diff --git a/app/crawler/crawl.js b/app/crawler/crawl.js index ac4825d..002e17a 100644 --- a/app/crawler/crawl.js +++ b/app/crawler/crawl.js @@ -7,8 +7,13 @@ */ const OlxCrawler = require("./specificCrawlers/olx"); const RentalCrawler = require("./specificCrawlers/rental"); +const ProstorCrawler = require("./specificCrawlers/prostor"); -const { OLX_CONFIG, RENTAL_CONFIG } = require("./crawlerConfig"); +const { + OLX_CONFIG, + RENTAL_CONFIG, + PROSTOR_CONFIG +} = require("./crawlerConfig"); const PostgresSaver = require("./savers/postgres"); async function crawlAll() { @@ -32,6 +37,15 @@ async function crawlAll() { RENTAL_CONFIG.RENTAL_MAX_RESULTS_PER_PAGE, RENTAL_CONFIG.RENTAL_IGNORED_USERNAMES, RENTAL_CONFIG.RENTAL_DELAY_BETWEEN_PAGES + ), + new ProstorCrawler( + [postgresSaver], + PROSTOR_CONFIG.PROSTOR_CRAWLER_AD_TYPE, + PROSTOR_CONFIG.PROSTOR_CRAWLER_AD_CATEGORIES, + PROSTOR_CONFIG.PROSTOR_MAX_PAGES, + PROSTOR_CONFIG.PROSTOR_MAX_RESULTS_PER_PAGE, + PROSTOR_CONFIG.PROSTOR_IGNORED_USERNAMES, + PROSTOR_CONFIG.PROSTOR_DELAY_BETWEEN_PAGES ) ]; diff --git a/app/crawler/specificConfigs/prostor.js b/app/crawler/specificConfigs/prostor.js index cde72d0..098fc95 100644 --- a/app/crawler/specificConfigs/prostor.js +++ b/app/crawler/specificConfigs/prostor.js @@ -24,7 +24,7 @@ const transformedProstorCrawlerAdCategories = prostorParsedCrawlerAdCategories module.exports = { PROSTOR_MAX_PAGES: parseInt(process.env.PROSTOR_MAX_PAGES) || 100, PROSTOR_MAX_RESULTS_PER_PAGE: - parseInt(process.env.PROSTOR_MAX_RESULTS_PER_PAGE) || 50, + parseInt(process.env.PROSTOR_MAX_RESULTS_PER_PAGE) || 5000, PROSTOR_CRAWLER_AD_TYPE: prostorCrawlerAdType || CRAWLER_AD_TYPE.NONE, PROSTOR_CRAWLER_AD_CATEGORIES: transformedProstorCrawlerAdCategories, PROSTOR_IGNORED_USERNAMES: prostorIgnoredUsernames || [], diff --git a/app/crawler/specificCrawlers/prostor.js b/app/crawler/specificCrawlers/prostor.js new file mode 100644 index 0000000..16bcce1 --- /dev/null +++ b/app/crawler/specificCrawlers/prostor.js @@ -0,0 +1,248 @@ +"use strict"; + +const fetch = require("node-fetch"); +const cheerio = require("cheerio"); + +const { + AD_TYPE, + AD_CATEGORY, + AD_AGENCY, + AD_STATUS, + CRAWLER_AD_TYPE +} = require("../../common/enums"); + +const PROSTOR_ENUMS = { + PROSTOR_AD_TYPE: { + [CRAWLER_AD_TYPE.ALL]: "&action=0", + [CRAWLER_AD_TYPE.ONLY_SELL]: "&action=1", + [CRAWLER_AD_TYPE.ONLY_RENT]: "&action=2" + }, + PROSTOR_AD_CATEGORY: { + [AD_CATEGORY.ALL.id]: "", + [AD_CATEGORY.FLAT.id]: "&type=7", + [AD_CATEGORY.HOUSE.id]: "&type=8", + [AD_CATEGORY.LAND.id]: "&type=10", + [AD_CATEGORY.OFFICE.id]: "&type=9", + [AD_CATEGORY.APARTMENT.id]: "&type=11", + [AD_CATEGORY.GARAGE.id]: "&type=14" + //[AD_CATEGORY.COTTAGE.id]: "" + }, + PROSTOR_PUBLISHED_DATE_FORMAT: "YYYY-MM-DD HH:mm:ss", + PROSTOR_RENEWED_DATE_FORMAT: "YYYY-MM-DD u HH:mm:ss" +}; + +class ProstorCrawler { + constructor( + savers = [], + crawlerAdTypes = CRAWLER_AD_TYPE.ALL, + crawlerAdCategories = [AD_CATEGORY.FLAT, AD_CATEGORY.HOUSE], + maxPages = 5000, + maxResultsPerPage = 5000, + ignoredUsernames = [], + delayBetweenPages = 1000 + ) { + this.savers = savers; + this.baseUrl = "https://prostor.ba/pretraga"; + this.crawlerAdTypes = crawlerAdTypes; + this.crawlerAdCategories = crawlerAdCategories; + this.maxResultsPerPage = maxResultsPerPage; + } + + async crawl() { + const crawlAdCategories = this.crawlerAdCategories; + const newRealEstates = []; + + if (crawlAdCategories) { + for (const adCategory of crawlAdCategories) { + const urlAdTypePart = + PROSTOR_ENUMS.PROSTOR_AD_TYPE[this.crawlerAdTypes]; + const urlCategoryPart = PROSTOR_ENUMS.PROSTOR_AD_CATEGORY[adCategory]; + if (urlAdTypePart !== undefined && urlCategoryPart !