371 lines
13 KiB
JavaScript
371 lines
13 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 indexPages(urls) {
|
|
const indexers = [];
|
|
|
|
urls.forEach(url => {
|
|
indexers.push(new Indexer(url));
|
|
});
|
|
|
|
return Promise.map(indexers, function (indexer) {
|
|
return indexer.indexWithPagination();
|
|
}).then(async (results) => {
|
|
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);
|
|
console.log("Final crawler results");
|
|
if (results[0]) {
|
|
console.log(results[0].length);
|
|
|
|
for (const finalResult of results[0]) {
|
|
|
|
if (null !== finalResult) {
|
|
if (finalResult.lat !== undefined && finalResult.lat !== null && finalResult.lat !== "") {
|
|
const pointInsideBoundingBox = await findPointInsideBoundingBox([finalResult.lng, finalResult.lat], finalResult.email);
|
|
|
|
|
|
if (pointInsideBoundingBox[0].length !== 0) {
|
|
finalResult.hasLocation = true
|
|
filteredResults.push(finalResult);
|
|
} else {
|
|
finalResult.hasLocation = false
|
|
filteredResults.push(finalResult);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
console.log("OLX CRAWLER: number of olx crawler results, after geo location filtering: " + filteredResults.length);
|
|
return filteredResults;
|
|
}
|
|
return []
|
|
}
|
|
|
|
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}&${sizeMax}&stranica=`,
|
|
email: request.email,
|
|
uuid: request.uniqueId
|
|
}
|
|
console.log(olxUrl.url);
|
|
urls.push(olxUrl);
|
|
}
|
|
|
|
return urls;
|
|
}
|
|
};
|
|
|
|
|
|
class Indexer {
|
|
|
|
/**
|
|
*
|
|
* @param {String|Array} olxUrl single or array of objects containing url email and uuid
|
|
* @param {Array} hrefResutls array contaning urls from crawler results
|
|
*/
|
|
|
|
constructor(olxUrl, hrefResutls) {
|
|
this.olxUrl = olxUrl;
|
|
this.hrefResutls = hrefResutls;
|
|
}
|
|
|
|
async indexWithPagination(pageNumber = 1) {
|
|
|
|
console.log("This is olxUrl:" + this.olxUrl.url);
|
|
const pageNr = this.olxUrl.url.match(/\d+$/);
|
|
const indexers = this.prepareIndexers(pageNumber ? [pageNumber] : pageNr);
|
|
|
|
try {
|
|
|
|
return Promise.map(indexers.indexers, function (indexer) {
|
|
return indexer.indexPage(pageNumber);
|
|
}).then(async (results) => {
|
|
let hasResults = false;
|
|
|
|
results.forEach(result => {
|
|
if (!hasResults) {
|
|
console.log("No results detected")
|
|
hasResults = result.hasResults
|
|
}
|
|
});
|
|
|
|
if (!hasResults) {
|
|
console.log("HAS NO MORE RESULTS, stop the paging, there are some results and they should contain only HREFS");
|
|
console.log(results.length);
|
|
const singlePageIndexers = this.prepareHrefIndexers(results);
|
|
if (singlePageIndexers.length === 0) {
|
|
console.log("THERE IS NOT EVEN SINGLE RESULT");
|
|
return []
|
|
}
|
|
|
|
return Promise.map(singlePageIndexers, function (indexer) {
|
|
return indexer.indexSingle();
|
|
}).then(async (results) => {
|
|
console.log("SinglePageMethod in HAS NO RESULTS, MarketAralms");
|
|
console.log(results.length);
|
|
return results;
|
|
});
|
|
|
|
} else {
|
|
console.log("HAS MORE RESULTS, should only contain HREFS");
|
|
console.log(results.length);
|
|
const newResults = await this.indexWithPagination(results[0].pageNumber + 5);
|
|
const singlePageIndexers = this.prepareHrefIndexers(results);
|
|
|
|
const newerResults = await Promise.map(singlePageIndexers, function (indexer) {
|
|
return indexer.indexSingle();
|
|
}).then(async (results) => {
|
|
console.log("SinglePageMethod HAS RESULTS, should contain MarketAlerts only");
|
|
console.log(results.length);
|
|
return results;
|
|
});
|
|
|
|
Array.prototype.push.apply(newResults, newerResults);
|
|
return newResults;
|
|
|
|
}
|
|
});
|
|
} catch (e) {
|
|
console.error("Error has accured", e);
|
|
}
|
|
|
|
}
|
|
|
|
prepareIndexers(pageNr) {
|
|
console.log("Entering prepareIndexers : page nr - " + pageNr);
|
|
const indexers = [];
|
|
let lastPageNumber;
|
|
if (pageNr) {
|
|
for (let index = Number(pageNr[0]); index <= Number(pageNr[0]) + 5; index++) {
|
|
lastPageNumber = index;
|
|
