Moved crawler for enqueue places to main crawler

This commit is contained in:
JakubDrobnik 2018-12-10 15:27:34 +01:00
parent e02ccfe8c5
commit 0689be833c
4 changed files with 127 additions and 142 deletions

View File

@ -1,11 +1,6 @@
{ {
"name": "crawler-google-places", "name": "crawler-google-places",
"actId": null, "version": "0.1",
"version": {
"versionNumber": "0.1",
"buildTag": "latest", "buildTag": "latest",
"envVars": [], "env": null
"sourceType": "TARBALL",
"tarballUrl": null
}
} }

View File

@ -1,9 +1,11 @@
const Apify = require('apify'); const Apify = require('apify');
const { sleep } = Apify.utils;
const { injectJQuery } = Apify.utils.puppeteer;
const { MAX_PAGE_RETRIES, DEFAULT_TIMEOUT, LISTING_PAGINATION_KEY } = require('./consts');
const waitForGoogleMapLoader = (page) => page.waitFor(() => !document.querySelector('#searchbox').classList.contains('loading'), { timeout: DEFAULT_TIMEOUT }); const { sleep } = Apify.utils;
const { DEFAULT_TIMEOUT, LISTING_PAGINATION_KEY } = require('./consts');
const waitForGoogleMapLoader = (page) => page.waitFor(() => !document.querySelector('#searchbox')
.classList
.contains('loading'), { timeout: DEFAULT_TIMEOUT });
const enqueueAllUrlsFromPagination = async (page, requestQueue) => { const enqueueAllUrlsFromPagination = async (page, requestQueue) => {
let results = await page.$$('.section-result'); let results = await page.$$('.section-result');
@ -36,14 +38,7 @@ const enqueueAllUrlsFromPagination = async (page, requestQueue) => {
* @param listingPagination * @param listingPagination
* @param retries * @param retries
*/ */
const enqueueAllPlaceDetailsCrawler = async (startUrl, searchString, launchPuppeteerOptions, requestQueue, listingPagination, retries = 0) => { const enqueueAllPlaceDetailsCrawler = async (page, searchString, launchPuppeteerOptions, requestQueue, listingPagination) => {
let browser;
try {
browser = await Apify.launchPuppeteer(launchPuppeteerOptions);
const page = await browser.newPage();
await page._client.send('Emulation.clearDeviceMetricsOverride');
await page.goto(startUrl);
await injectJQuery(page);
await page.type('#searchboxinput', searchString); await page.type('#searchboxinput', searchString);
await sleep(5000); await sleep(5000);
await page.click('#searchbox-searchbutton'); await page.click('#searchbox-searchbutton');
@ -61,10 +56,10 @@ const enqueueAllPlaceDetailsCrawler = async (startUrl, searchString, launchPuppe
await requestQueue.addRequest({ url, userData: { label: 'detail' } }); await requestQueue.addRequest({ url, userData: { label: 'detail' } });
return; return;
} }
const nextButtonSelector = '#section-pagination-button-next'; const nextButtonSelector = '[jsaction="pane.paginationSection.nextPage"]';
while (true) { while (true) {
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT }); await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
const paginationText = await page.$eval('.section-pagination-right', (el) => el.innerText); const paginationText = await page.$eval('.n7lv7yjyC35__right', (el) => el.innerText);
const [fromString, toString] = paginationText.match(/\d+/g); const [fromString, toString] = paginationText.match(/\d+/g);
const from = parseInt(fromString); const from = parseInt(fromString);
const to = parseInt(toString); const to = parseInt(toString);
@ -78,29 +73,19 @@ const enqueueAllPlaceDetailsCrawler = async (startUrl, searchString, launchPuppe
} }
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT }); await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
const isNextPaginationDisabled = await page.evaluate((nextButtonSelector) => { const isNextPaginationDisabled = await page.evaluate((nextButtonSelector) => {
return !!$(nextButtonSelector).attr('disabled'); return !!$(nextButtonSelector)
.attr('disabled');
}, nextButtonSelector); }, nextButtonSelector);
const noResultsEl = await page.$('.section-no-result-title'); const noResultsEl = await page.$('.section-no-result-title');
if (isNextPaginationDisabled || noResultsEl) { if (isNextPaginationDisabled || noResultsEl) {
break; break;
} else { } else {
// NOTE: puppeteer API click() didn't work :( // NOTE: puppeteer API click() didn't work :(
await page.evaluate((sel) => $(sel).click(), nextButtonSelector); await page.evaluate((sel) => $(sel)
.click(), nextButtonSelector);
await waitForGoogleMapLoader(page); await waitForGoogleMapLoader(page);
} }
} }
} catch (err) {
if (retries < MAX_PAGE_RETRIES) {
++retries;
console.log(`Retiring enqueueAllPlaceDetails for ${retries} time, error:`);
console.error(err);
await browser.close();
await enqueueAllPlaceDetailsCrawler(startUrl, searchString, launchPuppeteerOptions, requestQueue, listingPagination, ++retries);
}
throw err;
} finally {
if (browser) await browser.close();
}
}; };
module.exports = { run: enqueueAllPlaceDetailsCrawler }; module.exports = { run: enqueueAllPlaceDetailsCrawler };

