// choose http GET/method to download
func connectByHttp(p *context.Response, 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 {
reporter.Log.Println(err.Error())
p.SetStatus(true, err.Error())
//fmt.Printf("client do error %v \r\n", err)
return nil, err
}
}
return resp, nil
}
func (self *HttpDownloader) downloadText(p *context.Response, req *context.Request) *context.Response {
p, destbody := self.downloadFile(p, req)
if !p.IsSucc() {
return p
}
p.SetBodyStr(destbody).SetStatus(false, "")
return p
}
func (self *HttpDownloader) downloadHtml(p *context.Response, req *context.Request) *context.Response {
var err error
p, destbody := self.downloadFile(p, req)
//fmt.Printf("Destbody %v \r\n", destbody)
if !p.IsSucc() {
//fmt.Print("Response error \r\n")
return p
}
bodyReader := bytes.NewReader([]byte(destbody))
var doc *goquery.Document
if doc, err = goquery.NewDocumentFromReader(bodyReader); err != nil {
reporter.Log.Println(err.Error())
p.SetStatus(true, err.Error())
return p
}
var body string
if body, err = doc.Html(); err != nil {
reporter.Log.Println(err.Error())
p.SetStatus(true, err.Error())
return p
}
p.SetBodyStr(body).SetHtmlParser(doc).SetStatus(false, "")
return p
}
func (self *HttpDownloader) downloadJson(p *context.Response, req *context.Request) *context.Response {
var err error
p, destbody := self.downloadFile(p, req)
if !p.IsSucc() {
return p
}
var body []byte
body = []byte(destbody)
mtype := req.GetRespType()
if mtype == "jsonp" {
tmpstr := util.JsonpToJson(destbody)
body = []byte(tmpstr)
}
var r *simplejson.Json
if r, err = simplejson.NewJson(body); err != nil {
reporter.Log.Println(string(body) + "\t" + err.Error())
p.SetStatus(true, err.Error())
return p
}
// json result
p.SetBodyStr(string(body)).SetJson(r).SetStatus(false, "")
return p
}
// Download file and change the charset of response charset.
func (self *HttpDownloader) downloadFile(p *context.Response, req *context.Request) (*context.Response, string) {
var err error
var urlstr string
if urlstr = req.GetUrl(); len(urlstr) == 0 {
reporter.Log.Println("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
bodyStr := self.changeCharsetEncodingAuto(resp.Header.Get("Content-Type"), resp.Body)
//fmt.Printf("utf-8 body %v \r\n", bodyStr)
defer resp.Body.Close()
return p, bodyStr
}
// core processer
func (self *crawler) Process(req *context.Request) {
// 声明response
var resp *context.Response
defer func() {
if err := recover(); err != nil { // do not affect other
if strerr, ok := err.(string); ok {
reporter.Log.Println(strerr)
} else {
reporter.Log.Println("Process error：", err)
}
}
}()
// reporter.Log.Println("**************断点 1 ***********")
// download page
for i := 0; i < 3; i++ {
self.sleep()
resp = self.Downloader.Download(req)
if resp.IsSucc() { // if fail retry 3 times
break
}
}
// reporter.Log.Println("**************断点 2 ***********")
if !resp.IsSucc() { // if fail do not need process
return
}
// reporter.Log.Println("**************断点 3 ***********")
// 过程处理，提炼数据
self.Spider.GoRule(resp)
// reporter.Log.Println("**************断点 5 ***********")
// 该条请求结果存入pipeline
datas := resp.GetItems()
for i, count := 0, len(datas); i < count; i++ {
self.Pipeline.Collect(
resp.GetRuleName(), //DataCell.RuleName
datas[i], //DataCell.Data
resp.GetUrl(), //DataCell.Url
resp.GetParent(), //DataCell.ParentUrl
time.Now().Format("2006-01-02 15:04:05"),
)
}
// reporter.Log.Println("**************断点 end ***********")
}
// 获取任务规则采集语义字段
func (self *Spider) GetOutFeild(resp *context.Response, index int) string {
return self.RuleTree.Nodes[resp.GetRuleName()].OutFeild[index]
}
// 用指定规则解析响应流
func (self *Spider) CallRule(ruleName string, resp *context.Response) {
resp.SetRuleName(ruleName)
self.GoRule(resp)
}
// 根据响应流运行指定解析规则
func (self *Spider) GoRule(resp *context.Response) {
self.RuleTree.Nodes[resp.GetRuleName()].ParseFunc(self, resp)
}