612 lines
18 KiB
JavaScript
612 lines
18 KiB
JavaScript
"use strict";
|
|
|
|
const fetch = require("node-fetch");
|
|
const cheerio = require("cheerio");
|
|
const moment = require("moment-timezone");
|
|
|
|
const {
|
|
AD_TYPE,
|
|
AD_CATEGORY,
|
|
AD_AGENCY,
|
|
AD_STATUS,
|
|
CRAWLER_AD_TYPE,
|
|
FURNISHING_TYPE,
|
|
HEATING_TYPE
|
|
} = require("../../common/enums");
|
|
|
|
const {
|
|
PRINT_CRAWLER_DEBUG,
|
|
DEFAULT_TIMEZONE
|
|
} = require("../../config/appConfig");
|
|
const { SALJIC_FORCE_CRAWL } = require("../specificConfigs/saljic");
|
|
|
|
const SALJIC_ENUMS = {
|
|
SALJIC_AD_TYPE: {
|
|
[CRAWLER_AD_TYPE.ALL]: "&input_vrsta=",
|
|
[CRAWLER_AD_TYPE.ONLY_SELL]: "&input_vrsta=1",
|
|
[CRAWLER_AD_TYPE.ONLY_RENT]: "&input_vrsta=2"
|
|
},
|
|
SALJIC_AD_CATEGORY: {
|
|
[AD_CATEGORY.ALL.id]: "&input_kategorija=",
|
|
[AD_CATEGORY.FLAT.id]: "&input_kategorija=15",
|
|
[AD_CATEGORY.HOUSE.id]: "&input_kategorija=9",
|
|
[AD_CATEGORY.LAND.id]: "&input_kategorija=5", //3 and 4 also gradjevinsko
|
|
[AD_CATEGORY.OFFICE.id]: "&input_kategorija=8",
|
|
[AD_CATEGORY.APARTMENT.id]: "&input_kategorija=1",
|
|
[AD_CATEGORY.GARAGE.id]: "&input_kategorija=2"
|
|
//[AD_CATEGORY.COTTAGE.id]: ""
|
|
}
|
|
};
|
|
|
|
class SaljicCrawler {
|
|
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://www.saljicnekretnine.ba/v2/nekretnine_search";
|
|
this.crawlerAdTypes = crawlerAdTypes;
|
|
this.crawlerAdCategories = crawlerAdCategories;
|
|
this.maxResultsPerPage = maxResultsPerPage;
|
|
this.delayBetweenPages = delayBetweenPages;
|
|
}
|
|
|
|
async crawl() {
|
|
const crawlAdCategories = this.crawlerAdCategories;
|
|
|
|
const newRealEstates = [];
|
|
|
|
if (crawlAdCategories) {
|
|
const indexGenerators = [];
|
|
for (const adCategory of crawlAdCategories) {
|
|
indexGenerators.push(this.categoryIndexer(adCategory));
|
|
}
|
|
//
|
|
//console.log(indexGenerators);
|
|
//
|
|
let done = false;
|
|
while (!done) {
|
|
const categoryIndexerPromises = [];
|
|
const generatorsToRemove = [];
|
|
for (const indexGenerator of indexGenerators) {
|
|
categoryIndexerPromises.push(indexGenerator.next());
|
|
generatorsToRemove.push(false);
|
|
}
|
|
|
|
const singlePageResults = await Promise.all(categoryIndexerPromises);
|
|
const entries = singlePageResults.entries();
|
|
|
|
for (const [index, { value: singlePageResult }] of entries) {
|
|
if (singlePageResult) {
|
|
const saveResults = await this.saveCrawledResults(singlePageResult);
|
|
const { newRecords } = saveResults;
|
|
|
|
newRealEstates.push(...newRecords);
|
|
|
|
if (
|
|
Array.isArray(newRecords) &&
|
|
newRecords.length === 0 &&
|
|
!SALJIC_FORCE_CRAWL
|
|
) {
|
|
generatorsToRemove[index] = true;
|
|
}
|
|
} else {
|
|
//Generator returned undefined, remove this generator from array
|
|
generatorsToRemove[index] = true;
|
|
// console.log("Generator ", index + 1, "has no more pages");
|
|
}
|
|
}
|
|
|
|
// console.log("Generators state : ", generatorsToRemove);
|
|
for (let i = generatorsToRemove.length - 1; i >= 0; i--) {
|
|
if (generatorsToRemove[i]) {
|
|
// console.log("\tRemove generator ", i + 1);
|
|
indexGenerators.splice(i, 1);
|
|
}
|
|
}
|
|
if (indexGenerators.length === 0) {
