replace old crawler, without specific crawler and saver implementation
This commit is contained in:
Bilal Catic
@@ -1,435 +0,0 @@
|
||||
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(hrefs = []) {
|
||||
this.hrefs = hrefs;
|
||||
}
|
||||
|
||||
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() {
|
||||
const filteredResults = [];
|
||||
const realEstateRequests = await allRERequest();
|
||||
const urls = this.createRequestUrls(realEstateRequests);
|
||||
let results = await this.indexPages(
|
||||
urls,
|
||||
this.fromPage,
|
||||
this.toPage,
|
||||
this.maxResults
|
||||
);
|
||||
const flatResults = results.flat();
|
||||
if (flatResults) {
|
||||
for (const finalResult of flatResults) {
|
||||
if (null !== finalResult) {
|
||||
if (
|
||||
finalResult.lat !== undefined &&
|
||||
finalResult.lat !== null &&
|
||||
finalResult.lat !== ""
|
||||
) {
|
||||
const pointInsideBoundingBox = await findPointInsideBoundingBox(
|
||||
[finalResult.lng, finalResult.lat],
|
||||
finalResult.email,
|
||||
finalResult.uuid
|
||||
);
|
||||
|
||||
if (pointInsideBoundingBox[0].length !== 0) {
|
||||
finalResult.hasLocation = true;
|
||||
filteredResults.push(finalResult);
|
||||
} else {
|
||||
finalResult.hasLocation = false;
|
||||
filteredResults.push(finalResult);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
return filteredResults;
|
||||
}
|
||||
return [];
|
||||
}
|
||||
|
||||
createRequestUrls(realEstateRequests) {
|
||||
const urls = [];
|
||||
|
||||
for (const request of realEstateRequests) {
|
||||
const {
|
||||
realEstateType,
|
||||
region,
|
||||
sizeMin,
|
||||
sizeMax,
|
||||
priceMin,
|
||||
priceMax
|
||||
} = request;
|
||||
|
||||
const urlRealEstateParams = [
|
||||
{
|
||||
paramName: "kanton",
|
||||
paramValue: region,
|
||||
useParam: false
|
||||
},
|
||||
{
|
||||
paramName: "kategorija",
|
||||
paramValue: getRealEstateTypeEnum(realEstateType).olxid,
|
||||
useParam: true
|
||||
},
|
||||
{
|
||||
paramName: "kvadrata_min",
|
||||
paramValue: sizeMin,
|
||||
useParam: true
|
||||
},
|
||||
{
|
||||
paramName: "kvadrata_max",
|
||||
paramValue: sizeMax,
|
||||
useParam: true
|
||||
},
|
||||
{
|
||||
paramName: "od",
|
||||
paramValue: priceMin,
|
||||
useParam: true
|
||||
},
|
||||
{
|
||||
paramName: "do",
|
||||
paramValue: priceMax,
|
||||
useParam: true
|
||||
}
|
||||
];
|
||||
const urlResultsParams = [
|
||||
{
|
||||
paramName: "vrstapregleda",
|
||||
paramValue: "tabela",
|
||||
useParam: true
|
||||
},
|
||||
{
|
||||
paramName: "sort_order",
|
||||
paramValue: "desc",
|
||||
useParam: true
|
||||
},
|
||||
{
|
||||
paramName: "vrsta",
|
||||
paramValue: "samoprodaja",
|
||||
useParam: true
|
||||
},
|
||||
{
|
||||
paramName: "stranica",
|
||||
paramValue: "0",
|
||||
useParam: true
|
||||
}
|
||||
];
|
||||
|
||||
const paramsReduceFunction = (accumulatedValue, currentParam) => {
|
||||
const { paramName, paramValue, useParam } = currentParam;
|
||||
if (useParam) {
|
||||
return `${accumulatedValue}&${paramName}=${paramValue}`;
|
||||
} else {
|
||||
return accumulatedValue;
|
||||
}
|
||||
};
|
||||
|
||||
const reducedRealEstateParams = urlRealEstateParams.reduce(
|
||||
paramsReduceFunction,
|
||||
""
|
||||
);
|
||||
const reducedResultsParams = urlResultsParams.reduce(
|
||||
paramsReduceFunction,
|
||||
""
|
||||
);
|
||||
|
||||
const olxUrl = {
|