View File

@ -1,7 +1,5 @@
const Apify = require('apify'); const Apify = require('apify');
const placesCrawler = require('./places_crawler'); const placesCrawler = require('./places_crawler');
const enqueueAllPlaceDetailsCrawler = require('./enqueue_places_crawler');
const { LISTING_PAGINATION_KEY } = require('./consts');
Apify.main(async () => { Apify.main(async () => {
const input = await Apify.getValue('INPUT'); const input = await Apify.getValue('INPUT');
@ -22,21 +20,11 @@ Apify.main(async () => {
console.log('Start url is', startUrl); console.log('Start url is', startUrl);
const requestQueue = await Apify.openRequestQueue(); const requestQueue = await Apify.openRequestQueue();
await requestQueue.addRequest({ url: startUrl, userData: { label: 'startUrl', searchString } });
// Store state of listing pagination
// NOTE: Ensured - If pageFunction failed crawler skipped already scraped pagination
const listingPagination = await Apify.getValue(LISTING_PAGINATION_KEY) || {};
const launchPuppeteerOptions = {}; const launchPuppeteerOptions = {};
if (proxyConfig) Object.assign(launchPuppeteerOptions, proxyConfig); if (proxyConfig) Object.assign(launchPuppeteerOptions, proxyConfig);
// Enqueue all links to scrape from listings
if (!listingPagination.isFinish) {
console.log(`Start enqueuing place details for search: ${searchString}`);
await enqueueAllPlaceDetailsCrawler.run(startUrl, searchString, launchPuppeteerOptions, requestQueue, listingPagination);
listingPagination.isFinish = true;
await Apify.setValue(LISTING_PAGINATION_KEY, listingPagination);
}
// Scrape all place detail links // Scrape all place detail links
const crawler = placesCrawler.setUpCrawler(launchPuppeteerOptions, requestQueue); const crawler = placesCrawler.setUpCrawler(launchPuppeteerOptions, requestQueue);
await crawler.run(); await crawler.run();

View File

@ -4,7 +4,8 @@ const { sleep } = Apify.utils;
const infiniteScroll = require('./infinite_scroll'); const infiniteScroll = require('./infinite_scroll');
const { injectJQuery } = Apify.utils.puppeteer; const { injectJQuery } = Apify.utils.puppeteer;
const { MAX_PAGE_RETRIES, DEFAULT_TIMEOUT } = require('./consts'); const { MAX_PAGE_RETRIES, DEFAULT_TIMEOUT, LISTING_PAGINATION_KEY } = require('./consts');
const enqueueAllPlaceDetailsCrawler = require('./enqueue_places_crawler');
/** /**
* Method to set up crawler to get all place details and save them to default dataset * Method to set up crawler to get all place details and save them to default dataset
@ -18,17 +19,32 @@ const setUpCrawler = (launchPuppeteerOptions, requestQueue) => {
requestQueue, requestQueue,
maxRequestRetries: MAX_PAGE_RETRIES, maxRequestRetries: MAX_PAGE_RETRIES,
retireInstanceAfterRequestCount: 10, retireInstanceAfterRequestCount: 10,
handlePageTimeoutSecs: 600, handlePageTimeoutSecs: 2 * 3600, // Two hours because startUrl crawler
maxOpenPagesPerInstance: 1, // Because startUrl crawler crashes if we mixed it with details scraping
// maxConcurrency: 1, // maxConcurrency: 1,
gotoFunction: async ({ request, page }) => { gotoFunction: async ({ request, page }) => {
await page._client.send('Emulation.clearDeviceMetricsOverride'); await page._client.send('Emulation.clearDeviceMetricsOverride');
await page.goto(request.url, { timeout: 60000 }); await page.goto(request.url, { timeout: 60000 });
}, },
handlePageFunction: async ({ request, page }) => { handlePageFunction: async ({ request, page }) => {
const { label } = request.userData; const { label, searchString } = request.userData;
console.log(`Open ${request.url} with label: ${label}`); console.log(`Open ${request.url} with label: ${label}`);
// Get data from review
await injectJQuery(page); await injectJQuery(page);
if (label === 'startUrl') {
// enqueue all places
console.log(`Start enqueuing place details for search: ${searchString}`);
// Store state of listing pagination
// NOTE: Ensured - If pageFunction failed crawler skipped already scraped pagination
const listingPagination = await Apify.getValue(LISTING_PAGINATION_KEY) || {};
await enqueueAllPlaceDetailsCrawler.run(page, searchString, launchPuppeteerOptions, requestQueue, listingPagination);
listingPagination.isFinish = true;
await Apify.setValue(LISTING_PAGINATION_KEY, listingPagination);
} else {
// Timeout because timeout for handle page is 2 hours
setTimeout(() => {
throw new Error('HandlePagefunction timed out!');
}, 600000);
// Get data from review
await page.waitForSelector('h1.section-hero-header-title', { timeout: DEFAULT_TIMEOUT }); await page.waitForSelector('h1.section-hero-header-title', { timeout: DEFAULT_TIMEOUT });
const placeDetail = await page.evaluate(() => { const placeDetail = await page.evaluate(() => {
return { return {
@ -85,6 +101,7 @@ const setUpCrawler = (launchPuppeteerOptions, requestQueue) => {
} }
} }
await Apify.pushData(placeDetail); await Apify.pushData(placeDetail);
}
console.log(request.url, 'Done'); console.log(request.url, 'Done');
}, },