514 lines
13 KiB
JavaScript
514 lines
13 KiB
JavaScript
|
|
const Scraper = require('../helpers/scraper');
|
|
const cheerio = require('cheerio');
|
|
const path = require('path');
|
|
const jsonfile = require('jsonfile');
|
|
const removeAccents = require('remove-accents-diacritics');
|
|
const logger = require('log4js').getLogger('(PT)');
|
|
const url = require('url');
|
|
|
|
logger.level = process.env.LOGGER_LEVEL || 'warn';
|
|
|
|
class PTScrape extends Scraper {
|
|
|
|
constructor() {
|
|
super();
|
|
this.setID('PT');
|
|
|
|
this.on('done', () => {
|
|
this._done();
|
|
});
|
|
|
|
this.run = this._throttle(async () => {
|
|
await this.__run();
|
|
}, 5000);
|
|
|
|
if (process.env.NODE_ENV === 'production')
|
|
this._checkLock().then((l) => {
|
|
if(l)
|
|
this.run();
|
|
});
|
|
}
|
|
|
|
/**
|
|
*
|
|
* @param html
|
|
* @returns {Promise<void>}
|
|
*/
|
|
async extractEntityDetails(html) {
|
|
try {
|
|
const details = {};
|
|
const detailSequence = [
|
|
|
|
['field-name-field-tipo-ent-aut', 'institutionType'],
|
|
|
|
['field-name-field-estado-ent', 'state'],
|
|
|
|
['field-name-field-morada', 'address'],
|
|
|
|
['field-name-field-localidade', 'firstName'],
|
|
|
|
['field-name-field-cod-postal', 'postcode'],
|
|
|
|
['field-name-field-pais', 'country'],
|
|
|
|
['field-name-field-data-limite', 'beginningOfActivity'],
|
|
|
|
['field-name-field-capital-subscrito', 'subscribedCapital'],
|
|
|
|
['field-name-field-capital-realizado', 'paidUpCapital'],
|
|
|
|
['field-name-field-jel', 'institutionCodeNumber']
|
|
|
|
];
|
|
|
|
const $ = cheerio.load(html);
|
|
|
|
details.name = this._cleanUp($('h1.page-title').text()) ;
|
|
|
|
const mainDiv = $('div.content');
|
|
|
|
for(const item of detailSequence) {
|
|
const i = $(mainDiv).find(`.${item[0]} div.field-items`);
|
|
|
|
details[item[1]] = this._cleanUp($(i).text());
|
|
}
|
|
|
|
return details;
|
|
}
|
|
catch( err) {
|
|
logger.error(err);
|
|
}
|
|
}
|
|
|
|
/**
|
|
*
|
|
* @param serviceObject
|
|
* @returns {Promise<void>}
|
|
*/
|
|
async processIndex(serviceObject) {
|
|
logger.info(`Building the ${this.modeTitles[this.mode]} index...`);
|
|
await this._randomWait(this.page, 3, 5);
|
|
|
|
const body = await this.page.content();
|
|
|
|
const $ = cheerio.load(body);
|
|
|
|
if ($('div.view-empty').length > 0) {
|
|
// We have reached an empty page, so we assume we've scraped all links from this index
|
|
this.emit('indexdone');
|
|
|
|
return;
|
|
}
|
|
|
|
const links = $('div.views-field.views-field-title > span > a');
|
|
|
|
links.each((i, item) => {
|
|
const href = $(item).attr('href');
|
|
const text = $(item).text();
|
|
|
|
const newUrl = `https://www.bportugal.pt${href}`;
|
|
const id = this._makeFieldName(text);
|
|
|
|
serviceObject.links.push({ 'name':text, 'href':newUrl, 'id':id });
|
|
});
|
|
|
|
const filename = this.modeNames[this.mode];
|
|
|
|
const parsedUrl = url.parse(this.page.url(), true);
|
|
|
|
await this._makeScreenshotV2(this.page, `${this.path}/${filename}_main_${parsedUrl.query.page}`, null);
|
|
|
|
parsedUrl.query.page++;
|
|
parsedUrl.search = undefined; // Forces parsedUrl to use `query` property, as modified on line above
|
|
const nextPage = url.format(parsedUrl);
|
|
|
|
await this._goto(nextPage);
|
|
}
|
|
|
|
/**
|
|
*
|
|
* @param serviceObject
|
|
* @returns {Promise<void>}
|
|
*/
|
|
async buildIndex(serviceObject) {
|
|
// We have stopped using the "view all" button due to it breaking.
