cc-tracking/tasks/fresh Scraper.md
2019-05-21 16:40:16 +01:00

376 lines
9.0 KiB
Markdown

Fresh Scraper
========================
```javascript
const Scraper = require('../helpers/scraper');
const cheerio = require('cheerio');
const path = require('path');
const jsonfile = require('jsonfile');
const logger = require('log4js').getLogger('LV');
const url = require('url');
const removeAccents = require('remove-accents-diacritics');
logger.level = process.env.LOGGER_LEVEL || 'warn';
class LVScrape extends Scraper {
constructor() {
super();
this.id = 'LV';
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 serviceObject
* @returns {Promise<void>}
*/
async processEntityDetails(serviceObject) {
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 indexRedirector() {
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;
}
}
/**
*
* @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;
}
}
async processNewPage() {
// give the page a few seconds to settle
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;
}
switch (pageUrl.pathname) {
case '/en/our-registers/company-register/':
await this.indexRedirector();
break;
case '/en/our-registers/company-register/details':
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: ${pageUrl}`);
}
else {
logger.warn('processNewPage Fell through');
logger.warn('currentPage.location', pageUrl);
}
break;
}
}
/**
*
* @returns {Promise<void>}
*/
async attachEvents() {
this.on('serviceDone', async () => {
switch (this.mode) {
case 0:
this.emit('paymentServicesDone');
break;
case 1:
this.emit('emoneyServicesDone');
break;
case 2:
this.emit('creditServicesDone');
break;
}
});
this.on('psindexdone', async () => {
let newUrl;
this.paymentServices.items = this.paymentServices.links.length;
logger.info(`${this.paymentServices.items} items indexed`);
this.paymentServices.indexStep++;
if (this.paymentServices.indexStep >= this.paymentServices.urls.length)
newUrl = this.paymentServices.links[this.paymentServices.step].href;
else
newUrl = this.paymentServices.urls[this.paymentServices.indexStep];
await this._goto(newUrl);
});
this.on('emindexdone', async () => {
let newUrl;
this.emoneyServices.items = this.emoneyServices.links.length;
logger.info(`${this.emoneyServices.items} items indexed`);
this.emoneyServices.indexStep++;
if (this.emoneyServices.indexStep >= this.emoneyServices.urls.length)
newUrl = this.emoneyServices.links[this.emoneyServices.step].href;
else
newUrl = this.emoneyServices.urls[this.emoneyServices.indexStep];
await this._goto(newUrl);
});
this.on('ciindexdone', async () => {
let newUrl;
this.creditServices.items = this.creditServices.links.length;
logger.info(`${this.creditServices.items} items indexed`);
this.creditServices.indexStep++;
if (this.creditServices.indexStep >= this.creditServices.urls.length)
newUrl = this.creditServices.links[this.creditServices.step].href;
else
newUrl = this.creditServices.urls[this.creditServices.indexStep];
await this._goto(newUrl);
});
this.on('indexdone', async () => {
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 () => {
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 () => {
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 () => {
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() {
super._start();
try {
this.mode = 0;
this.rootURI = 'http://www.fktk.lv';
this.paymentServices = {
'items': 0,
'links': [],
'step': 0,
'indexStep': 0,
'visited': false,
'done' : false,
'urls': ['http://www.fktk.lv/en/market/payment-institutions/authorized-payment-institutions.html', 'http://www.fktk.lv/en/market/payment-institutions/registered-payment-institutions.html'],
'sections' : [],
'sectionLinks' : []
};
this.emoneyServices = {
'items': 0,
'links': [],
'step': 0,
'indexStep': 0,
'visited': false,
'done' : false,
'urls': ['http://www.fktk.lv/en/market/electronic-money-institutions/authorized-electronic-money-institutions.html', 'http://www.fktk.lv/en/market/electronic-money-institutions/registered-electronic-money-institutions.html'],
'sections' : [],
'sectionLinks' : []
};
this.creditServices = {
'items': 0,
'links': [],
'step': 0,
'indexStep': 0,
'visited': false,
'done' : false,
'searchDone' : false,
'started': false,
'urls': ['http://www.fktk.lv/en/market/credit-institutions/banks.html'],
'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/LV/FCMC`));
await this._doNonRepudiation().catch((err) => {
logger.warn(err);
});
await this._initBrowser();
await this._createBrowserPage();
this.page.on('domcontentloaded', this._throttle(async () => {
this.processNewPage().catch((err) => {
logger.error('processNewPage fail', err);
});
}, 2500));
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 = LVScrape;
```