|
|
done = true;
|
|
}
|
|
|
|
await this.sleep(this.delayBetweenPages);
|
|
}
|
|
}
|
|
return newRealEstates;
|
|
}
|
|
|
|
async *categoryIndexer(adCategory) {
|
|
let pageToIndex = 1;
|
|
|
|
const urlAdTypePart = SALJIC_ENUMS.SALJIC_AD_TYPE[this.crawlerAdTypes];
|
|
const urlCategoryPart = SALJIC_ENUMS.SALJIC_AD_CATEGORY[adCategory];
|
|
|
|
if (urlAdTypePart !== undefined && urlCategoryPart !== undefined) {
|
|
while (true) {
|
|
const urlPagePart = pageToIndex === 1 ? "" : (pageToIndex - 1) * 2 * 11;
|
|
const urlPageToCrawl = `${this.baseUrl}?order_by=${urlAdTypePart}${urlCategoryPart}&per_page=${urlPagePart}`;
|
|
|
|
const singlePageResults = await this.indexSinglePage(
|
|
urlPageToCrawl,
|
|
this.maxResultsPerPage
|
|
);
|
|
|
|
if (Array.isArray(singlePageResults) && singlePageResults.length > 0) {
|
|
yield singlePageResults;
|
|
} else {
|
|
return undefined;
|
|
}
|
|
|
|
++pageToIndex;
|
|
if (pageToIndex === this.maxPages) {
|
|
return undefined;
|
|
}
|
|
}
|
|
} else {
|
|
return undefined;
|
|
}
|
|
}
|
|
|
|
async indexSinglePage(url, maxResultsPerPage) {
|
|
if (PRINT_CRAWLER_DEBUG) {
|
|
console.log("[SALJIC] Index page : ", url);
|
|
}
|
|
|
|
try {
|
|
const res = await fetch(url);
|
|
const body = await res.text();
|
|
const $ = cheerio.load(body);
|
|
let hrefs = [];
|
|
|
|
$("#shop")
|
|
.find(".product")
|
|
.each((i, elem) => {
|
|
const href = $(elem)
|
|
.find("a")
|
|
.first()
|
|
.attr("href");
|
|
if (href) {
|
|
hrefs.push(href);
|
|
}
|
|
});
|
|
|
|
let adTypesTmp = [];
|
|
|
|
$("#shop")
|
|
.find(".product")
|
|
.each((i, elem) => {
|
|
const adType = $(elem)
|
|
.find(".trakica-search-page")
|
|
.text()
|
|
.trim();
|
|
if (adType) {
|
|
adTypesTmp.push(adType);
|
|
}
|
|
});
|
|
|
|
//Converting to AD_TYPE
|
|
const adTypes = adTypesTmp.map(adTypeText => {
|
|
return this.getAdTypeId(adTypeText);
|
|
});
|
|
|
|
//Converting to absolute URLs
|
|
const hrefsAbs = hrefs.map(link => {
|
|
return "https://www.saljicnekretnine.ba" + link;
|
|
});
|
|
|
|
let actualNoOfResults =
|
|
hrefsAbs.length <= maxResultsPerPage
|
|
? hrefsAbs.length
|
|
: maxResultsPerPage;
|
|
|
|
const asyncScraping = [];
|
|
for (let i = 0; i < actualNoOfResults; i++) {
|
|
asyncScraping.push(this.scrapeAd(hrefsAbs[i], adTypes[i]));
|
|
}
|
|
|
|
const scrapedData = await Promise.all(asyncScraping);
|
|
const filteredScrapedData = scrapedData.filter(adData => !!adData);
|
|
return filteredScrapedData;
|
|
} catch (e) {
|
|
console.error("[SALJIC] Exception caught:" + e);
|
|
return [];
|
|
}
|
|
}
|
|
|
|
async scrapeAd(url, adType) {
|
|
console.log("[SALJIC] Scraping : ", url);
|
|
try {
|
|
const adPageSource = await fetch(url);
|
|
const body = await adPageSource.text();
|
|
const $ = cheerio.load(body);
|
|
|
|
// No information for status ex. PRODAN
|
|
const status = AD_STATUS.STATUS_NORMAL;
|
|
//Extracting agency ID from url
|
|
const agencyObjectId = parseInt(url.substring(46, url.length));
|
|
|
|
//Extracting main properties
|
|
const propertySelectors = {
|
|
title:
|
|
"div.content-wrap > div.container > div.col-md-8.nobottommargin > div.single-post > div.entry > div.entry-title > h2",
|