||||
url: `https://www.olx.ba/pretraga?${reducedRealEstateParams}${reducedResultsParams}`,
|
||||
email: request.email,
|
||||
uuid: request.uniqueId,
|
||||
hrefs: this.hrefs
|
||||
};
|
||||
urls.push(olxUrl);
|
||||
}
|
||||
|
||||
return urls;
|
||||
}
|
||||
};
|
||||
|
||||
class Indexer {
|
||||
/**
|
||||
*
|
||||
* @param {String|Array} olxUrl single or array of objects containing url email and uuid
|
||||
* @param {Array} hrefResutls array contaning urls from crawler results
|
||||
*/
|
||||
|
||||
constructor(olxUrl, hrefResutls) {
|
||||
this.olxUrl = olxUrl;
|
||||
this.hrefResutls = hrefResutls;
|
||||
}
|
||||
|
||||
async indexWithPagination(pageNumber = 1) {
|
||||
const pageNr = this.olxUrl.url.match(/\d+$/);
|
||||
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) {
|
||||
hasResults = result.hasResults;
|
||||
}
|
||||
});
|
||||
|
||||
if (!hasResults) {
|
||||
const singlePageIndexers = this.prepareHrefIndexers(results);
|
||||
if (singlePageIndexers.length === 0) {
|
||||
return [];
|
||||
}
|
||||
|
||||
return Promise.map(singlePageIndexers, function(indexer) {
|
||||
return indexer.indexSingle();
|
||||
}).then(async results => {
|
||||
return results;
|
||||
});
|
||||
} else {
|
||||
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 => {
|
||||
return results;
|
||||
});
|
||||
|
||||
Array.prototype.push.apply(newResults, newerResults);
|
||||
return newResults;
|
||||
}
|
||||
});
|
||||
} catch (e) {
|
||||
console.error("Error has accured", e);
|
||||
}
|
||||
}
|
||||
|
||||
prepareIndexers(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,
|
||||
hrefs: this.olxUrl.hrefs
|
||||
};
|
||||
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,
|
||||
hrefs: this.olxUrl.hrefs
|
||||
};
|
||||
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,
|
||||
hrefs: this.olxUrl.hrefs
|
||||
};
|
||||
|
||||
indexers.push(new Indexer(newOlxUrl));
|
||||
});
|
||||
} else {
|
||||
results.forEach(result => {
|
||||
if (result !== null && result.hasOwnProperty("hrefs")) {
|
||||
result.hrefs.forEach(href => {
|
||||
const newOlxUrl = {
|
||||
url: href,
|
||||
email: result.olxUrl.email,
|
||||
uuid: result.olxUrl.uuid,
|
||||
hrefs: this.olxUrl.hrefs
|
||||
};
|
||||
|
||||
indexers.push(new Indexer(newOlxUrl));
|
||||
});
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
return indexers;
|
||||
}
|
||||
|
||||
async indexPage(pageNumber) {
|
||||
try {
|
||||
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);
|
||||
});
|
||||
return {
|
||||
hrefs: hrefs,
|
||||
hasResults: hasResults,
|
||||
pageNumber: pageNumber,
|
||||
olxUrl: this.olxUrl
|
||||
};
|
||||
} catch (e) {
|
||||
console.error("Exception caught:" + e);
|
||||
}
|
||||
}
|
||||
|
||||
async indexSingle() {
|
||||
try {
|
||||
if (this.olxUrl.url === undefined) {
|
||||
return {};
|
||||
}
|
||||
|
||||
// if (global.hrefs) {
|
||||
|
||||
if (
|
||||
this.olxUrl.hrefs[this.olxUrl.uuid] &&
|
||||
this.olxUrl.hrefs[this.olxUrl.uuid].includes(this.olxUrl.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 && location.length > 0 ? location.split(",") : [];
|
||||
const region = locationArray.length > 0 ? locationArray[0] : "";
|
||||
const municipality = locationArray.length > 1 ? 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 "";
|
||||
}
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user