spider_editor/src/util/util.go

package util

import (
	"bytes"
	"fmt"
	mgo "mongodb"
	qu "qfw/util"
	"regexp"
	"sort"
	sp "spiderutil"
	"strings"
	"time"

	"github.com/yuin/gopher-lua"
)

const Role_Admin, Role_Examine, Role_Dev = 3, 2, 1 //管理员，审核员，开发员
var (
	//MgoE            *mgo.MongodbSim //编辑器87
	MgoEB             *mgo.MongodbSim //编辑器163
	MgoS              *mgo.MongodbSim
	Province          map[string][]string
	City              map[string][]string
	DomainNameReg     = regexp.MustCompile(`(http|https)[:：]+`)
	DownLoadReg       = regexp.MustCompile(`download\(.*?\)`)
	CodeTypeReg       = regexp.MustCompile(`(utf8|utf-8|gbk)`)
	TitleFilterReg1   = regexp.MustCompile(`[\p{Han}]`)
	TitleFilterReg2   = regexp.MustCompile(`((上|下)一(页|篇)|阅读次数)`)
	DetailFilterReg1  = regexp.MustCompile(`((上|下)一(页|篇)|阅读次数|浏览次数|扫一扫|分享|区块链存证)`)
	Area              []string //省份
	DomainReg         = regexp.MustCompile(`(?://).+?(?:[:：/])`)
	SymbolReg         = regexp.MustCompile("[,，\\s\u3000\u2003\u00a0]+")
	ReplaceReg        = regexp.MustCompile(`[]:：/]+`)
	CheckText         = `item["spidercode"]="%s";item["site"]="%s";item["channel"]="%s"`
	CheckText_Code    = `item["spidercode"]="%s"`
	CheckText_Site    = `item["site"]="%s"`
	CheckText_Channel = `item["channel"]="%s"`
	JsonDataMap       = map[string]bool{ //jsondata
		"extweight":          true,
		"projecthref":        true,
		"sourcewebsite":      true,
		"sourcehref":         true,
		"area_city_district": true,
		"projectname":        true,
		"projectcode":        true,
		"approvalno":         true,
		"projectscope":       true,
		"item":               true,
		"buyer":              true,
		"agency":             true,
		"budget":             true,
		"buyer_info":         true,
		"buyerperson":        true,
		"buyertel":           true,
		"buyeraddr":          true,
		"projectaddr":        true,
		"publishdept":        true,
		"funds":              true,
		"paymenttype":        true,
		"projectscale":       true,
		"bidmethod":          true,
		"bidopentime":        true,
		"agency_info":        true,
		"agencyperson":       true,
		"agencytel":          true,
		"agencyaddr":         true,
		"isppp":              true,
		"winner":             true,
		"winneraddr":         true,
		"winnerperson":       true,
		"winnertel":          true,
		"bidamount":          true,
		"currency":           true,
		"experts":            true,
		"bidamounttype":      true,
		"contractname":       true,
		"countryprojectcode": true,
		"contractnumber":     true,
		"projectperiod":      true,
		"signaturedate":      true,
		"multipackage":       true,
		"package":            true,
		"supervisorrate":     true,
		"jsoncontent":        true,
		"purchasinglist":     true,
		"toptype":            true,
		"subtype":            true,
		"winnerorder":        true,
		"bidopendate":        true,
		"bidtype":            true,
	}
	Bu = "_bu" //创建采历史爬虫后缀

)

func InitMgo() {
	defer qu.Catch()
	//MgoE = &mgo.MongodbSim{
	//	MongodbAddr: sp.Config.Dbaddr,
	//	DbName:      sp.Config.Dbname,
	//	Size:        10,
	//}
	//MgoE.InitPool()
	MgoEB = &mgo.MongodbSim{
		MongodbAddr: sp.Config.BidEditor.Addr,
		DbName:      sp.Config.BidEditor.Db,
		Size:        sp.Config.BidEditor.Size,
		UserName:    sp.Config.BidEditor.Username,
		Password:    sp.Config.BidEditor.Password,
	}
	MgoEB.InitPool()
	MgoS = &mgo.MongodbSim{
		MongodbAddr: sp.Config.Dbaddr,
		DbName:      sp.Config.Dbname2,
		Size:        10,
	}
	MgoS.InitPool()
}

