浏览代码

refactor: 修改被抓取页面地址的定位逻辑

modood 7 年之前
父节点
当前提交
adfc313f77
共有 1 个文件被更改,包括 25 次插入52 次删除
  1. 25 52
      lib/crawler.js

+ 25 - 52
lib/crawler.js

@@ -13,37 +13,24 @@ const BufferHelper = require('bufferhelper')
  * 村级(村委会居委会,Village)    v
  */
 
+const pReg = /<td><a href='(.*?).html'>(.*?)<br\/><\/a><\/td>/g
+const casReg = /<tr class='.*?'><td><a href=.*?>(.*?)<\/a><\/td><td><a href=.*?>(.*?)<\/a><\/td><\/tr>/g
+const vReg = /<td>(.*?)<\/td><td>.*?<\/td><td>(.*?)<\/td>/g
+
 const host = 'www.stats.gov.cn'
-const links = [
-  {
-    path: '/tjsj/tjbz/tjyqhdmhcxhfdm/2016/index.html',
-    regexp: /<td><a href='(.*?).html'>(.*?)<br\/><\/a><\/td>/g
-  },
-  {
-    path: '/tjsj/tjbz/tjyqhdmhcxhfdm/2016/#{pCode}.html',
-    regexp: /<tr class='.*?'><td><a href=.*?>(.*?)<\/a><\/td><td><a href=.*?>(.*?)<\/a><\/td><\/tr>/g
-  },
-  {
-    path: '/tjsj/tjbz/tjyqhdmhcxhfdm/2016/#{pCode}/#{cCode}.html',
-    regexp: /<tr class='.*?'><td><a href=.*?>(.*?)<\/a><\/td><td><a href=.*?>(.*?)<\/a><\/td><\/tr>/g
-  },
-  {
-    path: '/tjsj/tjbz/tjyqhdmhcxhfdm/2016/#{pCode}/#{cCodeSuffix}/#{aCode}.html',
-    regexp: /<tr class='.*?'><td><a href=.*?>(.*?)<\/a><\/td><td><a href=.*?>(.*?)<\/a><\/td><\/tr>/g
-  },
-  {
-    path: '/tjsj/tjbz/tjyqhdmhcxhfdm/2016/#{pCode}/#{cCodeSuffix}/#{aCodeSuffix}/#{sCode}.html',
-    regexp: /<td>(.*?)<\/td><td>.*?<\/td><td>(.*?)<\/td>/g
-  }
-]
+const path = '/tjsj/tjbz/tjyqhdmhcxhfdm/2016/#{route}.html'
 
 /**
  * 抓取数据
  * @author modood <https://github.com/modood>
  * @datetime 2018-01-31 19:23
  */
-exports.fetch = (host, path, regexp, codeLen) =>
-  new Promise((resolve, reject) => http.get({ host, path, timeout: 3000 }, res => {
+exports.fetch = (host, route, regexp, codeLen) =>
+  new Promise((resolve, reject) => http.get({
+    host,
+    path: path.replace('#{route}', route),
+    timeout: 3000
+  }, res => {
     const bufferHelper = new BufferHelper()
     const statusCode = res.statusCode
 
@@ -72,10 +59,10 @@ exports.fetch = (host, path, regexp, codeLen) =>
  */
 exports.fetchProvinces = async () => {
   try {
-    return await exports.fetch(host, links[0].path, links[0].regexp, 2)
+    return await exports.fetch(host, 'index', pReg, 2)
   } catch (err) {
     console.log(`抓取省级数据失败(${err}),正在重试...`)
-    return this.fetchProvinces()
+    return exports.fetchProvinces()
   }
 }
 
@@ -85,13 +72,11 @@ exports.fetchProvinces = async () => {
  * @datetime 2018-01-31 19:51
  */
 exports.fetchCities = async (pCode) => {
-  const path = links[1].path.replace('#{pCode}', pCode)
-
   try {
-    return await exports.fetch(host, path, links[1].regexp, 4)
+    return await exports.fetch(host, pCode, casReg, 4)
   } catch (err) {
     console.log(`抓取省级(${pCode})的地级数据失败(${err}),正在重试...`)
-    return this.fetchCities(pCode)
+    return exports.fetchCities(pCode)
   }
 }
 
@@ -103,16 +88,12 @@ exports.fetchCities = async (pCode) => {
 exports.fetchAreas = async (cCode) => {
   cCode = cCode.toString()
   const pCode = cCode.substr(0, 2)
-  const path = links[2].path
-    .replace('#{pCode}', pCode)
-    .replace('#{cCode}', cCode)
 
   try {
-    return await exports.fetch(host, path, links[2].regexp, 6)
+    return await exports.fetch(host, `${pCode}/${cCode}`, casReg, 6)
   } catch (err) {
     console.log(`抓取地级(${cCode})的县级数据失败(${err}),正在重试...`)
-    console.log(path)
-    return this.fetchAreas(cCode)
+    return exports.fetchAreas(cCode)
   }
 }
 
@@ -121,21 +102,17 @@ exports.fetchAreas = async (cCode) => {
  * @author modood <https://github.com/modood>
  * @datetime 2018-01-31 20:08
  */
-exports.fetchStreets = async (aCode) => {
+exports.fetchStreets = async (aCode, route) => {
   aCode = aCode.toString()
   const pCode = aCode.substr(0, 2)
   const cCodeSuffix = aCode.substr(2, 2)
-  const path = links[3].path
-    .replace('#{pCode}', pCode)
-    .replace('#{cCodeSuffix}', cCodeSuffix)
-    .replace('#{aCode}', aCode)
+  const _route = route || `${pCode}/${cCodeSuffix}/${aCode}`
 
   try {
-    return await exports.fetch(host, path, links[3].regexp, 9)
+    return await exports.fetch(host, _route, casReg, 9)
   } catch (err) {
     console.log(`抓取县级(${aCode})的乡级数据失败(${err}),正在重试...`)
-    console.log(path)
-    return this.fetchStreets(aCode)
+    return exports.fetchStreets(aCode, route)
   }
 }
 
@@ -144,21 +121,17 @@ exports.fetchStreets = async (aCode) => {
  * @author modood <https://github.com/modood>
  * @datetime 2018-01-31 20:19
  */
-exports.fetchVillages = async (sCode) => {
+exports.fetchVillages = async (sCode, route) => {
   sCode = sCode.toString()
   const pCode = sCode.substr(0, 2)
   const cCodeSuffix = sCode.substr(2, 2)
   const aCodeSuffix = sCode.substr(4, 2)
-  const path = links[4].path
-    .replace('#{pCode}', pCode)
-    .replace('#{cCodeSuffix}', cCodeSuffix)
-    .replace('#{aCodeSuffix}', aCodeSuffix)
-    .replace('#{sCode}', sCode)
+  const _route = route || `${pCode}/${cCodeSuffix}/${aCodeSuffix}/${sCode}`
 
   try {
-    return await exports.fetch(host, path, links[4].regexp, 12)
+    return await exports.fetch(host, _route, vReg, 12)
   } catch (err) {
     console.log(`抓取乡级(${sCode})的村级数据失败(${err}),正在重试...`)
-    return this.fetchVillages(sCode)
+    return exports.fetchVillages(sCode, route)
   }
 }