jy-data-extract/src/jy/pretreated/analykv.go

package pretreated

import (
	u "jy/util"
	//qu "qfw/util"
	"regexp"
	"strings"
)

var Han = regexp.MustCompile("[\\p{Han}]")
var Han1 = regexp.MustCompile("[^:;；,：，。． \u3000\u2003\u00a0\\s]")
var Han2 = regexp.MustCompile("[^:;；,：，。.]")
var Key = regexp.MustCompile("[:：：]")
var Time = regexp.MustCompile("[\\d]")
var dh = regexp.MustCompile("[，,.]")
var space = regexp.MustCompile("[\\s\\n \u3000\u2003\u00a0]+")
var val = regexp.MustCompile("[^\\s\\n \u3000\u2003\u00a0,，。！；;]")
var matchkh = map[string]string{
	"(": ")",
	"（": "）",
	"【": "】",
	"[": "]",
	"［": "］",
	"〖": "〗",
}

func GetKvFromtxt(con, tag string, from int, isSite bool, codeSite string) ([]*u.Kv, map[string][]*u.Tag) {
	res := FindKv(TextAfterRemoveTable(con), tag, from)
	kvs := []*u.Kv{}
	for _, k := range res.Keys {
		v, _ := res.Map[k].(string)
		if k != "" && v != "" {
			kvs = append(kvs, &u.Kv{
				Key:   k,
				Value: v,
			})
		}
	}
	kvTags := GetKvTags(kvs, tag, nil, isSite, codeSite)
	return kvs, kvTags
}

type Line struct {
	PreLine  *Line
	NextLine *Line
	Strs     []string
	Str      string
	Pos      int
	Len      int
	KV       *SortMap
	IsKey    bool //是否只是key
	Kn       int  //冒号个数
	Spacen   int  //间隔空格个数
	DJh      int  //逗号句号
}

func NewLine() *Line {
	return &Line{
		Strs: []string{},
		KV:   NewSortMap(),
	}
}

var LineKey = regexp.MustCompile("^[^，。]{2,10}[:：]$")
var DJh = regexp.MustCompile("[,，。]")
var DunH = regexp.MustCompile("[、.]")

func GetLines(con string) (res []*Line) {
	res = []*Line{}
	l1 := NewLine()
	strings.IndexFunc(con, func(r rune) bool {
		if r == 10 {
			if len(l1.Strs) > 0 {
				l1.Str = strings.Join(l1.Strs, "")
				if !regexp.MustCompile("^[，,.。\\s \u3000\u2003\u00a0]$").MatchString(l1.Str) {
					l1.Str = u.TrimLRSpace(l1.Str, "")
					l1.Str = TimeHM.ReplaceAllString(l1.Str, ReplTimeHM)
					l1.Strs = strings.Split(l1.Str, "")
					res = append(res, l1)
				}
				l1 = NewLine()
			}
		} else {
			s := string(r)
			l1.Strs = append(l1.Strs, s)
		}
		return false
	})
	if len(l1.Strs) > 0 {
		res = append(res, l1)
	}
	for k, l := range res {
		if k == 0 && k < len(res)-1 {
			l.NextLine = res[k+1]
		} else if k == len(res)-1 {
			l.PreLine = res[k-1]
		} else {
			l.PreLine = res[k-1]
			l.NextLine = res[k+1]
		}
		if LineKey.MatchString(l.Str) {
			l.IsKey = true
		} else {
			l.Kn = len(Key.FindAllString(l.Str, -1))
			l.DJh = len(DJh.FindAllString(l.Str, -1))
		}
	}
	return
}

func FindKv_v2(con, tag string) (m *SortMap) {
	m = NewSortMap()
	resLine := GetLines(con)
	for i := 0; i < len(resLine); i++ {
		l1 := resLine[i]
		if l1.IsKey {
			continue
		} else {
			if l1.Kn > 0 {
				u.Debug("=--=", l1.Str)
			} else {
				if l1.Spacen == 1 && l1.DJh < 2 && l1.Len < 50 {
					u.Debug("===", l1.Str)
				} else {
					u.Debug("???", l1.Str)
				}
			}
		}
	}
	return
}

var TimeHM = regexp.MustCompile("[\\s \u3000\u2003\u00a0]*([01]{0,1}[0123456789]|2[0123])[：:]([012345][0-9])[:：]{0,1}")
var ReplTimeHM = "D${1}H${2}M"

