示例#1
0
func (this *HttpDownloader) downloadHtml(p *page.Page, req *context.Request) *page.Page {
	var err error
	p, destbody := this.downloadFile(p, req)
	//fmt.Printf("Destbody %v \r\n", destbody)
	if !p.IsSucc() {
		//fmt.Print("Page error \r\n")
		return p
	}
	bodyReader := bytes.NewReader([]byte(destbody))

	var doc *goquery.Document
	if doc, err = goquery.NewDocumentFromReader(bodyReader); err != nil {
		log.Error(err.Error())
		p.SetStatus(true, err.Error())
		return p
	}

	var body string
	if body, err = doc.Html(); err != nil {
		log.Error(err.Error())
		p.SetStatus(true, err.Error())
		return p
	}

	p.SetBodyStr(body).SetHtmlParser(doc).SetStatus(false, "")

	return p
}
示例#2
0
func (this *HttpDownloader) changeCharsetEncodingAutoGzipSupport(contentTypeStr string, sor io.ReadCloser) string {
	var err error
	gzipReader, err := gzip.NewReader(sor)
	if err != nil {
		log.Error(err.Error())
		return ""
	}
	defer gzipReader.Close()
	destReader, err := charset.NewReader(gzipReader, contentTypeStr)

	if err != nil {
		log.Error(err.Error())
		destReader = sor
	}

	var sorbody []byte
	if sorbody, err = ioutil.ReadAll(destReader); err != nil {
		log.Error(err.Error())
		// For gb2312, an error will be returned.
		// Error like: simplifiedchinese: invalid GBK encoding
		// return ""
	}
	//e,name,certain := charset.DetermineEncoding(sorbody,contentTypeStr)
	bodystr := string(sorbody)

	return bodystr
}
示例#3
0
// choose http GET/method to download
func connectByHttp(p *page.Page, req *context.Request) (*http.Response, error) {
	client := &http.Client{
		CheckRedirect: req.GetRedirectFunc(),
	}

	httpreq, err := http.NewRequest(req.GetMethod(), req.GetUrl(), strings.NewReader(req.GetPostdata()))
	if header := req.GetHeader(); header != nil {
		httpreq.Header = req.GetHeader()
	}

	if cookies := req.GetCookies(); cookies != nil {
		for i := range cookies {
			httpreq.AddCookie(cookies[i])
		}
	}

	var resp *http.Response
	if resp, err = client.Do(httpreq); err != nil {
		if e, ok := err.(*url.Error); ok && e.Err != nil && e.Err.Error() == "normal" {
			//  normal
		} else {
			log.Error(err.Error())
			p.SetStatus(true, err.Error())
			//fmt.Printf("client do error %v \r\n", err)
			return nil, err
		}
	}

	return resp, nil
}
示例#4
0
func (this *HttpDownloader) downloadJson(p *page.Page, req *context.Request) *page.Page {
	var err error
	p, destbody := this.downloadFile(p, req)
	if !p.IsSucc() {
		return p
	}

	var body []byte
	body = []byte(destbody)
	mtype := req.GetResponceType()
	if mtype == "jsonp" {
		tmpstr := utils.JsonpToJson(destbody)
		body = []byte(tmpstr)
	}

	var r *simplejson.Json
	if r, err = simplejson.NewJson(body); err != nil {
		log.Error(string(body) + "\t" + err.Error())
		p.SetStatus(true, err.Error())
		return p
	}

	// json result
	p.SetBodyStr(string(body)).SetJson(r).SetStatus(false, "")

	return p
}
示例#5
0
文件: page.go 项目: num5/spider
// GetHtmlParser returns goquery object binded to target crawl result.
func (this *Page) ResetHtmlParser() *goquery.Document {
	r := strings.NewReader(this.body)
	var err error
	this.docParser, err = goquery.NewDocumentFromReader(r)
	if err != nil {
		log.Error(err.Error())
		panic(err.Error())
	}
	return this.docParser
}
示例#6
0
func (this *HttpDownloader) Download(req *context.Request) *page.Page {
	var mtype string
	var p = page.NewPage(req)
	mtype = req.GetResponceType()
	switch mtype {
	case "html":
		return this.downloadHtml(p, req)
	case "json":
		fallthrough
	case "jsonp":
		return this.downloadJson(p, req)
	case "text":
		return this.downloadText(p, req)
	default:
		log.Error("error request type:" + mtype)
	}
	return p
}
示例#7
0
文件: request.go 项目: num5/spider
func readHeaderFromFile(headerFile string) http.Header {
	//read file , parse the header and cookies
	b, err := ioutil.ReadFile(headerFile)
	if err != nil {
		//make be:  share access error
		log.Error(err.Error())
		return nil
	}
	js, _ := simplejson.NewJson(b)
	//constructed to header

	h := make(http.Header)
	h.Add("User-Agent", js.Get("User-Agent").MustString())
	h.Add("Referer", js.Get("Referer").MustString())
	h.Add("Cookie", js.Get("Cookie").MustString())
	h.Add("Cache-Control", "max-age=0")
	h.Add("Connection", "keep-alive")
	return h
}
示例#8
0
// Download file and change the charset of page charset.
func (this *HttpDownloader) downloadFile(p *page.Page, req *context.Request) (*page.Page, string) {
	var err error
	var urlstr string
	if urlstr = req.GetUrl(); len(urlstr) == 0 {
		log.Error("url is empty")
		p.SetStatus(true, "url is empty")
		return p, ""
	}

	var resp *http.Response

	if proxystr := req.GetProxyHost(); len(proxystr) != 0 {
		//using http proxy
		//fmt.Print("HttpProxy Enter ",proxystr,"\n")
		resp, err = connectByHttpProxy(p, req)
	} else {
		//normal http download
		//fmt.Print("Http Normal Enter \n",proxystr,"\n")
		resp, err = connectByHttp(p, req)
	}

	if err != nil {
		return p, ""
	}

	//b, _ := ioutil.ReadAll(resp.Body)
	//fmt.Printf("Resp body %v \r\n", string(b))

	p.SetHeader(resp.Header)
	p.SetCookies(resp.Cookies())

	// get converter to utf-8
	var bodyStr string
	if resp.Header.Get("Content-Encoding") == "gzip" {
		bodyStr = this.changeCharsetEncodingAutoGzipSupport(resp.Header.Get("Content-Type"), resp.Body)
	} else {
		bodyStr = this.changeCharsetEncodingAuto(resp.Header.Get("Content-Type"), resp.Body)
	}
	//fmt.Printf("utf-8 body %v \r\n", bodyStr)
	defer resp.Body.Close()
	return p, bodyStr
}