spider_editor/spider/script.go

/**
脚本加载+调用 封装，
前期走文件系统加载
后期走数据库配置，
LUA中公共的方法需要抽出来，主脚本文件加载LUA公共文件
*/
package spider

import (
	"bytes"
	"compress/gzip"
	"crypto/aes"
	"encoding/base64"
	"encoding/json"
	"io/ioutil"
	mu "mfw/util"
	"net/http"
	"net/url"
	"path"
	qu "qfw/util"
	"regexp"
	util "spiderutil"
	"strconv"
	"strings"
	"time"

	"golang.org/x/text/encoding/simplifiedchinese"
	"golang.org/x/text/transform"

	"github.com/cjoudrey/gluahttp"
	lujson "github.com/yuin/gopher-json"
	"github.com/yuin/gopher-lua"
)

//脚本
type Script struct {
	SCode, ScriptFile string
	Encoding          string
	Downloader        string //下载器
	Timeout           int64  //超时时间秒
	L                 *lua.LState
	Test_luareqcount  int //脚本请求次数
	Test_goreqtime    int //go发起次数(时间)
	Test_goreqlist    int //go发起次数(列表)
	Test_goreqcon     int //go发起次数(正文)
}

//加载文件
func (s *Script) LoadScript(downloadnode, script string, isfile ...string) {
	s.ScriptFile = script
	options := lua.Options{
		RegistrySize:        256 * 20,
		CallStackSize:       256,
		IncludeGoStackTrace: false,
	}
	s.L = lua.NewState(options)
	//s.L.ScriptFileName = s.SCode
	s.L.PreloadModule("http", gluahttp.NewHttpModule(&http.Client{}).Loader)
	s.L.PreloadModule("json", lujson.Loader)
	if len(isfile) > 0 {
		if err := s.L.DoFile(script); err != nil {
			panic("加载lua脚本错误" + err.Error())
		}
	} else {
		if err := s.L.DoString(script); err != nil {
			panic("加载lua脚本错误" + err.Error())
		}
	}

	s.Encoding = s.GetVar("spiderPageEncoding")
	//暴露go方法
	//download(url,head) 普通下载
	s.L.SetGlobal("download", s.L.NewFunction(func(S *lua.LState) int {
		head := S.ToTable(-1)
		url := S.ToString(-2)
		ishttps := S.ToBool(-3)
		charset := S.ToString(-4)
		if charset == "" {
			charset = s.Encoding
		}
		ret := Download(downloadnode, s.Downloader, url, "get", util.GetTable(head), charset, false, ishttps, "", s.Timeout)
		S.Push(lua.LString(ret))
		s.Test_luareqcount++
		return 1
	}))
	s.L.SetGlobal("findContentText", s.L.NewFunction(func(S *lua.LState) int {
		gpath := S.ToString(-2)
		content := S.ToString(-1)
		ret := util.FindContentText(gpath, content)
		S.Push(ret)
		return 1
	}))
	//高级下载download(url,method,param,head,cookie)
	s.L.SetGlobal("downloadAdv", s.L.NewFunction(func(S *lua.LState) int {
		cookie := S.ToString(-1)
		head := S.ToTable(-2)
		param := S.ToTable(-3)
		method := S.ToString(-4)
		url := S.ToString(-5)
		ishttps := S.ToBool(-6)
		charset := S.ToString(-7)
		if charset == "" {
			charset = s.Encoding
		}
		var mycookie []*http.Cookie
		json.Unmarshal([]byte(cookie), &mycookie)
		var ret string
		var retcookie []*http.Cookie
		if param == nil {
			ptext := map[string]interface{}{"text": S.ToString(-3)}
			ret, retcookie = DownloadAdv(downloadnode, s.Downloader, url, method, ptext, util.GetTable(head), mycookie, charset, false, ishttps, "", s.Timeout)
		} else {
			ret, retcookie = DownloadAdv(downloadnode, s.Downloader, url, method, util.GetTable(param), util.GetTable(head), mycookie, charset, false, ishttps, "", s.Timeout)
		}

		S.Push(lua.LString(ret))
		scookie, _ := json.Marshal(retcookie)
		S.Push(lua.LString(scookie))
		s.Test_luareqcount++
		return 2
	}))

