239 lines
9.0 KiB
JavaScript
239 lines
9.0 KiB
JavaScript
const fetch = require('node-fetch');
|
|
const cheerio = require('cheerio');
|
|
const { allRERequest, findPointInsideBoundingBox } = require('../db/dbHelper');
|
|
const { getRealEstateTypeEnum } = require('../enums');
|
|
const { getRegion, getMunicipality } = require('../codes')
|
|
const Promise = require("bluebird");
|
|
|
|
module.exports = class OlxCrawler {
|
|
//TODO figure best way to handle paging
|
|
constructor(fromPage = 0, toPage = 10, maxResults = 1000) {
|
|
this.fromPage = fromPage;
|
|
this.toPage = toPage;
|
|
this.maxResults = maxResults;
|
|
}
|
|
|
|
async indexSingle(url, email, uuid) {
|
|
try {
|
|
const res = await fetch(url);
|
|
const body = await res.text();
|
|
const $ = cheerio.load(body);
|
|
|
|
//TODO figure out what to do with username
|
|
const username = $('#lg > div.desno2.profil > div:nth-child(2) > div.vrsta1.vrsta_desno > a > div.username > span').text();
|
|
|
|
// if (IGNORED_USERNAMES.includes((username || '').toLowerCase())) {
|
|
// return null;
|
|
// }
|
|
|
|
//TODO remove properties that are not needed, and add some if they are missing
|
|
const title = $('#naslovartikla').text().trim();
|
|
const realEstateType = $('#artikal_glavni_div > div.artikal_lijevo > div:nth-child(3) > div > span:nth-child(3) > a > span').text();
|
|
|
|
const price = $('#pc > p:nth-child(2)').text();
|
|
const size = $('#dodatnapolja1 > div:nth-child(1) > div.df2').text();
|
|
const rooms = $('#dodatnapolja1 > div:nth-child(2) > div.df2').text();
|
|
const address = $('#dodatnapolja1 > div:nth-child(5) > div.df2').text();
|
|
const gardenSize = $('#dodatnapolja1 > div:nth-child(6) > div.df2').text();
|
|
const location = $('#artikal_glavni_div > div.artikal_lijevo > div.op.pop.mobile-lokacija').attr('data-content');
|
|
|
|
const adType = $('#artikal_glavni_div > div.artikal_lijevo > div:nth-child(15) > div:nth-child(2) > div.df2').text();
|
|
const time = $('time').attr('datetime');
|
|
const olxId = $('#artikal_glavni_div > div.artikal_lijevo > div:nth-child(15) > div:nth-child(4) > div.df2').text();
|
|
|
|
const descriptions = $('.artikal_detaljniopis_tekst');
|
|
// const floor = $('#dodatnapolja1').find(':contains(Sprat)').last().nextAll().text();
|
|
const latLngRe = /LatLng\(([0-9]+\.[0-9]+)\,\s+([0-9]+\.[0-9]+)\)/g;
|
|
const imgRe = /href":("[^"]*")/g;
|
|
const matches = latLngRe.exec(body);
|
|
let lng = '',
|
|
lat = '';
|
|
|
|
|
|
const parseRooms = (rooms) => parseInt([...rooms].filter(c => !isNaN(c)).filter(c => c.trim()).join())
|
|
const parsePrice = (price) => parseFloat(price.replace(".", ""))
|
|
|
|
|
|
// TODO we dont save images ??
|
|
|
|
// const images = [];
|
|
// const imgMatches = body.match(imgRe);
|
|
|
|
// for (let i = 0; imgMatches && i < imgMatches.length; i++) {
|
|
// let img = imgMatches[i].replace("href\":", "")
|
|
// img = img.replace("\"", "");
|
|
// img = img.replace("\"", "");
|
|
// images.push(img);
|
|
// }
|
|
|
|
// const uploadPromises = images.map(img => {
|
|
// const imgFixed = eval(`'${img}'`);
|
|
// return cloudinary.uploader.upload(eval(`'${img}'`));
|
|
// });
|
|
|
|
// const uploadResults = await Promise.all(uploadPromises);
|
|
// const cloudinaryImages = uploadResults.map(ur => ur.url);
|
|
|
|
if (matches && matches.length >= 3) {
|
|
lat = matches[1];
|
|
lng = matches[2];
|
|
}
|
|
|
|
const parsedPrice = parsePrice(price);
|
|
|
|
console.log(location);
|
|
const locationArray = location.split(",");
|
|
const region = locationArray[0];
|
|
const municipality = locationArray[1];
|
|
|
|
const data = {
|
|
realEstateType: this.getCategoryId(realEstateType),
|
|
email: email,
|
|
uuid: uuid,
|
|
olxId: olxId,
|
|
// category: category,
|
|
url,
|
|
title,
|
|
price: isNaN(parsedPrice) ? 0 : parsedPrice,
|
|
size: parseFloat(size),
|
|
gardenSize: isNaN(parseFloat(gardenSize)) ? 0 : parseFloat(gardenSize),
|
|
address,
|
|
region,
|
|
municipality,
|
|
// adType: AD_TYPE_SALE,
|
|
time,
|
|
