Several updates, readme updated, new selectores fixed

This commit is contained in:
drobnikj 2019-01-08 09:35:01 +01:00
parent 76fdde6dbd
commit 9f1ae7fca0
4 changed files with 3878 additions and 162 deletions

3726
README.md

File diff suppressed because it is too large Load Diff

View File

@ -9,6 +9,7 @@ const waitForGoogleMapLoader = (page) => page.waitFor(() => !document.querySelec
const enqueueAllUrlsFromPagination = async (page, requestQueue, paginationFrom, maxPlacesPerCrawl) => {
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 });
@ -21,6 +22,7 @@ const enqueueAllUrlsFromPagination = async (page, requestQueue, paginationFrom,
await link.click();
await waitForGoogleMapLoader(page);
await page.waitForSelector('.section-back-to-list-button', { timeout: DEFAULT_TIMEOUT });
// After redirection to detail page, save the URL to Request queue to process it later
const url = page.url();
await requestQueue.addRequest({ url, userData: { label: 'detail' } });
log.info(`Added to queue ${url}`);
@ -28,20 +30,23 @@ const enqueueAllUrlsFromPagination = async (page, requestQueue, paginationFrom,
log.info(`Reach max places per crawl ${maxPlacesPerCrawl}, stopped enqueuing new places.`);
break;
}
await page.click('.section-back-to-list-button');
}
};
/**
* Crawler add all place detail from listing to queue
* Adds all places from listing to queue
* @param page
* @param searchString
* @param launchPuppeteerOptions
* @param requestQueue
* @param listingPagination
* @param maxPlacesPerCrawl
*/
const enqueueAllPlaceDetailsCrawler = async (page, searchString, launchPuppeteerOptions, requestQueue, listingPagination, maxPlacesPerCrawl) => {
const enqueueAllPlaceDetails = async (page, searchString, requestQueue, maxPlacesPerCrawl) => {
// Save state of listing pagination
// NOTE: If pageFunction failed crawler skipped already scraped pagination
const listingPagination = await Apify.getValue(LISTING_PAGINATION_KEY) || {};
await page.type('#searchboxinput', searchString);
await sleep(5000);
await page.click('#searchbox-searchbutton');
@ -50,8 +55,9 @@ const enqueueAllPlaceDetailsCrawler = async (page, searchString, launchPuppeteer
try {
await page.waitForSelector('h1.section-hero-header-title');
} catch (e) {
// It can happen, doesn't matter :)
// It can happen, if there are listing, not just detail page
}
// In case there is no listing, put just detail page to queue
const maybeDetailPlace = await page.$('h1.section-hero-header-title');
if (maybeDetailPlace) {
@ -59,6 +65,8 @@ const enqueueAllPlaceDetailsCrawler = async (page, searchString, launchPuppeteer
await requestQueue.addRequest({ url, userData: { label: 'detail' } });
return;
}
// In case there is listing, go through all details, limits with maxPlacesPerCrawl
const nextButtonSelector = '[jsaction="pane.paginationSection.nextPage"]';
while (true) {
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
@ -71,7 +79,8 @@ const enqueueAllPlaceDetailsCrawler = async (page, searchString, launchPuppeteer
} else {
log.debug(`Added links from pagination ${from} - ${to}`);
await enqueueAllUrlsFromPagination(page, requestQueue, from, maxPlacesPerCrawl);
listingPagination = { from, to };
listingPagination.from = from;
listingPagination.to = to;
await Apify.setValue(LISTING_PAGINATION_KEY, listingPagination);
}
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
@ -87,6 +96,9 @@ const enqueueAllPlaceDetailsCrawler = async (page, searchString, launchPuppeteer
await waitForGoogleMapLoader(page);
}
}
listingPagination.isFinish = true;
await Apify.setValue(LISTING_PAGINATION_KEY, listingPagination);
};
module.exports = { run: enqueueAllPlaceDetailsCrawler };
module.exports = { enqueueAllPlaceDetails };