	s.L.SetGlobal("findOneText", s.L.NewFunction(func(S *lua.LState) int {
		nodetype := S.ToString(-3)
		gpath := S.ToString(-2)
		content := S.ToString(-1)
		ret := util.FindOneText(gpath, content, nodetype)
		S.Push(ret)
		return 1
	}))
	s.L.SetGlobal("findOneHtml", s.L.NewFunction(func(S *lua.LState) int {
		nodetype := S.ToString(-3)
		gpath := S.ToString(-2)
		content := S.ToString(-1)
		ret := util.FindOneHtml(gpath, content, nodetype)
		S.Push(ret)
		return 1
	}))
	s.L.SetGlobal("findListText", s.L.NewFunction(func(S *lua.LState) int {
		gpath := S.ToString(-2)
		content := S.ToString(-1)
		ret := s.L.NewTable()
		util.FindListText(gpath, content, ret)
		S.Push(ret)
		return 1
	}))
	s.L.SetGlobal("findListHtml", s.L.NewFunction(func(S *lua.LState) int {
		gpath := S.ToString(-2)
		content := S.ToString(-1)
		ret := s.L.NewTable()
		util.FindListHtml(gpath, content, ret)
		S.Push(ret)
		return 1
	}))
	s.L.SetGlobal("findMap", s.L.NewFunction(func(S *lua.LState) int {
		qmap := S.ToTable(-2)
		content := S.ToString(-1)
		ret := s.L.NewTable()
		util.FindMap(qmap, content, ret)
		S.Push(ret)
		return 1
	}))
	//调用jsvm
	s.L.SetGlobal("jsvm", s.L.NewFunction(func(S *lua.LState) int {
		js := S.ToString(-1)
		ret := s.L.NewTable()
		if js == "" {
			ret.RawSet(lua.LString("val"), lua.LString(""))
			ret.RawSet(lua.LString("err"), lua.LString("js is null"))
		} else {
			rep := util.JsVmPost(util.Config.JsVmUrl, js)
			ret.RawSet(lua.LString("val"), lua.LString(qu.ObjToString(rep["val"])))
			ret.RawSet(lua.LString("err"), lua.LString(qu.ObjToString(rep["err"])))
		}
		S.Push(ret)
		return 1
	}))
	//指定下载器
	s.L.SetGlobal("changeDownloader", s.L.NewFunction(func(S *lua.LState) int {
		s.Downloader = GetOneDownloader()
		S.Push(lua.LString(s.Downloader))
		return 1
	}))
	//手工延时
	s.L.SetGlobal("timeSleep", s.L.NewFunction(func(S *lua.LState) int {
		time.Sleep(1 * time.Second)
		return 0
	}))
	//编码解码
	s.L.SetGlobal("transCode", s.L.NewFunction(func(S *lua.LState) int {
		codeType := strings.ToLower(S.ToString(-2))
		str := S.CheckString(-1)
		switch codeType {
		case "unicode":
			str = strings.Replace(str, "%u", "\\u", -1)
			str = transUnic(str)
		case "urlencode_gbk":
			data, _ := ioutil.ReadAll(transform.NewReader(bytes.NewReader([]byte(str)), simplifiedchinese.GBK.NewEncoder()))
			l, _ := url.Parse("http://a.com/?" + string(data))
			tmpstr := l.Query().Encode()
			if len(tmpstr) > 1 {
				str = tmpstr[0 : len(tmpstr)-1]
			} else {
				str = ""
			}
		case "urlencode_utf8":
			l, _ := url.Parse("http://a.com/?" + str)
			tmpstr := l.Query().Encode()
			if len(tmpstr) > 1 {
				str = tmpstr[0 : len(tmpstr)-1]
			} else {
				str = ""
			}
		case "urldecode_utf8":
			str, _ = url.QueryUnescape(str)
		case "decode64":
			str = util.DecodeB64(str)
		case "encodemd5":
			str = qu.GetMd5String(str)
		case "htmldecode": //html实体码
			//txt := `<div align="left" style="margin-left: 0pt;"><span style='font-family:; font-size:13px; color:#000000'>&#22826;&#38451;&#23707;&#29305;&#21220;&#28040;&#38450;&#31449;&#12289;&#26494;&#28006;&#29305;&#21220;&#28040;&#38450;&#31449;&#24314;&#35774;&#39033;&#30446;&#35774;&#35745;&#20013;&#26631;&#20844;&#31034;</span></div>`
			str = S.ToString(-1)
			reg, _ := regexp.Compile("&#\\d+;")
			str = reg.ReplaceAllStringFunc(str, func(src string) string {
				v, _ := strconv.Atoi(src[2 : len(src)-1])
				return string(rune(v))
			})
		}
		S.Push(lua.LString(str))
		return 1
	}))
	//保存错误日志
	s.L.SetGlobal("saveErrLog", s.L.NewFunction(func(S *lua.LState) int {

		return 0
	}))
	//添加改版日志
	s.L.SetGlobal("saveRevisionLog", s.L.NewFunction(func(S *lua.LState) int {

		return 0
	}))
	//如果服务端返回的html是gzip压缩过格式的 这里需要转一下
	s.L.SetGlobal("unGzip", s.L.NewFunction(func(S *lua.LState) int {
		html := S.ToString(-1)
		bs := []byte(html)
		gzipreader, _ := gzip.NewReader(bytes.NewReader(bs))
		bs, _ = ioutil.ReadAll(gzipreader)
		S.Push(lua.LString(bs))
		return 1
	}))

	s.L.SetGlobal("titleRepeatJudgement", s.L.NewFunction(func(S *lua.LState) int {
		bResult := false
		S.Push(lua.LBool(bResult))
		return 1
	}))

