Fixed issues from testing, structure code to files

This commit is contained in:
JakubDrobnik 2018-11-22 10:52:16 +01:00
parent 6230d31a57
commit c1203af221
5 changed files with 215 additions and 182 deletions

6
src/consts.js Normal file
View File

@ -0,0 +1,6 @@
// NOTE: This is not nice, it waits for implementing default timeout into puppeteer.
exports.DEFAULT_TIMEOUT = 60 * 1000; // 60 sec
exports.LISTING_PAGINATION_KEY = 'listingState';
exports.MAX_PAGE_RETRIES = 5;

View File

@ -0,0 +1,101 @@
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 enqueueAllUrlsFromPagination = async (page, requestQueue) => {
let results = await page.$$('.section-result');
const resultsCount = results.length;
for (let resultIndex = 0; resultIndex < resultsCount; resultIndex++) {
// Need to get results again, pupptr lost context..
await page.waitForSelector('.searchbox', { timeout: DEFAULT_TIMEOUT });
await waitForGoogleMapLoader(page);
await page.waitFor((resultIndex) => {
return document.querySelectorAll('.section-result h3').length >= resultIndex + 1;
}, { timeout: DEFAULT_TIMEOUT }, resultIndex);
results = await page.$$('.section-result');
const link = await results[resultIndex].$('h3');
await link.click();
await waitForGoogleMapLoader(page);
await page.waitForSelector('.section-back-to-list-button', { timeout: DEFAULT_TIMEOUT });
const url = page.url();
await requestQueue.addRequest({ url, userData: { label: 'detail' } });
console.log(`Added to queue ${url}`);
await page.click('.section-back-to-list-button');
}
};
/**
* Crawler add all place detail from listing to queue
* @param startUrl
* @param searchString
* @param launchPuppeteerOptions
* @param requestQueue
* @param listingPagination
* @param retries
*/
const enqueueAllPlaceDetailsCrawler = async (startUrl, searchString, launchPuppeteerOptions, requestQueue, listingPagination, retries = 0) => {
let browser;
try {
browser = await Apify.launchPuppeteer(launchPuppeteerOptions);
const page = await browser.newPage();
await page._client.send('Emulation.clearDeviceMetricsOverride');
await page.goto(startUrl);
await injectJQuery(page);
await page.type('#searchboxinput', searchString);
await sleep(5000);
await page.click('#searchbox-searchbutton');
await sleep(5000);
await waitForGoogleMapLoader(page);
// In case there is no listing, put just detail page to queue
const maybeDetailPlace = await page.$('h1.section-hero-header-title');
if (maybeDetailPlace) {
const url = page.url();
await requestQueue.addRequest({ url, userData: { label: 'detail' } });
return;
}
const nextButtonSelector = '#section-pagination-button-next';
while (true) {
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
const paginationText = await page.$eval('.section-pagination-right', (el) => el.innerText);
const [fromString, toString] = paginationText.match(/\d+/g);
const from = parseInt(fromString);
const to = parseInt(toString);
if (listingPagination.from && from <= listingPagination.from) {
console.log(`Skiped pagination ${from} - ${to}, already done!`);
} else {
console.log(`Added links from pagination ${from} - ${to}`);
await enqueueAllUrlsFromPagination(page, requestQueue);
listingPagination = { from, to };
await Apify.setValue(LISTING_PAGINATION_KEY, listingPagination);
}
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
const isNextPaginationDisabled = await page.evaluate((nextButtonSelector) => {
return !!$(nextButtonSelector).attr('disabled');
}, nextButtonSelector);
const noResultsEl = await page.$('.section-no-result-title');
if (isNextPaginationDisabled || noResultsEl) {
break;
} else {
// NOTE: puppeteer API click() didn't work :(
await page.evaluate((sel) => $(sel).click(), nextButtonSelector);
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 };