== undefined) { + const urlPageToCrawl = `${this.baseUrl}?remove_sold=1${urlAdTypePart}${urlCategoryPart}`; + const singleCategoryResults = await this.extractRealEstates( + urlPageToCrawl + ); + + const resultsSubset = singleCategoryResults.slice( + 0, + this.maxResultsPerPage + ); + + const saveResults = await this.saveCrawledResults(resultsSubset); + const { newRecords } = saveResults; + newRealEstates.push(...newRecords); + } + } + } + return newRealEstates; + } + + async extractRealEstates(url) { + console.log("[PROSTOR] Index page : ", url); + + try { + const res = await fetch(url); + const body = await res.text(); + const $ = cheerio.load(body); + + const scriptElement = $( + "body > div > div.container-fluid > script:nth-child(7)" + ); + + if ( + scriptElement[0] && + scriptElement[0].children && + scriptElement[0].children[0] && + scriptElement[0].children[0].data + ) { + const scriptData = scriptElement[0].children[0].data; + + try { + // script element data contains JS code and we need to extract only data for realEstates + // data string starts with : var map; var markers = [{"r ... + // so we remove first 23 characters + // + // real estate JSON data ends with ...}, ]; map = new... + // so we need to find index of that substring to know where to stop + // we will NOT include trailing comma because it breaks JSON parse, so we have to close ] bracket manually + + const jsonEndIndex = scriptData.indexOf(", ]; map = new"); + if (jsonEndIndex > -1) { + const jsonData = scriptData.substring(23, jsonEndIndex) + "]"; + const realEstates = JSON.parse(jsonData); + + const transformedRealEstates = []; + + for (const realEstate of realEstates) { + const transformedRealEstate = ProstorCrawler.transformRealEstateData( + realEstate + ); + if (transformedRealEstate) { + transformedRealEstates.push(transformedRealEstate); + } + } + + return transformedRealEstates; + } else { + throw { + message: "Something is wrong with JSON data or data is moved" + }; + } + } catch (e) { + console.log(e); + throw { message: "Can't find ad data JSON" }; + } + } + } catch (e) { + console.error("[PROSTOR] Exception caught:", e.message); + return []; + } + } + + static transformRealEstateData(realEstateData) { + try { + const { lat, lng, property_name, price, size, link } = realEstateData; + + // link contains part of the URL in the format of : /prodaja/stan/stup/9556 + // general form is : /actionType/realEstateType/location/realEstateID + // linkParts contains : ['', 'actionType', 'realEstateType', 'location', 'realEstateID'] + + const linkParts = link.split("/"); + + const adType = ProstorCrawler.getAdTypeId(linkParts[1]); + const realEstateType = ProstorCrawler.getAdCategoryId(linkParts[2]); + const prostorId = linkParts[4]; + const url = `https://prostor.ba${link}`; + + if (!adType || !realEstateType || !prostorId) { + return null; + } + + const adStatus = AD_STATUS.STATUS_NORMAL; + const parsedPrice = parseFloat(price.replace(/\./g, "")) || null; + const parsedArea = parseFloat(size); + + const data = { + url, + agencyObjectId: prostorId, + originAgencyName: AD_AGENCY.PROSTOR, + realEstateType, + adType, + title: property_name, + price: parsedPrice, + area: parsedArea, + gardenSize: null, + shortDescription: "", + longDescription: "", + streetNumber: 0, + streetName: "", + locality: "", + municipality: "", + city: "", + region: "", + entity: "", + country: "", + locationLat: lat, + locationLong: lng, + adStatus, + publishedDate: null, + renewedDate: null + }; + + return data; + } catch (e) { + console.error( + "[PROSTOR] Exception caught: " + e.message, + "\r\nURL:", + url + ); + return null; + } + } + + //======= HELPER FUNCTIONS ============= + + static getAdCategoryId(categoryText) { + switch (categoryText) { + case "stan": + return AD_CATEGORY.FLAT.id; + case "kuca": + return AD_CATEGORY.HOUSE.id; + case "apartman": + return AD_CATEGORY.APARTMENT.id; + case "poslovni-prostor": + return AD_CATEGORY.OFFICE.id; + case "garaza": + return AD_CATEGORY.GARAGE.id; + case "zemljiste": + return AD_CATEGORY.LAND.id; + default: + return undefined; + } + } + + static getAdTypeId(adTypeText) { + switch (adTypeText) { + case "prodaja": + return AD_TYPE.AD_TYPE_SALE; + case "najam": + return AD_TYPE.AD_TYPE_RENT; + default: + return undefined; + } + } + + async saveCrawledResults(results) { + const savers = this.savers; + + // for (const saver of savers) { + // await saver.save(results); + // } + + //For now, we use only Postgres saver, so ... + return await savers[0].save(results); + //so that we can use some sequelize options and information when data is inserted + } +} + +module.exports = ProstorCrawler;