shortDescription: descriptions.first().text(),
|
|
longDescription: descriptions.last().text(),
|
|
lat,
|
|
lng,
|
|
loc: [parseFloat(lat), parseFloat(lng)],
|
|
// images: cloudinaryImages
|
|
};
|
|
|
|
return data;
|
|
} catch (e) {
|
|
console.error('Exception caught: ' + e.message);
|
|
}
|
|
|
|
return null;
|
|
}
|
|
|
|
async indexPage(olxUrl, maxResults = 1000) {
|
|
try {
|
|
//TODO fix paging
|
|
// console.log('Starting to index page: ' + pageNr);
|
|
// const url = `http://www.olx.ba/pretraga?vrsta=samoprodaja&sort_order=desc&kategorija=23&sort_po=datum&kanton=9&stranica=${pageNr}`;
|
|
|
|
const res = await fetch(olxUrl.url);
|
|
const body = await res.text();
|
|
const $ = cheerio.load(body);
|
|
const hrefs = [];
|
|
const results = [];
|
|
|
|
$('#rezultatipretrage').find('.listitem').each((i, elem) => {
|
|
const href = $(elem).find('a').first().attr('href');
|
|
hrefs.push(href);
|
|
});
|
|
|
|
let actualNoOfResults = (hrefs.length <= maxResults) ? hrefs.length : maxResults;
|
|
|
|
for (let i = 0; i < hrefs.length; i++) {
|
|
console.log(`indexing: ${hrefs[i]}`);
|
|
|
|
const singleData = await this.indexSingle(hrefs[i], olxUrl.email, olxUrl.uuid);
|
|
|
|
if (singleData) {
|
|
results.push(singleData);
|
|
}
|
|
}
|
|
|
|
return results;
|
|
} catch (e) {
|
|
console.error('Exception caught:' + e);
|
|
}
|
|
}
|
|
|
|
getCategoryId(category) {
|
|
|
|
switch (category) {
|
|
case 'Stanovi':
|
|
return 'stan';
|
|
|
|
case 'Vikendice':
|
|
return 'vikendica'
|
|
|
|
case 'Kuće':
|
|
return 'kuca';
|
|
|
|
default:
|
|
return '';
|
|
}
|
|
}
|
|
|
|
async indexPages(urls, start, end, maxResults = 1000) {
|
|
//TODO fix paging
|
|
// let results = {};
|
|
// for (let i = start; i <= end; i++) {
|
|
// let result = await this.indexPage(i, maxResults);
|
|
// Object.assign(results, result)
|
|
// }
|
|
// return results;
|
|
|
|
let results = [];
|
|
for (let url of urls) {
|
|
let result = await this.indexPage(url, maxResults);
|
|
results.push(result);
|
|
}
|
|
return results;
|
|
}
|
|
|
|
async crawl() {
|
|
console.log("OLX CRAWLER: start crawl");
|
|
|
|
const filteredResults = [];
|
|
const realestateRequests = await allRERequest();
|
|
console.log("OLX CRAWLER: found " + realestateRequests.length + "subscribed RealEstateRequests");
|
|
const urls = this.createRequestUrls(realestateRequests);
|
|
let results = await this.indexPages(urls, this.fromPage, this.toPage, this.maxResults);
|
|
|
|
for (const result of results) {
|
|
for (const finalResult of result) {
|
|
if (finalResult.lat !== undefined && finalResult.lat !== null && finalResult.lat !== "") {
|
|
const pointInsideBoundingBox = await findPointInsideBoundingBox([finalResult.lng, finalResult.lat], finalResult.email);
|
|
|
|
if (pointInsideBoundingBox[0].length !== 0) {
|
|
filteredResults.push(finalResult);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
console.log("OLX CRAWLER: number of olx crawler results, after geo location filtering: " + filteredResults.length);
|
|
return filteredResults;
|
|
}
|
|
|
|
createRequestUrls(realestateRequests) {
|
|
const urls = []
|
|
|
|
for (const request of realestateRequests) {
|
|
const realsestateType = "kategorija=" + getRealEstateTypeEnum(request.realEstateType).olxCategory;
|
|
const region = "kanton=" + getRegion(request.region).olxid;
|
|
const municipality = "grad%5B%5D=" + getMunicipality(request.region, request.municipality).olxid;
|
|
const sizeMin = "kvadrata_min=" + request.sizeMin;
|
|
const sizeMax = "kvadrata_max=" + request.sizeMax;
|
|
const priceMin = "od=" + request.priceMin;
|
|
const priceMax = "do=" + request.priceMax;
|
|
|
|
const olxUrl = {
|
|
url: `https://www.olx.ba/pretraga?${realsestateType}&id=2&stanje=0&vrstapregleda=tabela&sort_order=desc&${region}&${municipality}&${priceMin}&${priceMax}&vrsta=samoprodaja&${sizeMin}&${sizeMax}`,
|
|
email: request.email,
|
|
uuid: request.uniqueId
|
|
}
|
|
console.log(olxUrl.url);
|
|
urls.push(olxUrl);
|
|
}
|
|
|
|
return urls;
|
|
}
|
|
};
|
|
|