View File

@ -28,7 +28,7 @@ const getPageScrollInfo = (page, elementToScroll) => page.evaluate((elementToScr
/**
* Scroll to down page until infinite scroll ends or reaches maxHeight
* @param page - instance of crawled page
* @param maxHeight - max height of document to scrollscrollHeight
* @param maxHeight - max height of document to scrollHeight
* @param elementToScroll - CSS selector of element where we want to scroll, default is 'body'
* @return {Promise.<void>}
*/
@ -96,30 +96,29 @@ module.exports = async (page, maxHeight, elementToScroll = 'body') => {
const pendingRequestsCount = resourcesStats.requested - (resourcesStats.finished + resourcesStats.failed + resourcesStats.forgotten);
// We have to wait if all xhrs are finished
if (pendingRequestsCount === 0) {
// If the page is scrolled to the very bottom or beyond maximum height, we are done
const isLoaderOnPage = await page.evaluate(() => {
const loader = $('.section-loading-spinner');
if (loader) {
return loader.parent().attr('style') !== 'display: none;';
}
if (loader) return loader.parent().attr('style') !== 'display: none;';
});
const reviewsCount = await page.evaluate(() => $('div.section-review').length);
// console.log(reviewsCount, previosReviewsCount, isLoaderOnPage);
/**
* If the page is scrolled to the very bottom or beyond
* maximum height and loader is not displayed and we don't find new reviews, we are done.
*/
if (reviewsCount === previosReviewsCount
&& (scrollInfo.scrollTop + scrollInfo.clientHeight >= Math.min(scrollInfo.scrollHeight, maxHeight))
&& !isLoaderOnPage
) break;
previosReviewsCount = reviewsCount;
// Otherwise we try to scroll down
await scrollTo(page, elementToScroll, maxHeight);
}
await sleep(defaultScrollDelay);
}
// Scroll back up, otherwise the screenshot of the browser would only show the bottom of
// the page
await scrollTo(page, elementToScroll, maxHeight);
logInfo(`Infinite scroll finished (${stringifyScrollInfo(scrollInfo)} resourcesStats=${JSON.stringify(resourcesStats)})`);
} catch (err) {
logError('An exception thrown in infiniteScroll()', err);

