node.js 基于cheerio的爬虫工具，需要登录权限的爬虫工具，直接导出到Excel文件中

本文转载自查看原文 2019-04-09 14:57 430 爬虫/ cheerio爬虫/ cookie爬虫/ nodejs

公司有过一个需求,需要拿一个网页的的表格数据，数据量达到30w左右；为了提高工作效率。

结合自身经验和网上资料。写了一套符合自己需求的nodejs爬虫工具。也许也会适合你的。

先上代码。在做讲解

'use strict';
// 引入模块
const superagent = require('superagent');
const cheerio = require('cheerio');
const Excel = require('exceljs');


var baseUrl = '';
var Cookies = 'PHPSESSID=1c948cafb361cb5dce87122846e649cd'; //伪装的cookie
let pageDatas = [];
let count = 1;
let limit = 3;

for (count; count < limit; count++) { 
    baseUrl = `http://bxjd.henoo.com/policy/policyList?page=${count}`;
    loadPage(baseUrl); 
}

function loadPage(baseUrl) { 
     getPageLoad(baseUrl);
} 
async function getPageLoad(baseUrl) {
    try {
        let body = await superagent.get(baseUrl)
            .set("Cookie", Cookies)
        var $ = cheerio.load(body.text);
        var trList = $("#tableList").children("tr");   
            for (var i = 0; i < trList.length; i++) { 
                let item = {};
                var tdArr = trList.eq(i).find("td");
                var id = tdArr.eq(0).text();
                item.sortId = id;
                var detailUrl = `http://bxjd.henoo.com/policy/view?id=${id}`;
                item.policyId = tdArr.eq(1).text();
                item.policyProductName = tdArr.eq(2).text();
                item.policyName = tdArr.eq(3).text();
                item.policyMoney = tdArr.eq(4).text();
                let detailBody = await superagent.get(detailUrl)
                    .set("Cookie", Cookies);
                var $$ = cheerio.load(detailBody.text);
                var detailT = $$(".table-view");          
                
                //投保人证件号
                item.policyIdNum = detailT.find("tr").eq(11).find("td").eq(1).text();
                //投保人手机号
                item.policyPhone = detailT.find("tr").eq(10).find("td").eq(1).text();
                //被保人手机号
                item.bePoliciedPhone = detailT.find("tr").eq(16).find("td").eq(1).text();           
                //被保人姓名
                item.bePoliciedName = detailT.find("tr").eq(13).find("td").eq(1).text();
                console.log(item.bePoliciedName)
                //被保人证件号
                item.bePoliciedIdNum = detailT.find("tr").eq(17).find("td").eq(1).text();
                pageDatas = [...pageDatas,item];
            }
        if (pageDatas.length / 15 == (count - 1)) { 
            writeXLS(pageDatas)
        }
    } catch (error) {
        
    }
    
}

function writeXLS(pageDatas) {
    const workbook = new Excel.Workbook();
    const sheet = workbook.addWorksheet('My Sheet');

    const reColumns=[
        {header:'序号',key:'sortId'},
        {header:'投保单号',key:'policyId'},
        {header: '产品名称', key: 'policyProductName'},
        {header: '投保人姓名', key: 'policyName' },
        {header: '投保人手机号', key: 'policyPhone' },
        {header: '投保人证件号', key: 'policyIdNum'},
        {header: '被保人姓名', key: 'bePoliciedName' },
        {header: '被保人手机号', key: 'bePoliciedPhone' },
        {header: '被保人证件号', key: 'bePoliciedIdNum' },
        {header:'保费',key:'policyMoney'},
    ];
    sheet.columns = reColumns;
    for(let trData of pageDatas){
        sheet.addRow(trData);
    }
    const filename = './projects.xlsx';
    workbook.xlsx.writeFile(filename)
    .then(function() {
      console.log('ok');
    }).catch(function (error) { 
          console.error(error); 
    });
  }