jy-data-extract/fullproject/src_dev5/task.go

package main

import (
	"encoding/json"
	"log"
	"qfw/util"
	"regexp"
	"strings"
	"sync"
	//"strings"
	"time"

	"github.com/robfig/cron"
)

/**
任务入口
全量、增量合并
更新、插入，内存清理
转换成info对象
**/

//项目合并对象
type ProjectTask struct {
	InitMinTime int64 //最小时间，小于0的处理一次
	name        string
	thread      int //线程数
	//查找锁
	findLock sync.Mutex
	wg       sync.WaitGroup
	//map锁
	AllIdsMapLock sync.Mutex
	//对应的id
	AllIdsMap map[string]*ID
	//采购单位、项目名称、项目编号
	mapPb, mapPn, mapPc map[string]*Key
	//更新或新增通道
	updatePool chan []map[string]interface{}
	//表名
	coll string
	//当前状态是全量还是增量
	currentType string //当前是跑全量还是跑增量
	//
	clearContimes int
	//当前时间
	currentTime int64
	//保存长度
	saveSize int
}

func NewPT() *ProjectTask {
	return &ProjectTask{
		InitMinTime: int64(1325347200),
		name:        "全/增量对象",
		thread:      3,
		updatePool:  make(chan []map[string]interface{}, 2000),
		wg:          sync.WaitGroup{},
		AllIdsMap:   make(map[string]*ID, 5000000),
		mapPb:       make(map[string]*Key, 5000000),
		mapPn:       make(map[string]*Key, 5000000),
		mapPc:       make(map[string]*Key, 5000000),
		saveSize:    200,
		coll:        ProjectColl,
	}
}

var P_QL *ProjectTask

//初始化全量合并对象
func init() {
	P_QL = NewPT()
	go P_QL.updateQueue()
	go P_QL.clearMem()

}

//项目保存和更新通道
func (p *ProjectTask) updateQueue() {
	arr := make([][]map[string]interface{}, p.saveSize)
	index := 0
	for {
		select {
		case v := <-p.updatePool:
			arr[index] = v
			index++
			if index == p.saveSize {
				MongoTool.UpSertBulk(p.coll, arr...)
				arr = make([][]map[string]interface{}, p.saveSize)
				index = 0
			}
		case <-time.After(2 * time.Second):
			if index > 0 {
				MongoTool.UpSertBulk(p.coll, arr[:index]...)
				arr = make([][]map[string]interface{}, p.saveSize)
				index = 0
			}
		}
	}
}

//项目合并内存更新
func (p *ProjectTask) clearMem() {
	c := cron.New()
	//在内存中保留最近6个月的信息
	validTime := int64(6 * 30 * 86400)
	//跑全量时每4分钟跑一次，跑增量时400分钟跑一次
	c.AddFunc("50 0/4 * * * *", func() {
		if p.currentType == "ql" || p.clearContimes >= 100 {
			//跳过的次数清零
			p.clearContimes = 0
			//信息进入查找对比全局锁
			p.findLock.Lock()
			defer p.findLock.Unlock()
			//合并进行的任务都完成
			p.wg.Wait()
			//遍历id
			//所有内存中的项目信息
			p.AllIdsMapLock.Lock()
			defer p.AllIdsMapLock.Unlock()
			//清除计数
			clearNum := 0
			for k, v := range p.AllIdsMap {
				if p.currentTime-v.P.LastTime > validTime {
					clearNum++
					//删除id的map
					delete(p.AllIdsMap, k)
					//删除pb
					if v.P.Buyer != "" {
						ids := p.mapPb[v.P.Buyer]
						if ids != nil {
							ids.Lock.Lock()
							ids.Arr = deleteSlice(ids.Arr, k)
							if len(ids.Arr) == 0 {
								delete(p.mapPb, v.P.Buyer)
							}
							ids.Lock.Unlock()
						}
					}
					//删除mapPn
					for _, vn := range append([]string{v.P.ProjectName}, v.P.MPN...) {
						if vn != "" {
							ids := p.mapPn[vn]
							if ids != nil {
								ids.Lock.Lock()
								ids.Arr = deleteSlice(ids.Arr, k)
								if len(ids.Arr) == 0 {
									delete(p.mapPn, vn)
								}
								ids.Lock.Unlock()
							}
						}
					}
					//删除mapPc
					for _, vn := range append([]string{v.P.ProjectCode}, v.P.MPC...) {
						if vn != "" {
							ids := p.mapPc[vn]
							if ids != nil {
								ids.Lock.Lock()
								ids.Arr = deleteSlice(ids.Arr, k)
								if len(ids.Arr) == 0 {
									delete(p.mapPc, vn)
								}
								ids.Lock.Unlock()
							}
						}
					}
					v = nil
				}
			}
			log.Println("清除完成:", clearNum, len(p.AllIdsMap))
		} else {
			p.clearContimes++
		}
	})
	c.Start()
	select {}
}