|
|
// Leaving the code below commented in case it is ever useful in future.
|
|
|
|
// await this.page.waitForSelector('#block-system-main > div > div > div.view-content-wrapper > ul > li.pager__item.pager__item_all', { 'visible':true, 'timeout':7500 }).then(async (elm) => {
|
|
// logger.debug('Extend menu list..');
|
|
// await elm.click({ 'delay':90 });
|
|
// }).catch(() => {
|
|
// logger.info('No show all button');
|
|
// });
|
|
|
|
await this._randomWait(this.page, 6, 9);
|
|
|
|
await this.processIndex(serviceObject);
|
|
}
|
|
|
|
/**
|
|
*
|
|
* @returns {Promise<void>}
|
|
*/
|
|
async indexRedirector() {
|
|
logger.debug('>> indexRedirector');
|
|
let doIndex = false;
|
|
await this.page.waitForSelector('input[value="Lista podmiotów"]', { 'visible':true, 'timeout':7500 }).then(async (elm) => {
|
|
logger.warn('Sent back to the main selector screen');
|
|
await elm.click({ 'delay':90 });
|
|
|
|
doIndex = false;
|
|
}).catch(() => {
|
|
// logger.info('No show all button');
|
|
doIndex = true;
|
|
});
|
|
|
|
if (doIndex)
|
|
switch (this.mode) {
|
|
|
|
case 0:
|
|
await this.buildIndex(this.paymentServices);
|
|
break;
|
|
|
|
case 1:
|
|
await this.buildIndex(this.emoneyServices);
|
|
break;
|
|
|
|
case 2:
|
|
await this.buildIndex(this.creditServices);
|
|
break;
|
|
|
|
}
|
|
}
|
|
|
|
/**
|
|
*
|
|
* @param serviceObject
|
|
* @returns {Promise<void>}
|
|
*/
|
|
async processEntityDetails(serviceObject) {
|
|
const noWhiteSpace = /\W/g;
|
|
|
|
const { name, id } = serviceObject.links[serviceObject.step];
|
|
// const id = serviceObject.links[serviceObject.step].id;
|
|
logger.info(`Process ${this.modeTitles[this.mode]} entity ${serviceObject.step + 1} of ${serviceObject.items} // ${name}`);
|
|
|
|
// 'h1.page-title'
|
|
|
|
await this.page.waitForSelector('h1.page-title', { 'visible':true, 'timeout':7500 });
|
|
|
|
await this._randomWait(this.page, 3, 5);
|
|
|
|
const entity = removeAccents.remove(id.trim());
|
|
|
|
const filename = [this.modePrefix[this.mode], entity.replace(noWhiteSpace, '_')].join('');
|
|
|
|
const filePath = `${this.path}/${filename}`.substring(0, 240);
|
|
|
|
await this._randomWait(this.page, 3, 5);
|
|
|
|
await this._makeScreenshotV2(this.page, `${filePath}_main`, null);
|
|
|
|
const body = await this.page.content();
|
|
|
|
const details = await this.extractEntityDetails(body);
|
|
|
|
await jsonfile.writeFile(`${filePath}.json`, { details });
|
|
|
|
await this._randomWait(this.page, 3, 5);
|
|
|
|
serviceObject.links[serviceObject.step].filename = `${filename}.json`;
|
|
serviceObject.step++;
|
|
|
|
if (serviceObject.step < serviceObject.items) {
|
|
const newUrl = serviceObject.links[serviceObject.step].href;
|
|
|
|
await this._goto(newUrl);
|
|
}
|
|
else
|
|
this.emit('serviceDone');
|
|
}
|
|
|
|
/**
|
|
*
|
|
* @returns {Promise<void>}
|
|
*/
|
|
async processRedirector() {
|
|
switch (this.mode) {
|
|
|
|
case 0:
|
|
await this.processEntityDetails(this.paymentServices);
|
|
break;
|
|
|
|
case 1:
|
|
await this.processEntityDetails(this.emoneyServices);
|
|
break;
|
|
|
|
case 2:
|
|
await this.processEntityDetails(this.creditServices);
