|
| 1 | +/* |
| 2 | + script/ceawler.js |
| 3 | +
|
| 4 | + resources/company_ids.jsにあるcompanyIDのリストを使ってスクレイピングをするツール |
| 5 | +*/ |
| 6 | + |
| 7 | +//Initialize instance |
| 8 | +var request = require("request"); |
| 9 | +var cheerio = require("cheerio"); |
| 10 | + |
| 11 | +var company_ids = require('../resources/company_ids'); |
| 12 | + |
| 13 | +var output = []; |
| 14 | + |
| 15 | +var i = 0; |
| 16 | + |
| 17 | +function outputFunc(){ |
| 18 | + if ( i >= company_ids.length ){ |
| 19 | + console.log( output ); |
| 20 | + } |
| 21 | +}; |
| 22 | + |
| 23 | +function crawl(name){ |
| 24 | + i++; //counter |
| 25 | + var requestUrl = "https://www.wantedly.com/companies/" + name + "/info"; |
| 26 | + request({url: requestUrl}, function(error, response, body) { |
| 27 | + if (!error && response.statusCode == 200) { |
| 28 | + $ = cheerio.load(body); |
| 29 | + |
| 30 | + // 募集人数 |
| 31 | + var job_count_raw = |
| 32 | + $("ul.wt-ui-tab > li:nth-child(3) > a").text(); |
| 33 | + var job_count_list = /\((\d+)\)/.exec(job_count_raw); |
| 34 | + var job_count; |
| 35 | + if(job_count_list) |
| 36 | + job_count = job_count_list.pop(); |
| 37 | + |
| 38 | + // 従業員数 |
| 39 | + var company_basic_info_html = |
| 40 | + $("section.company_basic_info > ul").text(); |
| 41 | + var employee_count_list = |
| 42 | + /(\d+)\n?人/.exec(company_basic_info_html); |
| 43 | + var employee_count; |
| 44 | + if (employee_count_list){ |
| 45 | + employee_count = employee_count_list[1]; |
| 46 | + } |
| 47 | + // 創立年月 |
| 48 | + var foundation_date_list = |
| 49 | + /(\d+年\d+月)/.exec(company_basic_info_html); |
| 50 | + var foundation_date; |
| 51 | + if (foundation_date_list){ |
| 52 | + foundation_date = foundation_date_list[1]; |
| 53 | + } |
| 54 | + |
| 55 | + // 会社情報 |
| 56 | + var company_name = $("h1.company-name > a").text(); |
| 57 | + var address = $("section.section_location.article > address").text(); |
| 58 | + var logo_url = $("div.profile-photo > a > span > img").attr("src"); |
| 59 | + var company_url = $("p.company-url > a").text(); |
| 60 | + |
| 61 | + // 私たちについて |
| 62 | + var about_me = $("section.section_origin > p").text(); |
| 63 | + |
| 64 | + // 経度緯度 (あとでちゃんと取る) |
| 65 | + var lats_raw_list = /Gmaps\.map\.markers\s\=\s\[(.*)\];/.exec(body); |
| 66 | + var lats_raw; |
| 67 | + if(lats_raw_list){ |
| 68 | + lats_raw = lats_raw_list.pop(); |
| 69 | + } |
| 70 | + |
| 71 | + var lats; |
| 72 | + if( lats_raw ){ |
| 73 | + lats = JSON.parse(lats_raw); |
| 74 | + } |
| 75 | + |
| 76 | + var company = { |
| 77 | + "id" : i, |
| 78 | + "name" : name, |
| 79 | + "company_name" : company_name, |
| 80 | + "address" : address, |
| 81 | + "logo_url" : logo_url, |
| 82 | + "company_url" : company_url, |
| 83 | + "job_count": job_count || null, |
| 84 | + "employee_count": employee_count || null, |
| 85 | + "foundation_date": foundation_date || null, |
| 86 | + "about_me": about_me || null, |
| 87 | + "lat": (lats)?lats.lat: null, |
| 88 | + "lng": (lats)?lats.lng: null |
| 89 | + }; |
| 90 | + output.push( company ); |
| 91 | + outputFunc(); |
| 92 | + } |
| 93 | + |
| 94 | + else { |
| 95 | + console.log("--- error occcured ---"); |
| 96 | + if (error && "code" in error) { |
| 97 | + console.log("Error No:" + error.errno); |
| 98 | + console.log("Error Code:" + error.code); |
| 99 | + console.log("Error Syscall:" + error.syscall); |
| 100 | + console.log("Status Code:" + response.statusCode); |
| 101 | + } |
| 102 | + } |
| 103 | + }); |
| 104 | +} |
| 105 | + |
| 106 | +function sleep(x) { |
| 107 | + return function(func) { |
| 108 | + setTimeout(function() { |
| 109 | + console.warn(i + company_ids[x]); |
| 110 | + crawl(company_ids[x]); |
| 111 | + func(); |
| 112 | + }, 500); |
| 113 | + } |
| 114 | +} |
| 115 | + |
| 116 | +function loop(x) { |
| 117 | + sleep(x)(function(){ |
| 118 | + if(x < company_ids.length -1){ |
| 119 | + loop(x+1); |
| 120 | + } |
| 121 | + }); |
| 122 | +} |
| 123 | + |
| 124 | +loop(0); |
0 commit comments