//全量合并
func (p *ProjectTask) taskQl(udpInfo map[string]interface{}) {
	defer util.Catch()
	//1、检查pubilshtime索引
	db, _ := udpInfo["db"].(string)
	if db == "" {
		db = MongoTool.DbName
	}
	coll, _ := udpInfo["coll"].(string)
	if coll == "" {
		coll = ExtractColl
	}
	sess := MongoTool.GetMgoConn()
	bcon := false
	if sess.DB(db).C(coll).EnsureIndexKey("publishtime_1", "publishtime_-1") == nil {
		bcon = true
	} else {
		log.Println("publishtime_1索引不存在")
	}
	MongoTool.DestoryMongoConn(sess)
	thread := util.IntAllDef(udpInfo["thread"], 3)
	if thread > 0 {
		p.thread = thread
	}
	if bcon {
		//生成查询语句执行
		p.enter(db, coll, map[string]interface{}{})

	}
}

//增量合并
func (p *ProjectTask) taskZl(udpInfo map[string]interface{}) {
	defer util.Catch()
	//1、检查pubilshtime索引
	db, _ := udpInfo["db"].(string)
	if db == "" {
		db = MongoTool.DbName
	}
	coll, _ := udpInfo["coll"].(string)
	if coll == "" {
		coll = ExtractColl
	}
	thread := util.IntAllDef(udpInfo["thread"], 3)
	if thread > 0 {
		p.thread = thread
	}
	//开始id和结束id
	q, _ := udpInfo["query"].(map[string]interface{})
	if q == nil {
		q = map[string]interface{}{
			"_id": map[string]interface{}{
				"$gt":  util.StringTOBsonId(udpInfo["gtid"].(string)),
				"$lte": util.StringTOBsonId(udpInfo["lteid"].(string)),
			},
		}
	}
	if q != nil {
		//生成查询语句执行
		p.enter(db, coll, q)
	}

}

func (p *ProjectTask) enter(db, coll string, q map[string]interface{}) {
	defer util.Catch()
	sess := MongoTool.GetMgoConn()
	defer MongoTool.DestoryMongoConn(sess)
	query := sess.DB(db).C(coll).Find(q).Sort("publishtime").Iter()
	pool := make(chan bool, p.thread)
	count := 0
	for tmp := make(map[string]interface{}); query.Next(tmp); count++ {
		info := ParseInfo(tmp)
		if info != nil && !((info.pnbval == 1 && info.Buyer != "") || info.pnbval == 0) {
			pool <- true
			go func(info *Info, tmp map[string]interface{}) {
				defer func() {
					p.currentTime = info.Publishtime
					<-pool
				}()
				p.startProjectMerge(info, tmp)
			}(info, tmp)
		} else {
			//信息错误,进行更新
		}
		if count%1000 == 0 {
			log.Println("current", count)
		}
		tmp = make(map[string]interface{})
	}
	//阻塞
	for n := 0; n < p.thread; n++ {
		pool <- true
	}
	log.Println("所有线程执行完成...", count)

}

var (
	//从标题获取项目编号
	titleGetPc  = regexp.MustCompile("^([-0-9a-zA-Z第号采招政询电审竞#]{8,}[-0-9a-zA-Z#]+)")
	titleGetPc1 = regexp.MustCompile("[\\[【(（](.{0,6}(编号|编码|项号|包号|代码|标段?号)[:：为])?([-0-9a-zA-Z第号采招政询电审竞#]{5,}([\\[\\]()（）][-0-9a-zA-Z第号采招审竞#]+[\\[\\]()（）][-0-9a-zA-Z第号采招审竞#]+)?)[\\]】)）]")
	titleGetPc2 = regexp.MustCompile("([-0-9a-zA-Z第号采政招询电审竞#]{8,}[-0-9a-zA-Z#]+)(.{0,5}公告)?$")
	//项目编号过滤
	pcReplace = regexp.MustCompile("([\\[【(（〖〔《｛﹝{](重|第?[二三四再]次.{0,4})[\\]】)）〗〕》｝﹞}])$|[\\[\\]【】()（）〖〗〔〕《》｛｝﹝﹞－；{}– 　]+|(号|重|第?[二三四五再]次(招标)?)$|[ _＿]+|((采购)?项目|采购(项目)?)$")
	//项目编号只是数字或只是字母4个以下
	StrOrNum = regexp.MustCompile("^[0-9_-]{1,4}$|^[a-zA-Z_-]{1,4}$")
	//纯数字或纯字母
	StrOrNum2 = regexp.MustCompile("^[0-9_-]+$|^[a-zA-Z_-]+$")
)