|
price:
|
|
"div.content-wrap > div.container > div.col-md-8.nobottommargin > div.single-post > div.entry > div.topmargin-sm.single-product > div.product > div.product-price > ins",
|
|
streetName:
|
|
"div.content-wrap > div.container > div.col-md-8.nobottommargin > div.single-post > div.entry > div.entry-content.topmargin > p",
|
|
|
|
descriptions:
|
|
"div.content-wrap > div.container > div.col-md-8.nobottommargin > div.single-post > div.entry > div.entry-content.topmargin > div.toggle.toggle-bg > div.togglec >p:nth-child(1)",
|
|
latAndLong:
|
|
"div.content-wrap > div.container > div.col-md-8.nobottommargin > div.single-post > div.entry > div.entry-content.topmargin > div.gmap.bottommargin > iframe"
|
|
};
|
|
const title = $(propertySelectors.title)
|
|
.text()
|
|
.replace(/(\r\n|\n|\r)/gm, "")
|
|
.replace(/ {1,}/g, " ")
|
|
.trim();
|
|
|
|
const priceText = $(propertySelectors.price)
|
|
.text()
|
|
.replace(/(\r\n|\n|\r)/gm, "")
|
|
.replace(/ {1,}/g, " ")
|
|
.trim();
|
|
const price =
|
|
priceText === "CIJENA NA UPIT"
|
|
? null
|
|
: parseFloat(
|
|
priceText.substring(8, priceText.length - 3).replace(",", "")
|
|
);
|
|
|
|
const streetName = $(propertySelectors.streetName)
|
|
.text()
|
|
.replace(/(\r\n|\n|\r)/gm, "")
|
|
.trim();
|
|
|
|
const descriptions = $(propertySelectors.descriptions)
|
|
.text()
|
|
.trim();
|
|
|
|
const latAndLongSrc = $(propertySelectors.latAndLong).attr("src");
|
|
const latText = latAndLongSrc.substring(
|
|
latAndLongSrc.indexOf("marker=") + 7,
|
|
latAndLongSrc.indexOf("%2C", latAndLongSrc.indexOf("marker="))
|
|
);
|
|
const longText = latAndLongSrc.substring(
|
|
latAndLongSrc.indexOf("%2C", latAndLongSrc.indexOf("marker=")) + 3,
|
|
latAndLongSrc.length
|
|
);
|
|
const locationLat = parseFloat(latText) || null;
|
|
const locationLong = parseFloat(longText) || null;
|
|
|
|
//====== DETAIL INFORMATION FIELDS ==========
|
|
let area,
|
|
gardenSize,
|
|
numberOfRooms = null,
|
|
numberOfFloors = null,
|
|
floor = null,
|
|
accessRoadType = null,
|
|
heatingType = null,
|
|
furnishingType = null,
|
|
balcony = null,
|
|
newBuilding = null,
|
|
elevator = null,
|
|
water = null,
|
|
electricity = null,
|
|
drainageSystem = null,
|
|
registeredInZkBooks = null,
|
|
recentlyAdapted = null,
|
|
parking = null,
|
|
garage = null,
|
|
gas = null,
|
|
antiTheftDoor = null,
|
|
airCondition = null,
|
|
phoneConnection = null,
|
|
cableTV = null,
|
|
internet = null,
|
|
basementAttic = null,
|
|
storeRoom = null,
|
|
videoSurveillance = null,
|
|
alarm = null,
|
|
suitableForStudents = null,
|
|
includingBills = null,
|
|
animalsAllowed = null,
|
|
pool = null,
|
|
urbanPlanPermit = null,
|
|
buildingPermit = null,
|
|
utilityConnection = null,
|
|
distanceToRiver = null;
|
|
let publishedDate = null;
|
|
let renewedDate = null;
|
|
let realEstateType;
|
|
let numberOfViewsAgency = null;
|
|
|
|
//Extracting data - Glavne karakteristike
|
|
let mainFieldIndex = 1;
|
|
do {
|
|
const mainFieldSelector = `div.content-wrap > div.container > div.col-md-8.nobottommargin > div.single-post > div.entry > div.entry-content.topmargin > div.col-md-12.bottommargin > ul > li.list-group-item:nth-child(${mainFieldIndex})`;