//from 1--全文 2--table td
func FindKv(con, tag string, from int) (m *SortMap) {
	if from == 2 || from == 3 {
		con = RemoveWarpOfTdVal(con)
	}
	//FindKv_v2(con, tag)
	matchMap := map[string]map[string]bool{
		"代理机构": map[string]bool{},
		"中标单位": map[string]bool{},
		"采购单位": map[string]bool{},
	}
	doubtMap := map[int]bool{}
	m = NewSortMap()
	strs := [][]string{}
	s1 := []string{}
	//断行
	strings.IndexFunc(con, func(r rune) bool {
		if r == 10 || r == 59 {
			if len(s1) > 0 {
				str := strings.Join(s1, "")
				str = u.TrimLRSpace(str, "")
				str = TimeHM.ReplaceAllString(str, ReplTimeHM)
				s1 = strings.Split(str, "")
				if len(s1) > 0 {
					strs = append(strs, s1)
				}
				s1 = []string{}
			}
		} else {
			s := string(r)
			s1 = append(s1, s)
		}
		return false
	})
	if len(s1) > 0 {
		str := strings.Join(s1, "")
		str = u.TrimLRSpace(str, "")
		str = TimeHM.ReplaceAllString(str, ReplTimeHM)
		s1 = strings.Split(str, "")
		if len(s1) > 0 {
			strs = append(strs, s1)
		}
	}
	//查找
	LastStr := ""
	for k0 := 0; k0 < len(strs); k0++ {
		s1 := strs[k0]
		//u.Debug(strings.Join(s1, ""))
		str1 := strings.Join(s1, "")
		k := ""
		v := ""
		flag := 0
		pos1, pos2 := -1, -1
		bkh := false
		skh := ""
		if from == 1 && DoubtReg.MatchString(str1) {
			doubtMap[k0] = true
		}
		if !Key.MatchString(str1) { //此行没有冒号
			if k0 > 0 {
				tm1 := strs[k0-1]
				if len([]rune(LastStr)) > 2 && len(tm1) < 8 && Key.MatchString(tm1[len(tm1)-1:][0]) && len([]rune(str1)) < 30 {
					//u.Debug(LastStr, str1)
					k = strings.Join(tm1[:len(tm1)-1], "")
					v = str1
					if k0 < len(strs)-1 {
						s2 := u.TrimLRSpace(strings.Join(strs[k0+1], ""), "")
						if len([]rune(s2)) < 10 && !regexp.MustCompile("^[0-9]+[、]+$").MatchString(s2) && !Key.MatchString(s2) {
							v += s2
							k0++
						}
					}
					keydetail(k, v, m, tag, k0, strs, matchMap, from, doubtMap)
				}
			}
			LastStr = ""
			continue
		} else {
			//u.Debug("---===----", str1)
			LastStr = str1
			for k1 := 0; k1 < len(s1); k1++ {
				s := s1[k1]
				if matchkh[s] != "" {
					skh = matchkh[s]
					bkh = true
				}
				if bkh {
					if skh == s {
						bkh = false
					}
					if flag == 1 {
						k += s
					} else if flag == 2 {
						v += s
					}
					continue
				}
				if flag == 0 {
					k = ""
					v = ""
					pos1, pos2 = -1, -1
					flag = 1
				}
				if flag == 1 {
					if Han1.MatchString(s) || (k != "" && Han2.MatchString(s)) {
						k += s
					} else if Key.MatchString(s) && k != "" {
						flag = 2
					} else {
						flag = 0
					}
				} else if flag == 2 {
					if val.MatchString(s) || (dh.MatchString(s) && k1 > 0 && k1 < len(s1)-1 && Time.MatchString(s1[k1-1]) && Time.MatchString(s1[k1+1])) {
						if pos1 < 0 {
							pos1 = k1
						}
						continue
					} else {
						be := false
						if space.MatchString(s) {
							temp := s1[k1+1:]
							//()（）[]【】
							m1 := k1
							bkh1 := false
							skh1 := ""
							for k2, v2 := range temp {
								if k2 == len(temp)-1 {
									be = true
								}
								if matchkh[v2] != "" {
									bkh1 = true
									skh1 = matchkh[v2]
									continue
								} else if bkh1 {
									if v2 == skh1 {
										bkh1 = false
									}
									continue
								} else if space.MatchString(v2) {
									continue
								} else if !val.MatchString(v2) {
									k1 = m1 + k2 + 1
									break
								} else {
									if pos1 < 0 {
										//u.Debug("-----", pos1)
										pos1 = k1 + k2 + 1
									}
								}
								if Key.MatchString(v2) && k2 > 0 && k2 < len(temp)-1 {
									if Time.MatchString(temp[k2-1]) && Time.MatchString(temp[k2+1]) {
										//u.Debug(v2, temp[k2-1], temp[k2+1])
										k1 = m1 + k2 + 1
									} else {
										//倒着
										for i := k2; i > k1-m1-1; i-- {
											if !val.MatchString(temp[i]) {
												k1 = m1 + i + 1
												break
											}
										}
										break
									}
								}
							}
						}
						if be {
							k1 = len(s1) //直接跳到最后
						}
						if pos2 < 0 && pos2 < pos1 {
							pos2 = k1
						}
						//					u.Debug(pos1, pos2, k1, len(s1))
						if pos1 > -1 && pos2 > pos1 {
							v = strings.Join(s1[pos1:pos2], "")
							flag = 0
							keydetail(k, v, m, tag, k0, strs, matchMap, from, doubtMap)
						} else {
							//u.Debug(k, pos1, pos2)
							flag = 0
						}
					}
				}
			}
			if flag == 2 {
				if pos2 > pos1 {
					v = strings.Join(s1[pos1:pos2], "")
				} else if pos1 > 0 {
					v = strings.Join(s1[pos1:], "")
				}
				if v != "" {
					flag = 0
					keydetail(k, v, m, tag, k0, strs, matchMap, from, doubtMap)
				}
				//u.Debug(k, v)
			}
		}
	}
	//	for _, kk := range m.Keys {
	//		u.Debug(kk, m.Map[kk])
	//	}
	return
}