View File

@ -26,7 +26,7 @@ Apify.main(async () => {
const launchPuppeteerOptions = {};
if (proxyConfig) Object.assign(launchPuppeteerOptions, proxyConfig);
// Scrape all place detail links
// Create and run crawler
const crawler = placesCrawler.setUpCrawler(launchPuppeteerOptions, requestQueue, maxCrawledPlaces);
await crawler.run();

View File

@ -1,11 +1,143 @@
const Apify = require('apify');
const { sleep, log } = Apify.utils;
const infiniteScroll = require('./infinite_scroll');
const { injectJQuery } = Apify.utils.puppeteer;
const { MAX_PAGE_RETRIES, DEFAULT_TIMEOUT, LISTING_PAGINATION_KEY } = require('./consts');
const enqueueAllPlaceDetailsCrawler = require('./enqueue_places_crawler');
const infiniteScroll = require('./infinite_scroll');
const { MAX_PAGE_RETRIES, DEFAULT_TIMEOUT } = require('./consts');
const { enqueueAllPlaceDetails } = require('./enqueue_places_crawler');
/**
* This is the worst part - parsing data from place detail
* @param page
*/
const extractPlaceDetail = async (page) => {
// Extracts basic information
const titleSel = 'h1.section-hero-header-title';
await page.waitForSelector(titleSel, { timeout: DEFAULT_TIMEOUT });
const detail = 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: $('[data-section-id="ol"] .widget-pane-link').text().trim(),
};
});
// Extracty histogram for popular times
const histogramSel = '.section-popular-times';
if (await page.$(histogramSel)) {
detail.popularTimesHistogram = await page.evaluate(() => {
const graphs = {};
const days = ['Su', 'Mo', 'Tu', 'We', 'Th', 'Fr', 'Sa'];
// Extract all days graphs
$('.section-popular-times-graph').each(function(i) {
const day = days[i];
graphs[day] = [];
let graphStartFromHour;
// Finds where x axis starts
$(this).find('.section-popular-times-label').each(function(labelIndex) {
if (graphStartFromHour) return;
const hourText = $(this).text().trim();
graphStartFromHour = hourText.includes('p')
? 12 + (parseInt(hourText) - labelIndex)
: parseInt(hourText) - labelIndex;
});
// Finds values from y axis
$(this).find('.section-popular-times-bar').each(function (barIndex) {
const occupancyMatch = $(this).attr('aria-label').match(/\d+\s+?%/);
if (occupancyMatch) {
const maybeHour = graphStartFromHour + barIndex;
graphs[day].push({
hour: maybeHour > 24 ? maybeHour - 24 : maybeHour,
occupancyPercent: parseInt(occupancyMatch[0]),
});
}
});
});
return graphs;
});
}
// Extracts reviews
detail.reviews = [];
const reviewsButtonSel = 'button[jsaction="pane.reviewChart.moreReviews"]';
if (detail.totalScore) {
detail.totalScore = parseFloat(detail.totalScore.replace(',', '.'));
detail.reviewsCount = await page.evaluate((selector) => {
const numberReviewsText = $(selector).text().trim();
return (numberReviewsText) ? numberReviewsText.match(/\d+/)[0] : null;
}, reviewsButtonSel);
// If we find consent dialog, close it!
if (await page.$('.widget-consent-dialog')) {
await page.click('.widget-consent-dialog .widget-consent-button-later');
}
// Get all reviews
await page.waitForSelector(reviewsButtonSel);
await page.click(reviewsButtonSel);
await page.waitForSelector('.section-star-display', { timeout: DEFAULT_TIMEOUT });
await sleep(5000);
// Sort reviews by newest, one click sometimes didn't work :)
try {
const sortButtonEl = '.section-tab-info-stats-button-flex';
for (let i = 0; i < 3; i++) {
await page.click(sortButtonEl);
await sleep(1000);
}
await page.click('.context-menu-entry[data-index="1"]');
} catch (err) {
// It can happen, it is not big issue :)
log.debug('Cannot select reviews by newest!');