View File

@ -1,96 +1,7 @@
/**
* Run the following example to perform a recursive crawl of a website using Puppeteer.
*/
const Apify = require('apify');
const infiniteScroll = require('./infinite_scroll');
const { sleep } = Apify.utils;
const { injectJQuery } = Apify.utils.puppeteer;
// NOTE: This is not nice, it waits for implementing default timeout into puppeteer.
const DEFAULT_TIMEOUT = 60 * 1000; // 60 sec
const LISTING_PAGINATION_KEY = 'listingState';
const MAX_PAGE_RETRIES = 5;
const waitForGoogleMapLoader = (page) => page.waitFor(() => !document.querySelector('#searchbox').classList.contains('loading'), { timeout: DEFAULT_TIMEOUT });
const enqueueAllUrlsFromPagination = async (page, requestQueue) => {
const detailLinks = [];
let results = await page.$$('.section-result');
const resultsCount = results.length;
for (let resultIndex = 0; resultIndex < resultsCount; resultIndex++) {
// Need to get results again, pupptr lost context..
await page.waitForSelector('.searchbox', { timeout: DEFAULT_TIMEOUT });
await waitForGoogleMapLoader(page);
await page.waitFor((resultIndex) => {
return document.querySelectorAll('.section-result h3').length >= resultIndex + 1;
}, { timeout: DEFAULT_TIMEOUT }, resultIndex);
results = await page.$$('.section-result');
const link = await results[resultIndex].$('h3');
await link.click();
await waitForGoogleMapLoader(page);
await page.waitForSelector('.section-back-to-list-button', { timeout: DEFAULT_TIMEOUT });
const url = page.url();
await requestQueue.addRequest({ url, userData: { label: 'detail' } });
console.log(`Added to queue ${url}`);
await page.click('.section-back-to-list-button');
}
return detailLinks;
};
const enqueueAllPlaceDetails = async (startUrl, searchString, launchPuppeteerOptions, requestQueue, listingPagination, retries = 0) => {
let browser;
try {
browser = await Apify.launchPuppeteer(launchPuppeteerOptions);
const page = await browser.newPage();
await page._client.send('Emulation.clearDeviceMetricsOverride');
await page.goto(startUrl);
await injectJQuery(page);
await page.type('#searchboxinput', searchString);
await sleep(5000);
await page.click('#searchbox-searchbutton');
await sleep(5000);
const nextButtonSelector = '#section-pagination-button-next';
while (true) {
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
const paginationText = await page.$eval('.section-pagination-right', (el) => el.innerText);
const [fromString, toString] = paginationText.match(/\d+/g);
const from = parseInt(fromString);
const to = parseInt(toString);
if (listingPagination.from && from <= listingPagination.from) {
console.log(`Skiped pagination ${from} - ${to}, already done!`);
} else {
console.log(`Added links from pagination ${from} - ${to}`);
await enqueueAllUrlsFromPagination(page, requestQueue);
listingPagination = { from, to };
await Apify.setValue(LISTING_PAGINATION_KEY, listingPagination);
}
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
const isNextPaginationDisabled = await page.evaluate((nextButtonSelector) => {
return !!$(nextButtonSelector).attr('disabled');
}, nextButtonSelector);
const noResultsEl = await page.$('.section-no-result-title');
if (isNextPaginationDisabled || noResultsEl) {
break;
} else {
// NOTE: puppeteer API click() didn't work :(
await page.evaluate((sel) => $(sel).click(), nextButtonSelector);
await waitForGoogleMapLoader(page);
}
}
} catch (err) {
if (retries < MAX_PAGE_RETRIES) {
++retries;
console.log(`Retiring enqueueAllPlaceDetails for ${retries} time.`);
await browser.close();
await enqueueAllPlaceDetails(startUrl, searchString, launchPuppeteerOptions, requestQueue, listingPagination, ++retries);
}
throw err;
} finally {
if (browser) await browser.close();
}
};
const placesCrawler = require('./places_crawler');
const enqueueAllPlaceDetailsCrawler = require('./enqueue_places_crawler');
const { LISTING_PAGINATION_KEY } = require('./consts');
Apify.main(async () => {
const input = await Apify.getValue('INPUT');
@ -110,100 +21,25 @@ Apify.main(async () => {
}
console.log('Start url is', startUrl);
const requestQueue = await Apify.openRequestQueue();
// 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 = {
// useApifyProxy: true,
// useChrome: true,
// apifyProxyGroups: ['CZECH_LUMINATI'],
// liveView: Apify.isAtHome(),
};
const launchPuppeteerOptions = {};
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 enqueueAllPlaceDetails(startUrl, searchString, launchPuppeteerOptions, requestQueue, listingPagination);
await enqueueAllPlaceDetailsCrawler.run(startUrl, searchString, launchPuppeteerOptions, requestQueue, listingPagination);
listingPagination.isFinish = true;
await Apify.setValue(LISTING_PAGINATION_KEY, listingPagination);
}
// Scrape all place detail links
const crawler = new Apify.PuppeteerCrawler({
launchPuppeteerOptions,
requestQueue,
maxRequestRetries: MAX_PAGE_RETRIES,
retireInstanceAfterRequestCount: 10,
handlePageTimeoutSecs: 600,
gotoFunction: async ({ request, page }) => {
await page._client.send('Emulation.clearDeviceMetricsOverride');
await page.goto(request.url, { timeout: 60000 });
},
handlePageFunction: async ({ request, page }) => {
const { label } = request.userData;
console.log(`Open ${request.url} with label: ${label}`);
// Get data from review
await injectJQuery(page);
await page.waitForSelector('h1.section-hero-header-title', { timeout: DEFAULT_TIMEOUT });
const placeDetail = await page.evaluate(() => {
return {
title: $('h1.section-hero-header-title').text().trim(),
totalScore: $('span.section-star-display').eq(0).text().trim(),
categoryName: $('[jsaction="pane.rating.category"]').text().trim(),
address: $('[data-section-id="ad"] .widget-pane-link').text().trim(),
plusCode: $('[data-section-id="ol"] .widget-pane-link').text().trim(),
};
});
placeDetail.url = request.url;
placeDetail.reviews = [];
if (placeDetail.totalScore) {
placeDetail.reviewsCount = await page.evaluate(() => {
const numberReviewsText = $('button.section-reviewchart-numreviews').text().trim();
return (numberReviewsText) ? numberReviewsText.match(/\d+/)[0] : null;
});
// Get all reviews
await page.click('button.section-reviewchart-numreviews');
await page.waitForSelector('.section-star-display', { timeout: DEFAULT_TIMEOUT });
await infiniteScroll(page, 99999999999, '.section-scrollbox.section-listbox');
await sleep(2000);
const reviewEls = await page.$$('div.section-review');
for (const reviewEl of reviewEls) {
const moreButton = await reviewEl.$('.section-expand-review');
if (moreButton) {
await moreButton.click();
await sleep(1000);
}
const review = await page.evaluate((reviewEl) => {
const $review = $(reviewEl);
return {
name: $review.find('.section-review-title').text().trim(),
text: $review.find('.section-review-text').text(),
stars: $review.find('.section-review-stars').attr('aria-label').trim(),
publishAt: $review.find('.section-review-publish-date').text().trim(),
likesCount: $review.find('.section-review-thumbs-up-count').text().trim(),
};
}, reviewEl);
placeDetail.reviews.push(review);
}
}
await Apify.pushData(placeDetail);
console.log(request.url, 'Done');
},
handleFailedRequestFunction: async ({ request }) => {
// This function is called when crawling of a request failed too many time
await Apify.pushData({
url: request.url,
succeeded: false,
errors: request.errorMessages,
});
},
});
const crawler = placesCrawler.setUpCrawler(launchPuppeteerOptions, requestQueue);
await crawler.run();
console.log('Done!');
});