// 初始化省市行政区划信息
func InitAreaCity() {
	//qu.ReadConfig("areacity.json", &Province)
	//Area = append(Area, "全国")
	//for area, _ := range Province {
	//	if area == "全国" {
	//		continue
	//	}
	//	Area = append(Area, area)
	//}
	Province = map[string][]string{}
	City = map[string][]string{}
	Area = append(Area, "全国")
	list, _ := MgoEB.Find("address", nil, nil, nil, false, -1, -1)
	for _, tmp := range *list {
		province := qu.ObjToString(tmp["province"])
		city := qu.ObjToString(tmp["city"])
		district := qu.ObjToString(tmp["district"])
		if province != "" && city == "" && district == "" { //area
			Area = append(Area, province)
		} else if province != "" && city != "" && district == "" { //city
			cityArr := Province[province]
			cityArr = append(cityArr, city)
			Province[province] = cityArr
		} else if province != "" && city != "" && district != "" { //district
			districtArr := City[city]
			districtArr = append(districtArr, district)
			City[city] = districtArr
		}
	}
}

// 爬虫整体测试时校验爬虫代码
func SpiderPassCheckLua(liststr, contentstr string, lua map[string]interface{}) (msg []string) {
	//校验含过滤方法stringFind但没有过滤注释“--关键词过滤”
	//if strings.Contains(liststr, "stringFind") && !strings.Contains(liststr, "--关键词过滤") {
	//	msg = append(msg, "列表页代码有过滤方法stringFind但缺少注释：--关键词过滤")
	//}
	//if strings.Contains(contentstr, "--关键词过滤") && !strings.Contains(contentstr, "delete") {
	//	msg = append(msg, `三级页代码有过滤方法但缺少data["delete"]="true"`)
	//}
	//if !strings.Contains(contentstr, "s_title") {
	//	msg = append(msg, "三级页缺少s_title")
	//}
	if !strings.Contains(contentstr, "getFileAttachmentsArrayWithTag") && !strings.Contains(contentstr, "downloadFile") {
		msg = append(msg, "三级页缺少下载附件方法")
	}
	//1.检测spidercode、site、channel
	//if param, ok := lua["param_common"].([]interface{}); ok && len(param) >= 3 {
	//	spidercode := qu.ObjToString(param[0])
	//	site := qu.ObjToString(param[1])
	//	channel := qu.ObjToString(param[2])
	//	checkText := fmt.Sprintf(CheckText, spidercode, site, channel)
	//	if strings.Contains(liststr, `item["spidercode"]`) && !strings.Contains(liststr, checkText) {
	//		msg = append(msg, "检查代码spidercode、site、channel字段值")
	//	}
	//}
	//2.检测https
	//isHttps := false
	//for _, text := range DomainNameReg.FindAllString(liststr, -1) {
	//	if strings.Contains(text, "https") {
	//		isHttps = true
	//	}
	//}
	//if isHttps {
	//	for tmpStr, tmpText := range map[string]string{"列表页": liststr, "三级页": contentstr} {
	//		downLoadText := DownLoadReg.FindString(tmpText)
	//		if downLoadText != "" {
	//			textArr := strings.Split(downLoadText, ",")
	//			if len(textArr) < 4 {
	//				msg = append(msg, tmpStr+"download方法添加下载参数")
	//			} else if len(textArr) == 4 {
	//				if !CodeTypeReg.MatchString(textArr[0]) || (textArr[1] != "true" && textArr[1] != "false") {
	//					msg = append(msg, tmpStr+"download方法添加下载参数")
	//				}
	//			}
	//		}
	//	}
	//}
	//3.检测title
	//if strings.Contains(liststr, `item["title"]="a"`) {
	//	if !strings.Contains(contentstr, `data["title"]`) {
	//		msg = append(msg, "检查代码title的完整性")
	//	}
	//}
	// 4.检测sendListNum
	//if !strings.Contains(liststr, "sendListNum") {
	//	msg = append(msg, "sendListNum方法缺失")
	//}
	return
}

