235 lines
6.8 KiB
JavaScript
235 lines
6.8 KiB
JavaScript
import axios from "axios";
|
|
import fs from "fs";
|
|
import path from "path";
|
|
import { timestampToDate, loopCall, keywordsInclude } from "./utils.js";
|
|
import config from "./config.js";
|
|
import { SQLiteMessageQueue } from "./sqlite.js";
|
|
|
|
class GreatWall {
|
|
constructor() {
|
|
this.jsonMap = [
|
|
{
|
|
name: "长城公开寻源",
|
|
info: [],
|
|
options: {
|
|
name: "长城公开寻源",
|
|
url: "https://srm.gwm.cn/cloud-srm/api-sou/sou-firstPage/souReqlistPage",
|
|
},
|
|
},
|
|
{
|
|
name: "长城招募公示大厅",
|
|
info: [],
|
|
options: {
|
|
name: "长城招募公示大厅",
|
|
url: "https://srm.gwm.cn/cloud-srm/api-sou/api-ql/Recruit/visitList",
|
|
data: {
|
|
type: "Recruit",
|
|
lang: "zh-cn",
|
|
query: { "*": {} },
|
|
payload: {
|
|
filter: {},
|
|
page: { sort: "lastUpdateDate desc", pageNum: 1, pageSize: 8 },
|
|
},
|
|
action: "visitList",
|
|
tree: true,
|
|
},
|
|
},
|
|
},
|
|
];
|
|
console.log("长城 爬虫启动...");
|
|
this.queue = new SQLiteMessageQueue();
|
|
this.start();
|
|
}
|
|
|
|
async start() {
|
|
try {
|
|
await this.init();
|
|
} catch (err) {
|
|
console.error("启动失败:", err);
|
|
}
|
|
}
|
|
async init() {
|
|
for (let item of this.jsonMap) {
|
|
let announcements = this.queue.getAnnouncementsBySpider(item.name);
|
|
if (announcements.length > 0) {
|
|
this.loopFetchIncrement(item);
|
|
} else {
|
|
this.loopFetchFull(item);
|
|
}
|
|
}
|
|
}
|
|
// 全量爬取
|
|
loopFetchFull(props) {
|
|
try {
|
|
loopCall(this.getInfo.bind(this), {
|
|
time: config.fullFetchTime,
|
|
pagenumber: 1,
|
|
additional: props.options,
|
|
stopWhen: (pagenumber, result) => {
|
|
return (
|
|
pagenumber >= result.pages || pagenumber >= config.pageNumberLimit
|
|
);
|
|
},
|
|
readyForNext: (pagenumber, result) => {
|
|
props.info.push(...result.info);
|
|
return pagenumber + 1;
|
|
},
|
|
complete: (result) => {
|
|
props.info.push(...result.info);
|
|
console.log(`爬取完成,共获取 ${props.info.length} 条有效数据`);
|
|
try {
|
|
if (props.info.length > 0) {
|
|
this.queue.saveAnnouncements(props.name, props.info);
|
|
// this.writeFile(props);
|
|
this.queue.addMessage(props.name, props.info);
|
|
}
|
|
} catch (error) {
|
|
console.error("数据库操作失败:", error);
|
|
}
|
|
this.loopFetchIncrement(props);
|
|
},
|
|
});
|
|
} catch (error) {
|
|
console.error(`${props.options.name}全量爬取失败:`, error);
|
|
}
|
|
}
|
|
loopFetchIncrement(props) {
|
|
try {
|
|
loopCall(this.getInfo.bind(this), {
|
|
time: config.incrementFetchTime, // 5分钟间隔
|
|
pagenumber: 1,
|
|
additional: props.options,
|
|
readyForNext: (pagenumber, result) => {
|
|
try {
|
|
let newInfo = this.queue.filterNewAnnouncements(
|
|
props.name,
|
|
result.info
|
|
);
|
|
// 存在新数据
|
|
if (newInfo.length > 0) {
|
|
console.log(`发现 ${newInfo.length} 条新数据`);
|
|
// props.info.push(...newInfo);
|
|
this.queue.saveAnnouncements(props.name, newInfo);
|
|