func keydetail(k, v string, m *SortMap, tag string, pos int, strs [][]string, matchMap map[string]map[string]bool, from int, doubtMap map[int]bool) {
	if regexp.MustCompile("^[0-9]+[、]+$").MatchString(v) {
		return
	}
	k = space.ReplaceAllString(k, "")
	if len([]rune(k)) > 1 {
		if len([]rune(k)) < 5 && strings.HasPrefix(k, "联系") || ContactInfoMustReg.MatchString(k) {
			/*
				5ded053fe9d1f601e4c9e3ee
				中标人：XXXXXXXXXXXXXXXXXX
				相关竞价人对成交结果有异议的，可自本公告发布之日起三日内书面提出。
				联系方式：卢明珠　0871-66136373
			*/
			if doubtMap[pos-1] && len(m.Map) == 1 { //当识别到中标、采购、代理标签后，对其后的联系人、电话等信息判断是否属于该标签
				goto L
			}
			num := 0
			bf := false
			for i := len(m.Keys) - 1; i > -1; i-- {
				//u.Debug("k", k)
				num++
				if from == 1 && !ContactType["代理机构"].MatchString(k) && ContactType["代理机构"].MatchString(m.Keys[i]) && !IsContactKvHandle(k, matchMap["代理机构"]) {
					matchMap["代理机构"][k] = true
					k = "代理机构" + k
					bf = true
					break
				}
				//if !filter_zbdw_ky.MatchString(k) && filter_zbdw_ky.MatchString(m.Keys[i]) && !IsContactKvHandle(k, matchMap["中标单位"]) {
				if from == 1 && !ContactType["中标单位"].MatchString(k) && ContactType["中标单位"].MatchString(m.Keys[i]) && !IsContactKvHandle(k, matchMap["中标单位"]) {
					matchMap["中标单位"][k] = true
					k = "中标单位" + k
					bf = true
					break
				}
				if from == 1 && !ContactType["采购单位"].MatchString(k) && ContactType["采购单位"].MatchString(m.Keys[i]) && !IsContactKvHandle(k, matchMap["采购单位"]) {
					matchMap["采购单位"][k] = true
					k = "采购单位" + k
					bf = true
					break
				}
				//if num > 0 {
				break
				//}
			}
			if !bf {
				//k = "采购人" + k
				//取出上一行
				if pos > 0 {
					if len(strs[pos-1]) < 20 {
						str := space.ReplaceAllString(strings.Join(strs[pos-1], ""), "")
						if from == 1 && ContactType["代理机构"].MatchString(str) && !IsContactKvHandle(k, matchMap["代理机构"]) {
							matchMap["代理机构"][k] = true
							k = "代理机构" + k
						} else if from == 1 && ContactType["中标单位"].MatchString(str) && !IsContactKvHandle(k, matchMap["中标单位"]) {
							matchMap["中标单位"][k] = true
							k = "中标单位" + k
						} else if from == 1 && ContactType["采购单位"].MatchString(str) && !IsContactKvHandle(k, matchMap["采购单位"]) {
							matchMap["采购单位"][k] = true
							k = "采购单位" + k
						}
					}
				}
			}
		} else if len([]rune(k)) == 2 {
			if filter_zbje_jd.MatchString(k) { //钱
				if tag != "" && filter_tag_zb.MatchString(tag) {
					k = "中标" + k
				} else {
					num := 0
					for i := len(m.Keys) - 1; i > -1; i-- {
						num++
						if filter_zbdw_ky.MatchString(m.Keys[i]) {
							k = "中标" + k
							break
						}
						if num > 2 {
							break
						}

