jy-data-extract/src/jy/extract/extractudp.go

// extractudp
package extract

import (
	"encoding/json"
	"fmt"
	db "jy/mongodbutil"
	ju "jy/util"
	mu "mfw/util"
	"net"
	qu "qfw/util"
	"sync"

	log "github.com/donnie4w/go-logger/logger"
	"gopkg.in/mgo.v2/bson"
)

var Udpclient mu.UdpClient //udp对象
var nextNodes []map[string]interface{}

//udp通知抽取
func ExtractUdp() {
	nextNodes = qu.ObjArrToMapArr(ju.Config["nextNode"].([]interface{}))
	Udpclient = mu.UdpClient{Local: ":" + qu.ObjToString(ju.Config["udpport"]), BufSize: 1024}
	Udpclient.Listen(processUdpMsg)
}

func processUdpMsg(act byte, data []byte, ra *net.UDPAddr) {
	switch act {
	case mu.OP_TYPE_DATA:
		var rep map[string]interface{}
		err := json.Unmarshal(data, &rep)
		if err != nil {
			log.Debug(err)
		} else {
			sid, _ := rep["gtid"].(string)
			eid, _ := rep["lteid"].(string)
			stype, _ := rep["stype"].(string)
			if sid == "" || eid == "" {
				log.Debug("err", "sid=", sid, "eid=", eid)
			} else {
				go Udpclient.WriteUdp([]byte("udpok"), mu.OP_NOOP, ra)
				if stype == "distributed" { //分布式抽取分支
					log.Debug("分布式抽取id段", sid, eid)
					InstanceId := qu.ObjToString(rep["InstanceId"])
					db.Mgo.Update("ecs", `{"InstanceId":"`+InstanceId+`"}`,
						map[string]interface{}{
							"$set": map[string]interface{}{
								"extstatus": "running",
							},
						}, true, false)
					ExtractByUdp(sid, eid, qu.ObjToString(rep["InstanceId"]))
					db.Mgo.Update("ecs", `{"InstanceId":"`+InstanceId+`"}`,
						map[string]interface{}{
							"$set": map[string]interface{}{
								"extstatus": "ok",
							},
						}, true, false)
					log.Debug("分布式抽取完成", sid, eid, "释放esc实例", qu.ObjToString(rep["ip"]))
				} else {
					log.Debug("udp通知抽取id段", sid, eid)
					ExtractByUdp(sid, eid)
					log.Debug("udp通知抽取完成,eid=", eid)
					for _, m := range nextNodes {
						by, _ := json.Marshal(map[string]interface{}{
							"gtid":  sid,
							"lteid": eid,
							"stype": qu.ObjToString(m["stype"]),
						})
						err := Udpclient.WriteUdp(by, mu.OP_TYPE_DATA, &net.UDPAddr{
							IP:   net.ParseIP(m["addr"].(string)),
							Port: qu.IntAll(m["port"]),
						})
						if err != nil {
							log.Debug(err)
						}
					}
				}
			}
		}
	case mu.OP_NOOP: //下个节点回应
		var rep map[string]interface{}
		err := json.Unmarshal(data, &rep)
		if err != nil {
			log.Debug(err)
		} else {
			log.Debug(rep)
		}
	}
}

var ext *ExtractTask

//根据id区间抽取
func ExtractByUdp(sid, eid string, instanceId ...string) {
	defer qu.Catch()
	if ext == nil {
		ext = &ExtractTask{}
		ext.Id = qu.ObjToString(ju.Config["udptaskid"])
		ext.InitTaskInfo()
		ext.TaskInfo.FDB = db.MgoFactory(3, 5, 600, ext.TaskInfo.FromDbAddr, ext.TaskInfo.FromDB)
		ext.TaskInfo.TDB = db.MgoFactory(3, 5, 600, ext.TaskInfo.ToDbAddr, ext.TaskInfo.ToDB)
		ext.InitRulePres()
		ext.InitRuleBacks()
		ext.InitRuleCore()
		ext.InitBlockRule()
		ext.InitTag()
		ext.InitClearFn()
		if ext.IsExtractCity { //版本上控制是否开始城市抽取
			//初始化城市DFA信息
			ext.InitCityDFA()
			ext.InitAreaCode()
			ext.InitPostCode()
		}
		//质量审核
		ext.InitAuditFields()
		ext.InitAuditRule()
		ext.InitAuditClass()
		ext.InitAuditRecogField()