}
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(2000);
}
const review = await page.evaluate((reviewEl) => {
const $review = $(reviewEl);
const reviewData = {
name: $review.find('.section-review-title').text().trim(),
text: $review.find('.section-review-review-content .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(),
};
const $response = $review.find('.section-review-owner-response');
if ($response) {
reviewData.responseFromOwnerText = $response.find('.section-review-text').text().trim();
}
return reviewData;
}, reviewEl);
detail.reviews.push(review);
}
await page.click('button.section-header-back-button');
}
// Extracts place images
await page.waitForSelector(titleSel, { timeout: DEFAULT_TIMEOUT });
const imagesButtonSel = '[jsaction="pane.imagepack.button"]';
if (await page.$(imagesButtonSel)) {
await page.click(imagesButtonSel);
await infiniteScroll(page, 99999999999, '.section-scrollbox.section-listbox');
detail.imageUrls = await page.evaluate(() => {
const urls = [];
$('.gallery-image-high-res').each(function () {
const urlMatch = $(this).attr('style').match(/url\("(.*)"\)/);
if (!urlMatch) return;
let imageUrl = urlMatch[1];
if (imageUrl[0] === '/') imageUrl = `https:${imageUrl}`;
urls.push(imageUrl);
});
return urls;
});
}
return detail;
};
/**
* Method to set up crawler to get all place details and save them to default dataset
* @param launchPuppeteerOptions
@ -19,13 +151,15 @@ const setUpCrawler = (launchPuppeteerOptions, requestQueue, maxCrawledPlaces) =>
requestQueue,
maxRequestRetries: MAX_PAGE_RETRIES,
retireInstanceAfterRequestCount: 10,
handlePageTimeoutSecs: 2 * 3600, // Two hours because startUrl crawler
maxOpenPagesPerInstance: 1, // Because startUrl crawler crashes if we mixed tabs with details scraping
// maxConcurrency: 1,
handlePageTimeoutSecs: 15 * 60, // 15 min because startUrl enqueueing
maxOpenPagesPerInstance: 1, // Because startUrl enqueueing crashes if we mixed tabs with details scraping
};
if (maxCrawledPlaces) {
crawlerOpts.maxRequestsPerCrawl = maxCrawledPlaces + 1; // The first one is startUrl
}
if (!Apify.isAtHome()) {
crawlerOpts.maxConcurrency = 2;
}
return new Apify.PuppeteerCrawler({
...crawlerOpts,
gotoFunction: async ({ request, page }) => {
@ -34,142 +168,22 @@ const setUpCrawler = (launchPuppeteerOptions, requestQueue, maxCrawledPlaces) =>
},
handlePageFunction: async ({ request, page }) => {
const { label, searchString } = request.userData;
log.info(`Open ${request.url} with label: ${label}`);
await injectJQuery(page);
if (label === 'startUrl') {
// enqueue all places
log.info(`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, maxCrawledPlaces);
listingPagination.isFinish = true;
await Apify.setValue(LISTING_PAGINATION_KEY, listingPagination);
log.info(`Start enqueuing places details for search: ${searchString}`);
await enqueueAllPlaceDetails(page, searchString, requestQueue, maxCrawledPlaces);
log.info('Enqueuing places finished!');
} else {
// Timeout because timeout for handle page is 2 hours
setTimeout(() => {
throw new Error('HandlePagefunction timed out!');
}, 600000);
// Get data from review
const titleSel = 'h1.section-hero-header-title';
await page.waitForSelector(titleSel, { 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(),
};
});
// Get data for place and save it to dataset
log.info(`Extracting details from place url ${request.url}`);
const placeDetail = await extractPlaceDetail(page);
placeDetail.url = request.url;
const histogramSel = '.section-popular-times';
if (await page.$(histogramSel)) {