91
src/places_crawler.js Normal file
View File

@ -0,0 +1,91 @@
const Apify = require('apify');
const { sleep } = Apify.utils;
const infiniteScroll = require('./infinite_scroll');
const { injectJQuery } = Apify.utils.puppeteer;
const { MAX_PAGE_RETRIES, DEFAULT_TIMEOUT } = require('./consts');
/**
* Method to set up crawler to get all place details and save them to default dataset
* @param launchPuppeteerOptions
* @param requestQueue
* @return {Apify.PuppeteerCrawler}
*/
const setUpCrawler = (launchPuppeteerOptions, requestQueue) => {
return new Apify.PuppeteerCrawler({
launchPuppeteerOptions,
requestQueue,
maxRequestRetries: MAX_PAGE_RETRIES,
retireInstanceAfterRequestCount: 10,
handlePageTimeoutSecs: 600,
maxConcurrency: 1,
gotoFunction: async ({ request, page }) => {
await page._client.send('Emulation.clearDeviceMetricsOverride');
await page.goto(request.url, { timeout: 60000 });
},
handlePageFunction: async ({ request, page }) => {
const { label } = request.userData;
console.log(`Open ${request.url} with label: ${label}`);
// Get data from review
await injectJQuery(page);
await page.waitForSelector('h1.section-hero-header-title', { timeout: DEFAULT_TIMEOUT });
const placeDetail = await page.evaluate(() => {
return {
title: $('h1.section-hero-header-title').text().trim(),
totalScore: $('span.section-star-display').eq(0).text().trim(),
categoryName: $('[jsaction="pane.rating.category"]').text().trim(),
address: $('[data-section-id="ad"] .widget-pane-link').text().trim(),
plusCode: $('[data-section-id="ol"] .widget-pane-link').text().trim(),
};
});
placeDetail.url = request.url;
placeDetail.reviews = [];
if (placeDetail.totalScore) {
placeDetail.reviewsCount = await page.evaluate(() => {
const numberReviewsText = $('button.section-reviewchart-numreviews').text().trim();
return (numberReviewsText) ? numberReviewsText.match(/\d+/)[0] : null;
});
// Get all reviews
await page.click('button.section-reviewchart-numreviews');
await page.waitForSelector('.section-star-display', { timeout: DEFAULT_TIMEOUT });
await sleep(2000);
// Sort reviews by newest
await page.click('.section-tab-info-stats-button-flex');
await page.waitForSelector('.context-menu-entry[data-index="1"]');
await page.click('.context-menu-entry[data-index="1"]');
await infiniteScroll(page, 99999999999, '.section-scrollbox.section-listbox');
const reviewEls = await page.$$('div.section-review');
for (const reviewEl of reviewEls) {
const moreButton = await reviewEl.$('.section-expand-review');
if (moreButton) {
await moreButton.click();
await sleep(1000);
}
const review = await page.evaluate((reviewEl) => {
const $review = $(reviewEl);
return {
name: $review.find('.section-review-title').text().trim(),
text: $review.find('.section-review-text').text(),
stars: $review.find('.section-review-stars').attr('aria-label').trim(),
publishAt: $review.find('.section-review-publish-date').text().trim(),
likesCount: $review.find('.section-review-thumbs-up-count').text().trim(),
};
}, reviewEl);
placeDetail.reviews.push(review);
}
}
await Apify.pushData(placeDetail);
console.log(request.url, 'Done');
},
handleFailedRequestFunction: async ({ request }) => {
// This function is called when crawling of a request failed too many time
await Apify.pushData({
url: request.url,
succeeded: false,
errors: request.errorMessages,
});
},
});
};
module.exports = { setUpCrawler };