	//解析附件中的word、pdf
	s.L.SetGlobal("officeAnalysis", s.L.NewFunction(func(S *lua.LState) int {

		ext := map[string]byte{"pdf": byte(0), "doc": byte(1), "docx": byte(2)}

		str := S.ToString(-2)
		extension := S.ToString(-1)

		bs, _ := base64.StdEncoding.DecodeString(str)
		bs = append([]byte{ext[extension]}, bs...)
		msgid := mu.UUID(8)
		Msclient.Call("", msgid, mu.SERVICE_OFFICE_ANALYSIS, mu.SENDTO_TYPE_ALL_RECIVER, bs, 60)
		return 1
	}))
	//下载附件download(url,method,param,head,cookie,fileName)
	s.L.SetGlobal("downloadFile", s.L.NewFunction(func(S *lua.LState) int {
		cookie := S.ToString(-1)
		head := S.ToTable(-2)
		param := S.ToTable(-3)
		method := S.ToString(-4)
		url := S.ToString(-5)
		fileName := S.ToString(-6)
		ishttps := strings.Contains(url, "https")
		var mycookie []*http.Cookie
		if cookie != "{}" {
			json.Unmarshal([]byte(cookie), &mycookie)
		} else {
			mycookie = make([]*http.Cookie, 0)
		}
		fileName = strings.TrimSpace(fileName)
		url = strings.TrimSpace(url)
		ret := DownloadFile(s.Downloader, url, method, util.GetTable(param), util.GetTable(head), mycookie, s.Encoding, false, ishttps, "", s.Timeout)
		name, size, ftype, fid := "", "", "", ""
		qu.Debug(GarbledCodeReg.FindAllString(string(ret), -1), len(ret))
		if ret == nil || len(ret) < 1024*5 {
			qu.Debug("下载文件出错！")
		} else {
			ftype = qu.GetFileType(ret)
			if (ftype == "docx" || ftype == "doc") && len(GarbledCodeReg.FindAllString(string(ret), -1)) > 10 {
				url, name, size, ftype, fid = "附件中含有乱码", "附件中含有乱码", "", "", ""
			} else {
				url, name, size, ftype, fid = util.UploadFile(s.SCode, fileName, url, ret)
			}
		}
		if strings.TrimSpace(ftype) == "" {
			if len(path.Ext(name)) > 0 {
				ftype = path.Ext(name)[1:]
			}
		}
		S.Push(lua.LString(url))
		S.Push(lua.LString(name))
		S.Push(lua.LString(size))
		S.Push(lua.LString(ftype))
		S.Push(lua.LString(fid))
		return 5