// 爬虫整体测试时校验列表页和详情页内容
func SpiderPassCheckListAndDetail(result map[int64][]map[string]interface{}, data map[string]interface{}) (msg []string) {
	msgMap := map[string]bool{}
	//校验列表页信息
	for _, list := range result {
		for _, l := range list {
			//校验title
			title := qu.ObjToString(l["title"])
			if !TitleFilterReg1.MatchString(title) {
				msgMap["列表页title中无汉字"] = true
			} else if TitleFilterReg2.MatchString(title) {
				msgMap["列表页title中含有上(下)一页"] = true
			}
			//校验发布时间
			publishtime := qu.ObjToString(l["publishtime"])
			if publishtime == "0" || publishtime == "" {
				msgMap["列表页publishtime取值异常"] = true
			} else {
				t, err := time.ParseInLocation(qu.Date_Full_Layout, publishtime, time.Local)
				if err != nil || t.Unix() <= 0 {
					msgMap["列表页publishtime取值异常"] = true
				}
			}
		}
	}
	if len(data) > 0 {
		//校验publishtime
		if l_np_publishtime, ok := data["l_np_publishtime"].(lua.LNumber); ok {
			if l_np_publishtime <= 0 || l_np_publishtime > 0 && l_np_publishtime < 1000000000 {
				msgMap["三级页publishtime取值异常"] = true
			}
		} else if l_np_publishtime, ok := data["l_np_publishtime"].(int64); ok {
			if l_np_publishtime <= 0 || l_np_publishtime > 0 && l_np_publishtime < 1000000000 {
				msgMap["三级页publishtime取值异常"] = true
			}
		} else {
			msgMap["三级页publishtime值类型异常"] = true
		}
		contenthtml := qu.ObjToString(data["contenthtml"])
		if strings.Contains(contenthtml, "img") {
			msgMap["contenthtml中含有img是否下载"] = true
		}
		if strings.Contains(contenthtml, "iframe") {
			msgMap["contenthtml中含有iframe是否下载"] = true
		}
		detail := qu.ObjToString(data["detail"])
		if DetailFilterReg1.MatchString(detail) {
			msgMap["三级页正文提取包含无效内容"] = true
		}
		//校验jsondata
		if jsondata, ok := data["jsondata"].(map[string]interface{}); ok && len(jsondata) > 0 {
			for field, _ := range jsondata {
				if !JsonDataMap[field] {
					msgMap["jsondata中"+field+"属性错误"] = true
				}
			}
		}
	}
	for text, _ := range msgMap {
		msg = append(msg, text)
	}
	return
}

// 爬虫整体测试时校验列表页和详情页内容
func SpiderPassCheckListAndDetail_back(list []map[string]interface{}, data map[string]interface{}) (msg []string) {
	if len(list) > 0 {
		p_zero := 0
		h_flag := true
		n_flag := true
		l_flag := true
		for _, l := range list {
			//校验title
			title := qu.ObjToString(l["title"])
			if !TitleFilterReg1.MatchString(title) && h_flag {
				msg = append(msg, "列表页title中无汉字")
				h_flag = false
			} else if TitleFilterReg2.MatchString(title) && n_flag {
				msg = append(msg, "列表页title中含有上(下)一页")
				n_flag = false
			}
			publishtime := qu.ObjToString(l["publishtime"])
			if publishtime == "0" {
				p_zero++
			} else if l_flag {
				t, _ := time.ParseInLocation(qu.Date_Full_Layout, publishtime, time.Local)
				if t.Unix() <= 0 {
					msg = append(msg, "列表页数据发布时间异常")
					l_flag = false
				}
			}
		}
		if len(data) > 0 {
			//校验publishtime
			if l_np_publishtime := data["l_np_publishtime"].(lua.LNumber); l_np_publishtime <= 0 {
				msg = append(msg, "三级页发布时间小于0")
			} else if p_zero == len(list) && l_np_publishtime == 0 {
				msg = append(msg, "三级页发布时间异常")
			}
			contenthtml := qu.ObjToString(data["contenthtml"])
			if strings.Contains(contenthtml, "img") {
				msg = append(msg, "contenthtml中含有img是否下载")
			}
			detail := qu.ObjToString(data["detail"])
			if TitleFilterReg2.MatchString(detail) {
				msg = append(msg, "三级页正文提取异常")
			}
			//校验jsondata
			if jsondata, ok := data["jsondata"].(map[string]interface{}); ok && len(jsondata) > 0 {
				for field, _ := range jsondata {
					if !JsonDataMap[field] {
						msg = append(msg, "jsondata中"+field+"属性错误")
					}
				}
			}
		}
	}
	return
}

