6
\$\begingroup\$

Description

As an exercise of learning go concurrency patterns, I decided too build a concurrent web crawler.

I made use of the argparse module I put up for review a while back.

I'm looking for feedback on my concurrency pattern, but any and all aspect of the code is open to be flamed :)

Code

package main

import (
    "fmt"
    "sync"
    "net/http"
    "io"
    "golang.org/x/net/html"
    "strings"
    "sort"
    "argparse"
)

func min(vars ...int) int {
    m := vars[0]
    for i := 1; i < len(vars); i++ {
        if vars[i] < m {
            m = vars[i]
        }
    }
    return m
}

type Crawler struct {
    base string
    pop chan []string
    push chan string
    wg *sync.WaitGroup
    visited map[string]bool
    hrefs []string
    queue []string
    maxChannels int
}

func newCrawler(base string, maxChannels int) Crawler {
    c := Crawler {
        base: base,
        maxChannels: maxChannels,
        pop: make(chan []string, maxChannels),
        push: make(chan string, maxChannels),
        wg: new(sync.WaitGroup),
        visited: make(map[string]bool),
        queue: make([]string, 1),
    }

    c.queue[0] = base
    c.visited[base] = true

    return c
}

func (c *Crawler) run() []string {
    defer func() {
        c.wg.Wait()
    }()

    for len(c.queue) > 0 {
        l := min(len(c.queue), c.maxChannels)
        
        for i := 0; i < l; i++ {
            url := c.queue[0]
            c.queue = c.queue[1:]
            c.hrefs = append(c.hrefs, url)
            c.runWorker(url)
            c.push <- url
        }

        for i := 0; i < l; i++ {
            hrefs := <- c.pop
            c.filterHrefs(hrefs)
        }
    }
    return c.hrefs
}

func (c *Crawler) filterHrefs(hrefs []string) {
    for _, href := range hrefs {
        if _, f := c.visited[href]; !f && strings.Contains(href, c.base) {
            c.visited[href] = true
            c.queue = append(c.queue, href)
        }
    }
} 

func (c *Crawler) runWorker(url string) {
    w := Worker {
        base: c.base,
        push: c.pop,
        pop: c.push,
        wg: c.wg,
    }
    c.wg.Add(1)
    go w.run()
}

type  Worker struct {
    base string
    push chan []string
    pop chan string
    wg *sync.WaitGroup
}

func (w *Worker) parseHref(href string) string {
    var url string
    switch {
        case strings.HasPrefix(href, "/"):
            url = w.base + href
        case strings.HasPrefix(href, "http"):
            url = href
    }
    return url
}

func (w *Worker) getAllHrefs(body io.Reader) []string {
    hrefs := make([]string, 0)
    page := html.NewTokenizer(body)
    for page.Next() != html.ErrorToken {
        token := page.Token()
        if token.Data == "a" {
            for _, a := range token.Attr {
                if a.Key == "href" {
                    hrefs = append(hrefs, w.parseHref(a.Val))
                }
            }
        }
    }
    return hrefs
}

func (w *Worker) fetch(url string) (io.Reader, error) {
    resp, err := http.Get(url)
    if err != nil {
        return nil, err
    }
    return resp.Body, nil
}

func(w *Worker) run() {
    defer func() {
        w.wg.Done()
    }()

    url := <- w.pop
    hrefs := make([]string, 0)
    body, err := w.fetch(url)
    if err == nil {
        hrefs = w.getAllHrefs(body)
    }
    w.push <- hrefs
}

func parseArguments() map[string]interface{} {
    parser := argparse.Argparse {
        Description: "Site crawler by @Ludisposed",
    }

    parser.AddArgument(
        argparse.Argument {
            ShortFlag: "b", LongFlag: "base", Type: "string", 
            Required: true, Help: "The base of the url",
        },
    )

    parser.AddArgument(
        argparse.Argument {
            ShortFlag: "m", LongFlag: "max", Type: 10, 
            Help: "Max amount of channels", Default: 10,
        },
    )

    return parser.Parse()
}

func main() {
    args := parseArguments()

    crawler := newCrawler(
        args["base"].(string), 
        args["max"].(int),
    )
    hrefs := crawler.run()

    sort.Strings(hrefs) // Sorting because pretty
    for _, h := range hrefs {
        fmt.Println(h)
    }
    fmt.Println("\n[+] Total unique urls found:", len(hrefs))   
}
\$\endgroup\$
0

1 Answer 1

3
+50
\$\begingroup\$

Disclaimer: I haven't had much exposure to golang. I'm mostly trying to pick up the language by going through random projects.

Going over the code you've provided, it seems to be easily followed. A few pointers (questions? concerns?), which might be due to my lack of knowledge:

  1. Your min function uses a for loop, where the conditional statement calls len(vars) on each iteration. This seems inefficient. Later in your code, you've used for _, value := range iterable style syntax. I'd be preferring that over here as well; since we're interested in value only, and not the index.

  2. When extracting the href attribute for all a tags, you keep iterating over attributes even when you've successfully captured href. Break early?

     for _, a := range token.Attr {
         if a.Key == "href" {
             hrefs = append(hrefs, w.parseHref(a.Val))
             break
         }
     }
    
  3. The parseHref function uses a switch statement, without a fallback default. It should return an error if the provided value does not satisfy either of those, or if you're planning to return the same value, then a switch-case block seems overwhelming.

     func (w *Worker) parseHref(href string) string {
         url = href
         if strings.HasPrefix(href, "/") {
             url = w.base + href
         }
         return url
     }
    
\$\endgroup\$
2
  • \$\begingroup\$ I'm only starting with go myself, but you've raised some valid points! Do you have any thoughts about the concurrency pattern? As it was the reason that prompted the question. IMHO the pattern seems off, (waiting till all goroutines are done before firing off the next batch) \$\endgroup\$
    – Ludisposed
    Commented Oct 9, 2020 at 14:36
  • \$\begingroup\$ The functionality seems to be fine. Gathering results from channels is something which seems to me to be a personal preference. You can also try to take help of the merge and parallel digestion examples on this blog post: blog.golang.org/pipelines @Ludisposed \$\endgroup\$
    – hjpotter92
    Commented Oct 10, 2020 at 10:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.