	}))
	//支持正则
	s.L.SetGlobal("regexp", s.L.NewFunction(func(S *lua.LState) int {
		index := int(S.ToNumber(-1))
		regstr := S.ToString(-2)
		text := S.ToString(-3)
		reg := regexp.MustCompile(regstr)
		reps := reg.FindAllStringSubmatchIndex(text, -1)
		ret := s.L.NewTable()
		number := 0
		for _, v := range reps {
			number++
			ret.Insert(number, lua.LString(text[v[index]:v[index+1]]))
		}
		S.Push(ret)
		return 1
	}))
	//支持替换
	s.L.SetGlobal("replace", s.L.NewFunction(func(S *lua.LState) int {
		text := S.ToString(-3)
		old := S.ToString(-2)
		repl := S.ToString(-1)
		text = strings.Replace(text, old, repl, -1)
		S.Push(lua.LString(text))
		return 1
	}))
	//标题的关键词、排除词过滤
	s.L.SetGlobal("pagefilterword", s.L.NewFunction(func(S *lua.LState) int {
		keyWordReg := regexp.MustCompile(util.Config.Word["keyword"])
		notKeyWordReg := regexp.MustCompile(util.Config.Word["notkeyword"])
		data := S.ToTable(-1)
		dataMap := util.TableToMap(data)
		ret := s.L.NewTable()
		num := 1
		for _, v := range dataMap {
			tmp := v.(map[string]interface{})
			isOk := false
			if title := qu.ObjToString(tmp["title"]); title != "" {
				if keyWordReg.MatchString(title) && !notKeyWordReg.MatchString(title) {
					isOk = true
				}
			}
			if isOk {
				ret.Insert(num, util.MapToLuaTable(S, tmp))
				num++
			}
		}
		S.Push(ret)
		return 1
	}))
	//标题的关键词、排除词过滤
	s.L.SetGlobal("detailfilterword", s.L.NewFunction(func(S *lua.LState) int {
		keyWordReg := regexp.MustCompile(util.Config.Word["keyword"])
		notKeyWordReg := regexp.MustCompile(util.Config.Word["notkeyword"])
		data := S.ToTable(-1)
		dataMap := util.TableToMap(data)
		if title := qu.ObjToString(dataMap["title"]); title != "" {
			if keyWordReg.MatchString(title) && !notKeyWordReg.MatchString(title) {
				S.Push(lua.LBool(true))
				return 1
			} else {
				qu.Debug(s.SCode, dataMap["href"], "	title error")
			}
		} else {
			qu.Debug(s.SCode, dataMap["href"], "	title error")
		}
		S.Push(lua.LBool(false))
		return 1
	}))
	//detail过滤
	s.L.SetGlobal("filterdetail", s.L.NewFunction(func(S *lua.LState) int {
		/*
			1.长度判断 (特殊处理：详情请访问原网页！;详见原网页;见原网页;无;无相关内容;无正文内容)
			2.是否含汉字
		*/
		reg1 := regexp.MustCompile("(原网页|无|无相关内容|无正文内容|详见附件|见附件)")
		reg2 := regexp.MustCompile("[\u4e00-\u9fa5]")
		detail := S.ToString(-1)
		if reg1.MatchString(detail) {
			S.Push(lua.LBool(true))
			return 1
		}
		if len([]rune(detail)) < 50 || !reg2.MatchString(detail) {
			S.Push(lua.LBool(false))
			return 1
		}
		S.Push(lua.LBool(false))
		return 1
	}))
	//匹配汉字
	s.L.SetGlobal("matchan", s.L.NewFunction(func(S *lua.LState) int {
		reg1 := regexp.MustCompile("(见附件|详见附件)")
		reg2 := regexp.MustCompile("[\u4e00-\u9fa5]")
		detail := S.ToString(-1)
		detail = reg1.ReplaceAllString(detail, "")
		ok := reg2.MatchString(detail)
		S.Push(lua.LBool(ok))
		return 1
	}))
	//aes ecb模式加密
	s.L.SetGlobal("aesEncryptECB", s.L.NewFunction(func(S *lua.LState) int {
		origData := S.ToString(-2)
		key := S.ToString(-1)
		bytekey := []byte(key)
		byteorigData := []byte(origData)
		cipher, _ := aes.NewCipher(generateKey([]byte(bytekey)))
		length := (len(byteorigData) + aes.BlockSize) / aes.BlockSize
		plain := make([]byte, length*aes.BlockSize)
		copy(plain, byteorigData)
		pad := byte(len(plain) - len(byteorigData))
		for i := len(byteorigData); i < len(plain); i++ {
			plain[i] = pad
		}
		encrypted := make([]byte, len(plain))
		// 分组分块加密
		for bs, be := 0, cipher.BlockSize(); bs <= len(byteorigData); bs, be = bs+cipher.BlockSize(), be+cipher.BlockSize() {
			cipher.Encrypt(encrypted[bs:be], plain[bs:be])
		}
		result := base64.StdEncoding.EncodeToString(encrypted)
		S.Push(lua.LString(result))
		return 1
	}))
}

//
func (s *Script) Reload() {
	s.L.Close()
	s.LoadScript("", s.ScriptFile)
}

//unicode转码
func transUnic(str string) string {
	buf := bytes.NewBuffer(nil)
	i, j := 0, len(str)
	for i < j {
		x := i + 6
		if x > j {
			buf.WriteString(str[i:])
			break
		}
		if str[i] == '\\' && str[i+1] == 'u' {
			hex := str[i+2 : x]
			r, err := strconv.ParseUint(hex, 16, 64)
			if err == nil {
				buf.WriteRune(rune(r))
			} else {
				buf.WriteString(str[i:x])
			}
			i = x
		} else {
			buf.WriteByte(str[i])
			i++
		}
	}
	return buf.String()
}

//取得变量
func (s *Script) GetVar(key string) string {
	return s.L.GetGlobal(key).String()
}

//
func (s *Script) GetIntVar(key string) int {
	lv := s.L.GetGlobal(key)
	if v, ok := lv.(lua.LNumber); ok {
		return int(v)
	}
	return -1
}

//
func (s *Script) GetBoolVar(key string) bool {
	lv := s.L.GetGlobal(key)
	if v, ok := lv.(lua.LBool); ok {
		return bool(v)
	}
	return false
}

func generateKey(key []byte) (genKey []byte) {
	genKey = make([]byte, 16)
	copy(genKey, key)
	for i := 16; i < len(key); {
		for j := 0; j < 16 && i < len(key); j, i = j+1, i+1 {
			genKey[j] ^= key[i]
		}
	}
	return genKey
}