// this.writeFile(props);
|
|
this.queue.addMessage(props.name, newInfo);
|
|
// 全是新数据,继续下一页
|
|
if (newInfo.length === result.info.length) {
|
|
return pagenumber + 1;
|
|
} else {
|
|
// 有部分重复数据,重新从第一页开始
|
|
return 1;
|
|
}
|
|
} else {
|
|
console.log("没有发现新数据,继续监控...");
|
|
return 1; // 重新从第一页开始
|
|
}
|
|
} catch (error) {
|
|
console.error("数据库操作失败:", error);
|
|
}
|
|
},
|
|
});
|
|
} catch (error) {
|
|
console.error(`${props.options.name}增量爬取失败:`, error);
|
|
}
|
|
}
|
|
async getInfo(pagenumber = 1, config) {
|
|
let info = [];
|
|
console.log(`${config.name}--获取第 ${pagenumber} 页数据...`);
|
|
let result = await this.getList(pagenumber, config);
|
|
if (result[0]) {
|
|
// 出错, 记录错误日志
|
|
console.error("获取页面数据失败:", result[0]);
|
|
return { pages: 0, info: [] };
|
|
} else {
|
|
if (config.data) {
|
|
// 招募公示大厅
|
|
let arr = result[1].data.records;
|
|
let pages = result[1].data.pageCount;
|
|
for (let i = 0; i < arr.length; i++) {
|
|
let item = arr[i];
|
|
let endTime, publishTime;
|
|
endTime = item.deadlineTime;
|
|
publishTime = item.publishTime;
|
|
// 命中关键词
|
|
if (keywordsInclude(item.title)) {
|
|
info.push({
|
|
id: item.recruitId,
|
|
name: item.title,
|
|
publishTime: publishTime,
|
|
endTime: endTime,
|
|
urls: `https://srm.gwm.cn/#/portalBidding/vendorBiddingDetail?id=${item.recruitId}`,
|
|
});
|
|
}
|
|
}
|
|
return { pages, info };
|
|
} else {
|
|
// 公开寻源
|
|
let arr = result[1].data.list;
|
|
let pages = result[1].data.pages;
|
|
|
|
for (let i = 0; i < arr.length; i++) {
|
|
let item = arr[i];
|
|
let endTime, publishTime;
|
|
endTime = item.publicEndTime;
|
|
publishTime = item.releaseDate;
|
|
// 命中关键词
|
|
if (keywordsInclude(item.projectName)) {
|
|
info.push({
|
|
id: item.reqHeadId,
|
|
name: item.projectName,
|
|
publishTime: publishTime,
|
|
endTime: endTime,
|
|
urls: `https://srm.gwm.cn/#/portal?id=${item.reqHeadId}`,
|
|
});
|
|
}
|
|
}
|
|
return { pages, info };
|
|
}
|
|
}
|
|
}
|
|
// 分页获取数据
|
|
getList(pagenumber, config) {
|
|
let data = {};
|
|
if (config.data) {
|
|
data = config.data;
|
|
data.payload.page.pageNum = pagenumber;
|
|
} else {
|
|
data = { pageNum: pagenumber, pageSize: 8 };
|
|
}
|
|
return axios({
|
|
url: config.url,
|
|
data: data,
|
|
method: "post",
|
|
})
|
|
.then((res) => {
|
|
let result = res.data;
|
|
if (result.code == "0") {
|
|
return [null, result];
|
|
} else {
|
|
return ["err", null];
|
|
}
|
|
})
|
|
.catch((err) => {
|
|
return [err, null];
|
|
});
|
|
}
|
|
|
|
// writeFile(props) {
|
|
// fs.writeFileSync(props.filepath, JSON.stringify(props.info), "utf-8");
|
|
// }
|
|
|
|
// extractDeadlineTime(html) {
|
|
// // 匹配"预告报名截止时间:"后面的时间格式
|
|
// const regex = /预告报名截止时间:(\d{4}-\d{2}-\d{2}\s+\d{2}:\d{2}:\d{2})/;
|
|
// const match = html.match(regex);
|
|
|
|
// if (match) {
|
|
// return match[1];
|
|
// }
|
|
|
|
// return null;
|
|
// }
|
|
}
|
|
|
|
new GreatWall();
|