		//品牌抽取是否开启
		ju.IsBrandGoods, _ = ju.Config["brandgoods"].(bool)

		ext.ResultSave(true)
		ext.BidSave(true)
		ext.IsRun = true
	}
	if len(instanceId) > 0 { //分布式抽取进度
		query := bson.M{"_id": bson.M{"$gte": bson.ObjectIdHex(sid), "$lte": bson.ObjectIdHex(eid)}}
		count1 := ext.TaskInfo.FDB.Count(ext.TaskInfo.FromColl, query)
		count2 := ext.TaskInfo.FDB.Count(ext.TaskInfo.FromColl+"_back", query)
		count := count1 + count2
		pageNum := (count + PageSize - 1) / PageSize
		limit := PageSize
		if count < PageSize {
			limit = count
		}
		fmt.Printf("count=%d,pageNum=%d,query=%v", count, pageNum, query)

		startI := 0 //接着上次任务执行
		sidback := sid
		esc, _ := db.Mgo.FindOne("ecs", `{"InstanceId":"`+instanceId[0]+`"}`)
		startI = qu.IntAll((*esc)["pagecurrent"])
		if qu.ObjToString((*esc)["lastId"]) != "" {
			sid = qu.ObjToString((*esc)["lastId"])
		}
		if qu.ObjToString((*esc)["lastIdback"]) != "" {
			sidback = qu.ObjToString((*esc)["lastIdback"])
		}

		for i := startI; i < pageNum; i++ {
			query = bson.M{"_id": bson.M{"$gte": bson.ObjectIdHex(sid), "$lte": bson.ObjectIdHex(eid)}}
			fmt.Printf("page=%d,query=%v", i+1, query)
			if ext.TaskInfo.FDB.Count(ext.TaskInfo.FromColl, query) > 0 {
				list, _ := ext.TaskInfo.FDB.Find(ext.TaskInfo.FromColl, query, nil, Fields, false, 0, limit)
				for _, v := range *list {
					if qu.ObjToString(v["sensitive"]) != "" { //去除含敏感词数据
						continue
					}
					_id := qu.BsonIdToSId(v["_id"])
					log.Debug(_id)
					var j, jf *ju.Job
					if ext.IsFileField && v["projectinfo"] != nil {
						v["isextFile"] = true
						j, jf = ext.PreInfo(v)
					} else {
						j, _ = ext.PreInfo(v)
					}
					ext.TaskInfo.ProcessPool <- true
					go ext.ExtractProcess(j, jf)
					sid = _id
				}
				db.Mgo.Update("ecs", `{"InstanceId":"`+instanceId[0]+`"}`,
					map[string]interface{}{"$set": map[string]interface{}{
						"lastId": sid,
					}}, true, false)
			}
			queryback := bson.M{"_id": bson.M{"$gte": bson.ObjectIdHex(sidback)}}
			fmt.Printf("page=%d,queryback=%v", i+1, queryback)
			if ext.TaskInfo.FDB.Count(ext.TaskInfo.FromColl+"_back", queryback) > 0 {
				list2, _ := ext.TaskInfo.FDB.Find(ext.TaskInfo.FromColl+"_back", queryback, nil, Fields, false, 0, limit)
				for _, v := range *list2 {
					if qu.ObjToString(v["sensitive"]) != "" { //去除含敏感词数据
						continue
					}
					_id := qu.BsonIdToSId(v["_id"])
					log.Debug(_id)
					var j, jf *ju.Job
					if ext.IsFileField && v["projectinfo"] != nil {
						v["isextFile"] = true
						j, jf = ext.PreInfo(v)
					} else {
						j, _ = ext.PreInfo(v)
					}
					ext.TaskInfo.ProcessPool <- true
					go ext.ExtractProcess(j, jf)
					sidback = _id
				}
				db.Mgo.Update("ecs", `{"InstanceId":"`+instanceId[0]+`"}`,
					map[string]interface{}{"$set": map[string]interface{}{
						"lastIdback": sidback,
					}}, true, false)
			}
			db.Mgo.Update("ecs", `{"InstanceId":"`+instanceId[0]+`"}`,
				map[string]interface{}{"$set": map[string]interface{}{
					"pagetotal":   pageNum,
					"pagecurrent": i + 1,
				}}, true, false)
		}
	} else { //普通抽取
		query := bson.M{"_id": bson.M{"$gte": bson.ObjectIdHex(sid), "$lte": bson.ObjectIdHex(eid)}}
		count := ext.TaskInfo.FDB.Count(ext.TaskInfo.FromColl, query)
		pageNum := (count + PageSize - 1) / PageSize
		limit := PageSize
		if count < PageSize {
			limit = count
		}
		wg := sync.WaitGroup{}
		for i := 0; i < pageNum; i++ {
			query = bson.M{"_id": bson.M{"$gte": bson.ObjectIdHex(sid)}}
			fmt.Printf("page=%d,query=%v", i+1, query)
			list, _ := ext.TaskInfo.FDB.Find(ext.TaskInfo.FromColl, query, nil, Fields, false, 0, limit)
			for k, v := range *list {
				if qu.ObjToString(v["sensitive"]) != "" { //去除含敏感词数据
					continue
				}
				_id := qu.BsonIdToSId(v["_id"])
				var j, jf *ju.Job
				if ext.IsFileField && v["projectinfo"] != nil {
					v["isextFile"] = true
					j, jf = ext.PreInfo(v)
				} else {
					j, _ = ext.PreInfo(v)
				}
				ext.TaskInfo.ProcessPool <- true
				wg.Add(1)
				go func() {
					defer wg.Done()
					ext.ExtractProcess(j, jf)
				}()
				if k%1000 == 0 {
					log.Debug(i, k, _id)
				}
				sid = _id
			}
		}
		wg.Wait()
		ext.BidSave(false)
	}
}