|
|
|
|
const mainField = $(mainFieldSelector)
|
|
.text()
|
|
.replace(/[\n\r\t]/gm, "")
|
|
.trim();
|
|
|
|
const mainFieldTitle = mainField.substring(0, mainField.indexOf(" "));
|
|
const mainFieldValue = mainField
|
|
.substring(mainField.indexOf(" "), mainField.length)
|
|
.trim();
|
|
|
|
switch (mainFieldTitle) {
|
|
case "Površina":
|
|
area = parseFloat(
|
|
mainFieldValue.substring(0, mainFieldValue.indexOf(" "))
|
|
);
|
|
break;
|
|
case "Okućnica":
|
|
gardenSize = parseFloat(
|
|
mainFieldValue.substring(0, mainFieldValue.indexOf(" "))
|
|
);
|
|
break;
|
|
case "Broj soba":
|
|
numberOfRooms = parseInt(mainFieldValue);
|
|
break;
|
|
case "Broj spratova":
|
|
numberOfFloors = parseInt(mainFieldValue);
|
|
break;
|
|
case "Sprat":
|
|
floor = parseInt(mainFieldValue);
|
|
break;
|
|
case "Godina renoviranja":
|
|
recentlyAdapted = true;
|
|
break;
|
|
case "Broj parking mjesta":
|
|
parking = true;
|
|
break;
|
|
case "Dostupno od":
|
|
const day = mainFieldValue.substring(0, 2);
|
|
const month = mainFieldValue.substring(3, 5);
|
|
const year = mainFieldValue.substring(6, mainFieldValue.length);
|
|
publishedDate = new Date(`${month}/${day}/${year}`);
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
|
|
if (mainFieldTitle === "") {
|
|
break;
|
|
}
|
|
mainFieldIndex++;
|
|
} while (true);
|
|
|
|
//Extracting data - Sadrzaji
|
|
let additionalFieldIndex = 1;
|
|
do {
|
|
const additionalFieldSelector = `div.content-wrap > div.container > div.col-md-8.nobottommargin > div.single-post > div.entry > div.entry-content.topmargin > div.col-md-12.bottommargin > ul > li.border-color.col-md-5.col-md-offset-1.col-md-pull-1.list-group-item-bottom:nth-child(${additionalFieldIndex})`;
|
|
|
|
const additionalField = $(additionalFieldSelector)
|
|
.text()
|
|
.trim();
|
|
|
|
if (additionalFieldIndex === 1) {
|
|
//Extracting data of real estate type
|
|
const categoryTmp = additionalField
|
|
.replace(/[\n\r\t]/gm, "")
|
|
.substring(
|
|
additionalField.indexOf("Kategorija") + 10,
|
|
additionalField.length
|
|
)
|
|
.trim();
|
|
realEstateType = this.getAdCategoryId(categoryTmp);
|
|
} else {
|
|
switch (additionalField) {
|
|
case "Internet":
|
|
internet = true;
|
|
break;
|
|
case "Garaža":
|
|
garage = true;
|
|
break;
|
|
case "Klima":
|
|
airCondition = true;
|
|
break;
|
|
case "Balkon":
|
|
balcony = true;
|
|
break;
|
|
case "Ostava":
|
|
storeRoom = true;
|
|
break;
|
|
case "Podrum":
|
|
basementAttic = true;
|
|
break;
|
|
case "Blindirana vrata":
|
|
antiTheftDoor = true;
|
|
break;
|
|
case "Voda":
|
|
water = true;
|
|
break;
|
|
case "Kablovska":
|
|
cableTV = true;
|
|
break;
|
|
case "Uknjiženo":
|
|
registeredInZkBooks = true;
|
|
break;
|
|
case "Grijanje - centralno":
|
|
heatingType = HEATING_TYPE.CENTRAL_CITY.id;
|
|
break;
|
|
case "Grijanje - plin":
|
|
heatingType = HEATING_TYPE.GAS.id;
|
|
break;
|
|
case "Grijanje - struja":
|
|
heatingType = HEATING_TYPE.ELECTRICITY.id;
|
|
break;
|
|
case "Grijanje":
|
|
heatingType = HEATING_TYPE.OTHER.id;
|
|
break;
|
|
case "Plin":
|
|
gas = true;
|
|
break;
|
|
case "Namješten":
|
|
furnishingType = FURNISHING_TYPE.FURNISHED.id;