|
|
break;
|
|
|
|
}
|
|
}
|
|
|
|
/**
|
|
*
|
|
* @returns {Promise<void>}
|
|
*/
|
|
async processNewPage() {
|
|
// give the page a few seconds to settle
|
|
const pathSplitter = /(\/en\/.+?\/)/;
|
|
|
|
await this._randomWait(this.page, 3, 5);
|
|
|
|
const pageUrl = url.parse(await this.page.url());
|
|
|
|
if (pageUrl.href === 'chrome-error://chromewebdata/') {
|
|
logger.warn('Directed to: chrome-error://chromewebdata/');
|
|
this.emit('recover');
|
|
|
|
return;
|
|
}
|
|
|
|
const splitPath = pageUrl.pathname.match(pathSplitter);
|
|
|
|
const pathname = splitPath[0];
|
|
|
|
switch (pathname) {
|
|
|
|
case '/en/entidades-autorizadas/':
|
|
await this.indexRedirector();
|
|
break;
|
|
|
|
case '/en/entidadeautorizada/':
|
|
await this.processRedirector();
|
|
break;
|
|
case '/en/our-registers/company-register/gransoverskridandehandel/':
|
|
await this.crossBorderRedirector();
|
|
break;
|
|
|
|
default:
|
|
if (process.env.NODE_ENV) {
|
|
await this._uploadError();
|
|
throw new Error(`Unknown page: ${pageUrl}`);
|
|
}
|
|
else {
|
|
logger.warn('processNewPage Fell through');
|
|
logger.warn('currentPage.location', pageUrl);
|
|
}
|
|
break;
|
|
|
|
}
|
|
}
|
|
|
|
async attachEvents() {
|
|
this.on('entityComplete', () => {
|
|
this.handleEntityComplete();
|
|
});
|
|
|
|
this.on('serviceDone', async function() {
|
|
switch (this.mode) {
|
|
|
|
case 0:
|
|
this.emit('paymentServicesDone');
|
|
break;
|
|
|
|
case 1:
|
|
this.emit('emoneyServicesDone');
|
|
break;
|
|
|
|
case 2:
|
|
this.emit('creditServicesDone');
|
|
break;
|
|
|
|
}
|
|
});
|
|
|
|
//
|
|
|
|
this.on('psindexdone', async () => {
|
|
this.paymentServices.items = this.paymentServices.links.length;
|
|
logger.info(`${this.paymentServices.items} items indexed`);
|
|
|
|
const newUrl = this.paymentServices.links[this.paymentServices.step].href;
|
|
|
|
await this._goto(newUrl);
|
|
});
|
|
|
|
this.on('emindexdone', async () => {
|
|
this.emoneyServices.items = this.emoneyServices.links.length;
|
|
logger.info(`${this.emoneyServices.items} items indexed`);
|
|
|
|
const newUrl = this.emoneyServices.links[this.emoneyServices.step].href;
|
|
|
|
await this._goto(newUrl);
|
|
});
|
|
|
|
this.on('ciindexdone', async () => {
|
|
this.creditServices.items = this.creditServices.links.length;
|
|
logger.info(`${this.creditServices.items} items indexed`);
|
|
|
|
const newUrl = this.creditServices.links[this.creditServices.step].href;
|
|
|
|
await this._goto(newUrl);
|
|
});
|
|
|
|
this.on('indexdone', async function() {
|
|
switch (this.mode) {
|
|
|
|
case 0:
|
|
this.emit('psindexdone');
|
|
break;
|
|
|
|
case 1:
|
|
this.emit('emindexdone');
|
|
break;
|
|
|
|
case 2:
|
|
this.emit('ciindexdone');
|
|
break;
|
|
|
|
}
|
|
});
|
|
|
|
this.on('paymentServicesDone', async function() {
|
|
logger.warn('paymentServicesDone');
|
|
try{
|
|
this.paymentServices.done = true;
|
|
jsonfile.writeFileSync(`${this.path}/paymentServices.json`, { 'links': this.paymentServices.links });
|
|
jsonfile.writeFileSync(`${this.debugPath}/paymentServices.json`, this.paymentServices);
|
|
|
|
