.6 of The Go Programming Language. package main import ( "flag" "fmt" "log" "net/url" "sync" ) func main() { maxConcurrency := flag.Int("c", 0, "Maximum number of concurrent queue pushes") startRawURL := flag.String("url", "", "Entry-point URL") maxURLs := flag.Int("max", 0, "Maximum number of URLs to collect (omitted or 0 means no limit)") flag.Parse() if *maxConcurrency == 0 { log.Fatal("Missing -c argument") } if *maxConcurrency < 1 { log.Fatalf("Invalid -c argument: %d", *maxConcurrency) } if *startRawURL == "" { log.Fatal("Missing -url argument") } if *maxURLs < 0 { log.Fatalf("Invalid -max argument: %d", *maxURLs) } startURL, err := url.Parse(*startRawURL) if err != nil { log.Fatal(err) } pool(*startURL, *maxConcurrency, *maxURLs) } func pool(startURL url.URL, maxConcurrency, maxURLs int) { var wg sync.WaitGroup urls := make(chan url.URL) outlets := make([]<-chan []url.URL, maxConcurrency) for i := range maxConcurrency { outlets[i] = worker(&wg, i+1, urls) } out := fanIn(outlets...) go func() { urls <- startURL }() seen := make(map[url.URL]bool) for batch := range out { for _, u := range batch { if !seen[u] { seen[u] = true go func() { urls <- u }() fmt.Printf("(%d) %s\n", len(seen), &u) } } } wg.Wait() } func worker(wg *sync.WaitGroup, id int, urls <-chan url.URL) <-chan []url.URL { out := make(chan []url.URL) wg.Go(func() { defer func() { close(out) }() for u := range urls { //fmt.Printf("(%d) starting %s\n", id, &u) doc, err := fetch(u) if err != nil { log.Print(err) } batch := findURLs(u, doc) // fmt.Printf("(%d) finished %s\n", id, &u) out <- batch } }) return out } func fanIn(chans ...<-chan []url.URL) <-chan []url.URL { out := make(chan []url.URL) var wg sync.WaitGroup for _, ch := range chans { wg.Go(func() { for batch := range ch { out <- batch } }) } go func() { wg.Wait() close(out) }() return out }