const newOlxUrl = {
|
|
url: this.olxUrl.url.replace(/\d+$/, "") + index,
|
|
email: this.olxUrl.email,
|
|
uuid: this.olxUrl.uuid
|
|
}
|
|
indexers.push(new Indexer(newOlxUrl));
|
|
|
|
}
|
|
} else {
|
|
for (let index = 1; index <= 5; index++) {
|
|
lastPageNumber = index;
|
|
const newOlxUrl = {
|
|
url: this.olxUrl.url + index,
|
|
email: this.olxUrl.email,
|
|
uuid: this.olxUrl.uuid
|
|
}
|
|
indexers.push(new Indexer(newOlxUrl));
|
|
}
|
|
}
|
|
return {
|
|
indexers: indexers,
|
|
lastPageNumber: lastPageNumber
|
|
};
|
|
}
|
|
|
|
prepareHrefIndexers(results) {
|
|
const indexers = []
|
|
|
|
if (!Array.isArray(results)) {
|
|
results.hrefs.forEach(href => {
|
|
const newOlxUrl = {
|
|
url: href,
|
|
email: results.olxUrl.email,
|
|
uuid: results.olxUrl.uuid
|
|
}
|
|
|
|
indexers.push(new Indexer(newOlxUrl));
|
|
});
|
|
|
|
} else {
|
|
|
|
|
|
results.forEach(result => {
|
|
|
|
if (result !== null && result.hasOwnProperty('hrefs')) {
|
|
result.hrefs.forEach(href => {
|
|
// console.log(href);
|
|
const newOlxUrl = {
|
|
url: href,
|
|
email: result.olxUrl.email,
|
|
uuid: result.olxUrl.uuid
|
|
}
|
|
|
|
indexers.push(new Indexer(newOlxUrl));
|
|
})
|
|
}
|
|
|
|
});
|
|
}
|
|
|
|
return indexers;
|
|
}
|
|
|
|
async indexPage(pageNumber) {
|
|
console.log("Page number in index page, max page number :")
|
|
console.log(pageNumber);
|
|
|
|
try {
|
|
|
|
console.log("Indexing page: " + this.olxUrl.url);
|
|
const res = await fetch(this.olxUrl.url);
|
|
const body = await res.text();
|
|
const $ = cheerio.load(body);
|
|
const hrefs = [];
|
|
let hasResults = false
|
|
|
|
$('#rezultatipretrage').find('.listitem').each((i, elem) => {
|
|
hasResults = true
|
|
const href = $(elem).find('a').first().attr('href');
|
|
hrefs.push(href);
|
|
});
|
|
|
|
console.log("this is hrefs for olxUrl" + this.olxUrl.url);
|
|
console.log("NUMBER OF HREFS " + hrefs.length);
|
|
|
|
return {
|
|
hrefs: hrefs,
|
|
hasResults: hasResults,
|
|
pageNumber: pageNumber,
|
|
olxUrl: this.olxUrl
|
|
}
|
|
} catch (e) {
|
|
console.error('Exception caught:' + e);
|
|
}
|
|
}
|
|
|
|
async indexSingle() {
|
|
try {
|
|
console.log("Index single");
|
|
console.log(this.olxUrl.url);
|
|
|
|
if (this.olxUrl.url === undefined) {
|
|
return {}
|
|
}
|
|
|
|
if (global.hrefs) {
|
|
|
|
if (global.hrefs[this.olxUrl.uuid] && global.hrefs[this.olxUrl.uuid].includes(this.olxUrl.url)) {
|
|
|
|
console.log("We found duplicate URL");
|
|
return null
|
|
}
|
|
}
|
|
|
|
const res = await fetch(this.olxUrl.url);
|
|
const body = await res.text();
|
|
const $ = cheerio.load(body);
|
|
|
|
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 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 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 parsePrice = (price) => parseFloat(price.replace(".", ""))
|
|
|
|
if (matches && matches.length >= 3) {
|
|
lat = matches[1];
|
|
lng = matches[2];
|
|
}
|
|
|
|
const parsedPrice = parsePrice(price);
|
|
|
|
const locationArray = location.split(",");
|
|
const region = locationArray[0];
|
|
const municipality = locationArray[1];
|
|
|
|
const data = {
|
|
realEstateType: this.getCategoryId(realEstateType),
|
|
email: this.olxUrl.email,
|
|
uuid: this.olxUrl.uuid,
|
|
olxId: olxId,
|
|
url: this.olxUrl.url,
|
|
title,
|
|
price: isNaN(parsedPrice) ? 0 : parsedPrice,
|
|
size: parseFloat(size),
|
|
gardenSize: isNaN(parseFloat(gardenSize)) ? 0 : parseFloat(gardenSize),
|
|
address,
|
|
region,
|
|
municipality,
|
|
time,
|
|
shortDescription: descriptions.first().text(),
|
|
longDescription: descriptions.last().text(),
|
|
lat,
|
|
lng,
|
|
loc: [parseFloat(lat), parseFloat(lng)],
|
|
};
|
|
|
|
return data;
|
|
} catch (e) {
|
|
console.error('Exception caught: ' + e.message);
|
|
}
|
|
|
|
return null;
|
|
}
|
|
|
|
getCategoryId(category) {
|
|
|
|
switch (category) {
|
|
case 'Stanovi':
|
|
return 'stan';
|
|
|
|
case 'Vikendice':
|
|
return 'vikendica'
|
|
|
|
case 'Kuće':
|
|
return 'kuca';
|
|
|
|
default:
|
|
return '';
|
|
}
|
|
}
|
|
}
|
|
|