|
|
break;
|
|
case "Alarm":
|
|
alarm = true;
|
|
break;
|
|
case "Video nadzor":
|
|
videoSurveillance = true;
|
|
break;
|
|
case "Lift":
|
|
elevator = true;
|
|
break;
|
|
case "Novogradnja":
|
|
newBuilding = true;
|
|
break;
|
|
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (additionalField === "") {
|
|
break;
|
|
}
|
|
additionalFieldIndex++;
|
|
} while (true);
|
|
|
|
//If no published date it takes current date of crawling
|
|
if (publishedDate) {
|
|
renewedDate = new Date();
|
|
} else {
|
|
publishedDate = new Date();
|
|
renewedDate = new Date();
|
|
}
|
|
|
|
const data = {
|
|
url,
|
|
agencyObjectId,
|
|
originAgencyName: AD_AGENCY.SALJIC,
|
|
realEstateType,
|
|
adType,
|
|
title,
|
|
price,
|
|
area,
|
|
gardenSize,
|
|
shortDescription: descriptions.substring(0, descriptions.indexOf(".")),
|
|
longDescription: descriptions,
|
|
streetNumber: 0,
|
|
streetName,
|
|
locality: "",
|
|
municipality: "",
|
|
city: "",
|
|
region: "",
|
|
entity: "",
|
|
country: "",
|
|
locationLat,
|
|
locationLong,
|
|
adStatus: status,
|
|
publishedDate,
|
|
renewedDate,
|
|
numberOfRooms,
|
|
numberOfFloors,
|
|
floor,
|
|
accessRoadType,
|
|
heatingType,
|
|
furnishingType,
|
|
balcony,
|
|
newBuilding,
|
|
elevator,
|
|
water,
|
|
electricity,
|
|
drainageSystem,
|
|
registeredInZkBooks,
|
|
recentlyAdapted,
|
|
parking,
|
|
garage,
|
|
gas,
|
|
antiTheftDoor,
|
|
airCondition,
|
|
phoneConnection,
|
|
cableTV,
|
|
internet,
|
|
basementAttic,
|
|
storeRoom,
|
|
videoSurveillance,
|
|
alarm,
|
|
suitableForStudents,
|
|
includingBills,
|
|
animalsAllowed,
|
|
pool,
|
|
urbanPlanPermit,
|
|
buildingPermit,
|
|
utilityConnection,
|
|
distanceToRiver,
|
|
numberOfViewsAgency
|
|
};
|
|
console.log(data);
|
|
return data;
|
|
} catch (e) {
|
|
console.error("Exception caught: " + e.message, "\r\nURL:", url);
|
|
}
|
|
return null;
|
|
}
|
|
|
|
//======= HELPER FUNCTIONS =============
|
|
|
|
getAdCategoryId(categoryText) {
|
|
switch (categoryText) {
|
|
case "Stan":
|
|
return AD_CATEGORY.FLAT.id;
|
|
case "Građevinsko zemljiste":
|
|
return AD_CATEGORY.LAND.id;
|
|
case "Industrijsko zemljiste":
|
|
return AD_CATEGORY.LAND.id;
|
|
case "Poljoprivredno zemljiste":
|
|
return AD_CATEGORY.LAND.id;
|
|
case "Kuća":
|
|
return AD_CATEGORY.HOUSE.id;
|
|
case "Poslovni prostor":
|
|
return AD_CATEGORY.OFFICE.id;
|
|
case "Kancelarije":
|
|
return AD_CATEGORY.OFFICE.id;
|
|
case "Apartmani":
|
|
return AD_CATEGORY.APARTMENT.id;
|
|
case "Garaža":
|
|
return AD_CATEGORY.GARAGE.id;
|
|
case "Vikendica":
|
|
return AD_CATEGORY.COTTAGE.id;
|
|
default:
|
|
return undefined;
|
|
}
|
|
}
|
|
|
|
getAdTypeId(adTypeText) {
|
|
switch (adTypeText) {
|
|
case "PRODAJA":
|
|
return AD_TYPE.AD_TYPE_SALE.stringId;
|
|
case "NAJAM":
|
|
return AD_TYPE.AD_TYPE_RENT.stringId;
|
|
default:
|
|
return undefined;
|
|
}
|
|
}
|
|
|
|
async sleep(ms) {
|
|
return new Promise(resolve => setTimeout(resolve, ms));
|
|
}
|
|
|
|
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 savers[0].save(results);
|
|
//so that we can use some sequelize options and information when data is inserted
|
|
}
|
|
}
|
|
|
|
module.exports = SaljicCrawler;
|