func GetLuasInfoBySite(site, area, city, district string) (domain, status, event, platform, infotype, specialtype string, remarktime int64) {
	shelveUp := 0
	eventMap, platformMap := map[int]interface{}{}, map[string]interface{}{}
	infoformatMap := map[int]bool{}
	eventArr, platformArr, infoformatArr := []string{}, []string{}, []string{}
	//areaMap := map[string]int{}
	//areaCityMap := map[string]map[string]int{}
	//cityDistrictMap := map[string]map[string]int{}
	domainMap := map[string]bool{}
	domainArr := []string{}
	remarktime = time.Now().Unix()
	//luas, _ := MgoE.Find("luaconfig", `{"param_common.1":"`+site+`"}`, ``, `{"model":1,"event":1,"state":1,"platform":1,"param_common":1,"comeintime":1}`, false, -1, -1)
	luas, _ := MgoEB.Find("luaconfig", `{"site":"`+site+`"}`, ``, `{"projecthref":1,"model":1,"event":1,"state":1,"platform":1,"param_common":1,"comeintime":1,"infoformat":1}`, false, -1, -1)
	arr := [][]map[string]interface{}{}
	for _, l := range *luas {
		update := []map[string]interface{}{}
		set := map[string]interface{}{}
		if b, ok := l["projecthref"].(bool); ok && b { //爬虫采集的数据是流程性信息
			specialtype = "含流程数据"
		}
		//更新爬虫area、city、district
		if area != "" {
			set["model.area"] = area
		}
		if area == "全国" {
			set["model.city"] = ""
			set["model.district"] = ""
		} else if area != "" {
			if city != "" {
				set["model.city"] = city
			}
			if district != "" {
				set["model.district"] = district
			}
		}
		if len(set) > 0 {
			update = append(update, map[string]interface{}{"_id": l["_id"]})
			update = append(update, map[string]interface{}{"$set": set})
			arr = append(arr, update)
		}
		//remarktime
		if comeintime := qu.Int64All(l["comeintime"]); comeintime != int64(0) && comeintime < remarktime {
			remarktime = comeintime
		}
		//domain
		paramCommon := l["param_common"].([]interface{})
		if len(paramCommon) >= 12 {
			href := qu.ObjToString(paramCommon[11])
			domain := DomainReg.FindString(href)
			if domain != "" {
				domain = ReplaceReg.ReplaceAllString(domain, "")
				if !domainMap[domain] {
					domainArr = append(domainArr, domain)
					domainMap[domain] = true
				}
			}
		}
		//state、event、platform
		state := qu.IntAll(l["state"])
		event := qu.IntAll(l["event"])
		platform := qu.ObjToString(l["platform"])
		if state == 5 || state == 11 { //5:lua已上架；11:python已上线
			shelveUp++
		}
		eventMap[event] = true
		platformMap[platform] = true
		//infoformat
		infoformat := qu.IntAll(l["infoformat"])
		infoformatMap[infoformat] = true
		//area、city、district
		//if model, ok := l["model"].(map[string]interface{}); ok && model != nil {
		//	a := qu.ObjToString(model["area"])
		//	c := qu.ObjToString(model["city"])
		//	d := qu.ObjToString(model["district"])
		//	if a != "" {
		//		areaMap[a] = areaMap[a] + 1
		//		if c != "" {
		//			if cityNum := areaCityMap[a]; cityNum != nil {
		//				cityNum[c] = cityNum[c] + 1
		//			} else {
		//				areaCityMap[a] = map[string]int{c: 1}
		//			}
		//			if d != "" {
		//				if distrctNum := cityDistrictMap[c]; distrctNum != nil {
		//					distrctNum[d] = distrctNum[d] + 1
		//				} else {
		//					cityDistrictMap[c] = map[string]int{d: 1}
		//				}
		//			}
		//		}
		//
		//	}
		//}
	}
	//domain
	domain = strings.Join(domainArr, ";")

