const GetPageUrl = () => {
return new Array(50).fill('').map((_, i) => {
if (i === 0) {
return 'http://www.scjc.gov.cn/scjc/scdc/scdc.shtml'
}
return `http://www.scjc.gov.cn/scjc/scdc/scdc_${i + 1}.shtml`
})
}
const GetDocument = (url) => {
return new Promise((r) => {
let iframeEl = document.createElement('iframe')
iframeEl.src = url
iframeEl.sandbox = 'allow-same-origin'
iframeEl.onload = () => {
if (iframeEl.contentDocument) {
r({ document: iframeEl.contentDocument,iframeEl:iframeEl })
}else{
r(null)
}
}
document.body.appendChild(iframeEl)
})
}
const GettitleAndHref = (document) => {
let arrInfo = Array.from(document.querySelectorAll('.data-list > h4 a'))
let arr =arrInfo.map((info)=>({
title : info.innerText,
href : info.href
}))
let arrTime = Array.from(document.querySelectorAll('.data-list .right'))
arr = arr.map((info,index)=>({
...info,
PublishTime:arrTime[index].innerText
}))
return arr
}
const getContentAndTime = async(href) => {
let tarDocument = await GetDocument(href)
let tarDoc = tarDocument.document
let content = tarDoc.querySelector('.cont > .info').innerText.trim()
document.body.removeChild(tarDocument.iframeEl)
return content
}
const getAllInfo = async (arr) => {
let contentArr = await Promise.all(arr.map(({href})=>getContentAndTime(href)))
arr = arr.map((info,index)=>({
...info,
content:contentArr[index],
}))
return arr
}
const downloadToLocal = (data) => {
const aEle = document.createElement("a");
aEle.href = URL.createObjectURL(new Blob([JSON.stringify(data)]));
aEle.download = "ygz.cq.gov.cn.json";
document.body.appendChild(aEle);
aEle.click();
};
const GetAllData = async () => {
let allInfoArr=[]
let UrlArr = GetPageUrl()
for (let i = 0; i < UrlArr.length; i++) {
let PagePost = await GetDocument(UrlArr[i])
let titleAndHrefArr = GettitleAndHref(PagePost.document)
allInfoArr.push(...await getAllInfo(titleAndHrefArr))
document.body.removeChild(PagePost.iframeEl)
}
return allInfoArr
}
GetAllData().then(r=>{
console.log('抓取完毕')
downloadToLocal(r)
})
JS控制台爬虫
最新推荐文章于 2024-07-18 12:57:54 发布