func ParseInfo(tmp map[string]interface{}) (info *Info) {
	bys, _ := json.Marshal(tmp)
	var thisinfo *Info
	json.Unmarshal(bys, &thisinfo)
	if thisinfo == nil {
		return nil
	}
	if len(thisinfo.Topscopeclass) == 0 {
		thisinfo.Topscopeclass = []string{}
	}
	if len(thisinfo.Subscopeclass) == 0 {
		thisinfo.Subscopeclass = []string{}
	}

	//从标题中查找项目编号
	res := titleGetPc.FindStringSubmatch(thisinfo.Title)
	if len(res) > 1 && len(res[1]) > 6 && thisinfo.ProjectCode != res[1] && !numCheckPc.MatchString(res[1]) && !_zimureg1.MatchString(res[1]) {
		thisinfo.PTC = res[1]
	} else {
		res = titleGetPc1.FindStringSubmatch(thisinfo.Title)
		if len(res) > 3 && len(res[3]) > 6 && thisinfo.ProjectCode != res[3] && !numCheckPc.MatchString(res[3]) && !_zimureg1.MatchString(res[3]) {
			thisinfo.PTC = res[3]
		} else {
			res = titleGetPc2.FindStringSubmatch(thisinfo.Title)
			if len(res) > 1 && len(res[1]) > 6 && thisinfo.ProjectCode != res[1] && !numCheckPc.MatchString(res[1]) && !_zimureg1.MatchString(res[1]) {
				thisinfo.PTC = res[1]
			}
		}
	}

	if thisinfo.ProjectName != "" && len([]rune(thisinfo.ProjectName)) > 0 {
		thisinfo.ProjectName = pcReplace.ReplaceAllString(thisinfo.ProjectName, "")
		if thisinfo.ProjectName != "" {
			thisinfo.pnbval++
		}
	}

	if thisinfo.ProjectCode != "" || thisinfo.PTC != "" {
		if thisinfo.ProjectCode != "" {
			thisinfo.ProjectCode = pcReplace.ReplaceAllString(thisinfo.ProjectCode, "")
			if thisinfo.pnbval == 0 && len([]rune(thisinfo.ProjectCode)) < 5 {
				thisinfo.ProjectCode = StrOrNum.ReplaceAllString(thisinfo.ProjectCode, "")
			}
		} else {
			thisinfo.PTC = pcReplace.ReplaceAllString(thisinfo.PTC, "")
			if thisinfo.pnbval == 0 && len([]rune(thisinfo.PTC)) < 5 {
				thisinfo.PTC = StrOrNum.ReplaceAllString(thisinfo.PTC, "")
			}
		}
		if thisinfo.ProjectCode != "" || thisinfo.PTC != "" {
			thisinfo.pnbval++
		}
	}
	if thisinfo.ProjectCode == thisinfo.PTC || strings.Index(thisinfo.ProjectCode, thisinfo.PTC) > -1 {
		thisinfo.PTC = ""
	}

	if thisinfo.Buyer != "" && len([]rune(thisinfo.Buyer)) > 2 {
		thisinfo.pnbval++
	} else {
		thisinfo.Buyer = ""
	}
	//winners整理
	winner, _ := tmp["winner"].(string)
	m1 := map[string]bool{}
	winners := []string{}
	if winner != "" {
		m1[winner] = true
		winners = append(winners, winner)
	}
	if thisinfo.HasPackage {
		packageM, _ := tmp["package"].(map[string]interface{})
		for _, p := range packageM {
			pm, _ := p.(map[string]interface{})
			pw, _ := pm["winner"].(string)
			if pw != "" {
				m1[pw] = true
				winners = append(winners, pw)
			}
		}
	}
	thisinfo.Winners = winners

	thisinfo.LenPC = len([]rune(thisinfo.ProjectCode))
	thisinfo.LenPTC = len([]rune(thisinfo.PTC))
	thisinfo.LenPN = len([]rune(thisinfo.ProjectName))
	return thisinfo
}

//从数组中删除元素
func deleteSlice(arr []string, v string) []string {
	for k, v1 := range arr {
		if v1 == v {
			return append(arr[:k], arr[k+1:]...)
		}
	}
	return arr
}