					}
				}

			}

		}
	L:
		//qu.Debug(k, v)
		if m.Map[k] == nil {
			m.AddKey(k, v)
		} else {
			vals := []string{}
			if vvv, ok := m.Map[k].([]string); ok {
				vals = append(vals, vvv...)
			} else {
				vals = append(vals, v)
			}
			vals = append(vals, v)
			m.AddKey(k, vals)
		}
	}
}

//时间处理、换行优先级|
func FindKv_v1(con string) (m map[string]string) {
	m = map[string]string{}
	k := ""
	v := ""
	flag := 0
	strings.IndexFunc(con, func(r rune) bool {
		s := string(r)
		if flag == 0 {
			k = ""
			v = ""
			flag = 1
		}
		if flag == 1 {
			if Han.MatchString(s) {
				k += s
			} else if Key.MatchString(s) && k != "" {
				flag = 2
			} else {
				flag = 0
			}
		} else if flag == 2 {
			if v == "" {
				if space.MatchString(s) {

				} else if val.MatchString(s) && !Key.MatchString(s) {
					v += s
				} else {
					flag = 0
				}
			} else {
				if val.MatchString(s) {
					if Key.MatchString(k) {
						if (regexp.MustCompile("(时间|日期)").MatchString(v) || regexp.MustCompile("(时间|日期)").MatchString(k)) && regexp.MustCompile("[^\\d][012]?[0-9]").MatchString(k) {
							v += s
						} else if regexp.MustCompile("^[\\p{Han}]$").MatchString(v) {
							k = v
							v = ""
							flag = 1
						}
					} else {
						v += s
					}
				} else if k != "" && v != "" {
					u.Debug(k, "=", v)
					flag = 0
				}
			}
		}
		return false
	})
	if flag == 2 && k != "" && v != "" {
		u.Debug(k, "=", v)
	}
	return
}