	for e, _ := range eventMap {
		eventArr = append(eventArr, fmt.Sprint(e))
	}
	event = strings.Join(eventArr, ",")
	for p, _ := range platformMap {
		platformArr = append(platformArr, p)
	}
	sort.Strings(platformArr)
	platform = strings.Join(platformArr, ",")

	for infoformat, _ := range infoformatMap {
		text := "招标"
		if infoformat == 2 {
			text = "拟建/审批"
		} else if infoformat == 3 {
			text = "产权"
		} else if infoformat == 4 {
			text = "舆情"
		}
		infoformatArr = append(infoformatArr, text)
	}
	sort.Strings(infoformatArr)
	infotype = strings.Join(infoformatArr, ",")
	//
	status = fmt.Sprintf("%d%s%d", shelveUp, "/", len(*luas))
	//批量更新
	if len(arr) > 0 {
		MgoEB.UpdateBulk("luaconfig", arr...)
		arr = [][]map[string]interface{}{}
	}

	//an, cn, dn := 0, 0, 0
	//for at, num := range areaMap {
	//	if num > an {
	//		area = at
	//		an = num
	//	}
	//}
	//if area != "" {
	//	for ct, num := range areaCityMap[area] {
	//		if num > cn {
	//			city = ct
	//			cn = num
	//		}
	//	}
	//}
	//if city != "" {
	//	for dt, num := range cityDistrictMap[city] {
	//		if num > dn {
	//			district = dt
	//			dn = num
	//		}
	//	}
	//}
	return
}

type StringValSorter struct {
	Keys []string
	Vals []string
}

func MapStringValueSort(m map[string]string) *StringValSorter {
	vs := NewStringValSorter(m)
	vs.Sort()
	return vs
}

func NewStringValSorter(m map[string]string) *StringValSorter {
	vs := &StringValSorter{
		Keys: make([]string, 0, len(m)),
		Vals: make([]string, 0, len(m)),
	}
	for k, v := range m {
		vs.Keys = append(vs.Keys, k)
		vs.Vals = append(vs.Vals, v)
	}
	return vs
}

func (vs *StringValSorter) Sort() {
	sort.Sort(vs)
}

func (vs *StringValSorter) Len() int {
	return len(vs.Vals)
}
func (vs *StringValSorter) Less(i, j int) bool {
	return vs.Vals[i] < vs.Vals[j]
}
func (vs *StringValSorter) Swap(i, j int) {
	vs.Vals[i], vs.Vals[j] = vs.Vals[j], vs.Vals[i]
	vs.Keys[i], vs.Keys[j] = vs.Keys[j], vs.Keys[i]
}

type MyWrite struct {
	Byte *bytes.Buffer
}

func (m *MyWrite) Write(p []byte) (n int, err error) {
	n, err = m.Byte.Write(p)
	return
}

func (m *MyWrite) Reader(p []byte) (n int, err error) {
	n, err = m.Byte.Read(p)
	return
}

// 获取第day天凌晨的时间戳
func GetTime(day int) int64 {
	defer qu.Catch()
	nowTime := time.Now().AddDate(0, 0, day)
	timeStr := qu.FormatDate(&nowTime, qu.Date_Short_Layout)
	t, _ := time.ParseInLocation(qu.Date_Short_Layout, timeStr, time.Local)
	return t.Unix()
}