this.mode++;
|
|
this.inProgress = false;
|
|
|
|
await this._goto(this.emoneyServices.urls[0]);
|
|
}
|
|
catch (e) {
|
|
logger.error(e);
|
|
}
|
|
});
|
|
|
|
this.on('emoneyServicesDone', async function() {
|
|
logger.warn('emoneyServicesDone');
|
|
try{
|
|
this.emoneyServices.done = true;
|
|
jsonfile.writeFileSync(`${this.path}/emoneyServices.json`, { 'links':this.emoneyServices.links });
|
|
jsonfile.writeFileSync(`${this.debugPath}/emoneyServices.json`, this.emoneyServices);
|
|
this.mode++;
|
|
this.inProgress = false;
|
|
|
|
await this._goto(this.creditServices.urls[0]);
|
|
}
|
|
catch (e) {
|
|
logger.error(e);
|
|
}
|
|
});
|
|
|
|
this.on('creditServicesDone', async function() {
|
|
logger.warn('creditServicesDone');
|
|
try{
|
|
this.creditServices.done = true;
|
|
jsonfile.writeFileSync(`${this.path}/creditServices.json`, { 'links':this.creditServices.links });
|
|
jsonfile.writeFileSync(`${this.debugPath}/creditServices.json`, this.creditServices);
|
|
this.mode++;
|
|
this.inProgress = false;
|
|
|
|
this.emit('done');
|
|
}
|
|
catch (e) {
|
|
logger.error(e);
|
|
}
|
|
});
|
|
}
|
|
|
|
/**
|
|
*
|
|
* @returns {Promise<void>}
|
|
*/
|
|
async start() {
|
|
logger.debug(this.eventNames());
|
|
|
|
super._start();
|
|
try {
|
|
this.mode = 0;
|
|
|
|
this.paymentServices = {
|
|
'items': 0,
|
|
'links': [],
|
|
'step': 0,
|
|
'indexStep': 0,
|
|
'visited': false,
|
|
'done' : false,
|
|
'urls': ['https://www.bportugal.pt/en/entidades-autorizadas/75/all?page=0'],
|
|
'sections' : [],
|
|
'sectionLinks' : []
|
|
};
|
|
|
|
this.emoneyServices = {
|
|
'items': 0,
|
|
'links': [],
|
|
'step': 0,
|
|
'indexStep': 0,
|
|
'visited': false,
|
|
'done' : false,
|
|
'urls': ['https://www.bportugal.pt/en/entidades-autorizadas/72/all?page=0'],
|
|
'sections' : [],
|
|
'sectionLinks' : []
|
|
};
|
|
|
|
this.creditServices = {
|
|
'items': 0,
|
|
'links': [],
|
|
'step': 0,
|
|
'indexStep': 0,
|
|
'visited': false,
|
|
'done' : false,
|
|
'searchDone' : false,
|
|
'started': false,
|
|
'urls': ['https://www.bportugal.pt/en/entidades-autorizadas/67-68-1524-69/all?page=0'],
|
|
'sections' : [],
|
|
'sectionLinks' : []
|
|
};
|
|
|
|
this.startPage = this.paymentServices.urls[0];
|
|
this.emoneyUrl = this.emoneyServices.urls[0];
|
|
this.credit = this.creditServices.urls[0];
|
|
|
|
this.setPath(path.resolve(`${__dirname }/../artefacts/PT/BP`));
|
|
|
|
await this._doNonRepudiation().catch((err) => {
|
|
logger.warn(err);
|
|
});
|
|
|
|
await this._initBrowser();
|
|
await this._createBrowserPage();
|
|
await this._makeResponsive();
|
|
|
|
this.page.on('domcontentloaded', this._throttle(async () => {
|
|
this.processNewPage();
|
|
}, 5000));
|
|
|
|
if (this.eventNames().length === 2)
|
|
await this.attachEvents();
|
|
|
|
//
|
|
|
|
await this.page.setViewport({ 'width': 1200, 'height': 800 });
|
|
await this._goto(this.startPage, { 'waitUntil':'networkidle0' });
|
|
|
|
await this._randomWait(this.page, 3, 5);
|
|
}
|
|
catch(e) {
|
|
throw new Error(e);
|
|
}
|
|
}
|
|
|
|
async __run() {
|
|
await this.start();
|
|
}
|
|
|
|
}
|
|
|
|
module.exports = PTScrape;
|