placeDetail.popularTimesHistogram = await page.evaluate(() => {
const graphs = {};
const days = ['Su', 'Mo', 'Tu', 'We', 'Th', 'Fr', 'Sa'];
// Days graphs
$('.section-popular-times-graph').each(function(i) {
const day = days[i];
graphs[day] = [];
let graphStartFromHour;
$(this).find('.section-popular-times-label').each(function(labelIndex) {
if (graphStartFromHour) return;
const hourText = $(this).text().trim();
graphStartFromHour = hourText.includes('p')
? 12 + (parseInt(hourText) - labelIndex)
: parseInt(hourText) - labelIndex;
});
$(this).find('.section-popular-times-bar').each(function (barIndex) {
const occupancy = $(this).attr('aria-label').match(/\d+\s{1,}%/)[0];
const maybeHour = graphStartFromHour + barIndex;
graphs[day].push({
hour: maybeHour > 24 ? maybeHour - 24 : maybeHour,
occupancy,
});
});
});
return graphs;
});
}
placeDetail.reviews = [];
const reviewsButtonSel = 'button[jsaction="pane.reviewChart.moreReviews"]';
if (placeDetail.totalScore) {
placeDetail.reviewsCount = await page.evaluate((selector) => {
const numberReviewsText = $(selector).text().trim();
return (numberReviewsText) ? numberReviewsText.match(/\d+/)[0] : null;
}, reviewsButtonSel);
// If we find consent dialog, close it!
if (await page.$('.widget-consent-dialog')) {
await page.click('.widget-consent-dialog .widget-consent-button-later');
}
// Get all reviews
await page.waitForSelector(reviewsButtonSel);
await page.click(reviewsButtonSel);
await page.waitForSelector('.section-star-display', { timeout: DEFAULT_TIMEOUT });
await sleep(5000);
// Sort reviews by newest, one click sometimes didn't work :)
try {
const sortButtonEl = '.section-tab-info-stats-button-flex';
await page.click(sortButtonEl);
await sleep(1000);
await page.click(sortButtonEl);
await sleep(1000);
await page.click(sortButtonEl);
await sleep(5000);
await page.click('.context-menu-entry[data-index="1"]');
} catch (err) {
// It can happen, it is not big issue :)
log.debug('Cannot select reviews by newest!');
}
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(2000);
}
const review = await page.evaluate((reviewEl) => {
const $review = $(reviewEl);
const reviewData = {
name: $review.find('.section-review-title').text().trim(),
text: $review.find('.section-review-review-content .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(),
};
const $response = $review.find('.section-review-owner-response');
if ($response) {
reviewData.responseFromOwnerText = $response.find('.section-review-text').text().trim();
}
return reviewData;
}, reviewEl);
placeDetail.reviews.push(review);
}
await page.click('button.section-header-back-button');
}
await page.waitForSelector(titleSel, { timeout: DEFAULT_TIMEOUT });
const imagesButtonSel = '[jsaction="pane.imagepack.button"]';
console.log(imagesButtonSel);
if (await page.$(imagesButtonSel)) {
await page.click(imagesButtonSel);
await infiniteScroll(page, 99999999999, '.section-scrollbox.section-listbox');
placeDetail.imageUrls = await page.evaluate(() => {
const urls = [];
$('.gallery-image-high-res').each(function () {
const urlMatch = $(this).attr('style').match(/url\("(.*)"\)/);
if (!urlMatch) return;
let imageUrl = urlMatch[1];
if (imageUrl[0] === '/') imageUrl = `https:${imageUrl}`;
urls.push(imageUrl);
});
return urls;
});
}
await Apify.pushData(placeDetail);
log.info(`Finished place url ${request.url}`);
}
log.info('Finished', request.url);
},
handleFailedRequestFunction: async ({ request }) => {
// This function is called when crawling of a request failed too many time