mirror of
https://github.com/davidjohnbarton/crawler-google-places.git
synced 2025-12-12 16:38:45 +00:00
Moved crawler for enqueue places to main crawler
This commit is contained in:
parent
e02ccfe8c5
commit
0689be833c
11
apify.json
11
apify.json
|
|
@ -1,11 +1,6 @@
|
||||||
{
|
{
|
||||||
"name": "crawler-google-places",
|
"name": "crawler-google-places",
|
||||||
"actId": null,
|
"version": "0.1",
|
||||||
"version": {
|
"buildTag": "latest",
|
||||||
"versionNumber": "0.1",
|
"env": null
|
||||||
"buildTag": "latest",
|
|
||||||
"envVars": [],
|
|
||||||
"sourceType": "TARBALL",
|
|
||||||
"tarballUrl": null
|
|
||||||
}
|
|
||||||
}
|
}
|
||||||
|
|
|
||||||
|
|
@ -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');
|
||||||
|
|
@ -36,70 +38,53 @@ 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;
|
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
|
||||||
try {
|
try {
|
||||||
browser = await Apify.launchPuppeteer(launchPuppeteerOptions);
|
await page.waitForSelector('h1.section-hero-header-title');
|
||||||
const page = await browser.newPage();
|
} catch (e) {
|
||||||
await page._client.send('Emulation.clearDeviceMetricsOverride');
|
// It can happen, doesn't matter
|
||||||
await page.goto(startUrl);
|
}
|
||||||
await injectJQuery(page);
|
const maybeDetailPlace = await page.$('h1.section-hero-header-title');
|
||||||
await page.type('#searchboxinput', searchString);
|
if (maybeDetailPlace) {
|
||||||
await sleep(5000);
|
const url = page.url();
|
||||||
await page.click('#searchbox-searchbutton');
|
await requestQueue.addRequest({ url, userData: { label: 'detail' } });
|
||||||
await sleep(5000);
|
return;
|
||||||
await waitForGoogleMapLoader(page);
|
}
|
||||||
// In case there is no listing, put just detail page to queue
|
const nextButtonSelector = '[jsaction="pane.paginationSection.nextPage"]';
|
||||||
try {
|
while (true) {
|
||||||
await page.waitForSelector('h1.section-hero-header-title');
|
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
|
||||||
} catch (e) {
|
const paginationText = await page.$eval('.n7lv7yjyC35__right', (el) => el.innerText);
|
||||||
// It can happen, doesn't matter
|
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} - ${to}`);
|
||||||
|
await enqueueAllUrlsFromPagination(page, requestQueue);
|
||||||
|
listingPagination = { from, to };
|
||||||
|
await Apify.setValue(LISTING_PAGINATION_KEY, listingPagination);
|
||||||
}
|
}
|
||||||
const maybeDetailPlace = await page.$('h1.section-hero-header-title');
|
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
|
||||||
if (maybeDetailPlace) {
|
const isNextPaginationDisabled = await page.evaluate((nextButtonSelector) => {
|
||||||
const url = page.url();
|
return !!$(nextButtonSelector)
|
||||||
await requestQueue.addRequest({ url, userData: { label: 'detail' } });
|
.attr('disabled');
|
||||||
return;
|
}, nextButtonSelector);
|
||||||
|
const noResultsEl = await page.$('.section-no-result-title');
|
||||||
|
if (isNextPaginationDisabled || noResultsEl) {
|
||||||
|
break;
|
||||||
|
} else {
|
||||||
|
// NOTE: puppeteer API click() didn't work :(
|
||||||
|
await page.evaluate((sel) => $(sel)
|
||||||
|
.click(), nextButtonSelector);
|
||||||
|
await waitForGoogleMapLoader(page);
|
||||||
}
|
}
|
||||||
const nextButtonSelector = '#section-pagination-button-next';
|
|
||||||
while (true) {
|
|
||||||
await page.waitForSelector(nextButtonSelector, { timeout: DEFAULT_TIMEOUT });
|
|
||||||
const paginationText = await page.$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();
|
|
||||||
}
|
}
|
||||||
};
|
};
|
||||||
|
|
||||||
|
|
|
||||||
14
src/main.js
14
src/main.js
|
|
@ -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();
|
||||||
|
|
|
||||||
|
|
@ -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
|
||||||
|
|
@ -18,73 +19,89 @@ 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);
|
||||||
await page.waitForSelector('h1.section-hero-header-title', { timeout: DEFAULT_TIMEOUT });
|
if (label === 'startUrl') {
|
||||||
const placeDetail = await page.evaluate(() => {
|
// enqueue all places
|
||||||
return {
|
console.log(`Start enqueuing place details for search: ${searchString}`);
|
||||||
title: $('h1.section-hero-header-title').text().trim(),
|
// Store state of listing pagination
|
||||||
totalScore: $('span.section-star-display').eq(0).text().trim(),
|
// NOTE: Ensured - If pageFunction failed crawler skipped already scraped pagination
|
||||||
categoryName: $('[jsaction="pane.rating.category"]').text().trim(),
|
const listingPagination = await Apify.getValue(LISTING_PAGINATION_KEY) || {};
|
||||||
address: $('[data-section-id="ad"] .widget-pane-link').text().trim(),
|
await enqueueAllPlaceDetailsCrawler.run(page, searchString, launchPuppeteerOptions, requestQueue, listingPagination);
|
||||||
plusCode: $('[data-section-id="ol"] .widget-pane-link').text().trim(),
|
listingPagination.isFinish = true;
|
||||||
};
|
await Apify.setValue(LISTING_PAGINATION_KEY, listingPagination);
|
||||||
});
|
} else {
|
||||||
placeDetail.url = request.url;
|
// Timeout because timeout for handle page is 2 hours
|
||||||
placeDetail.reviews = [];
|
setTimeout(() => {
|
||||||
if (placeDetail.totalScore) {
|
throw new Error('HandlePagefunction timed out!');
|
||||||
placeDetail.reviewsCount = await page.evaluate(() => {
|
}, 600000);
|
||||||
const numberReviewsText = $('button.section-reviewchart-numreviews').text().trim();
|
// Get data from review
|
||||||
return (numberReviewsText) ? numberReviewsText.match(/\d+/)[0] : null;
|
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(),
|
||||||
|
};
|
||||||
});
|
});
|
||||||
// Get all reviews
|
placeDetail.url = request.url;
|
||||||
await page.click('button.section-reviewchart-numreviews');
|
placeDetail.reviews = [];
|
||||||
await page.waitForSelector('.section-star-display', { timeout: DEFAULT_TIMEOUT });
|
if (placeDetail.totalScore) {
|
||||||
await sleep(5000);
|
placeDetail.reviewsCount = await page.evaluate(() => {
|
||||||
// Sort reviews by newest, one click sometimes didn't work :)
|
const numberReviewsText = $('button.section-reviewchart-numreviews').text().trim();
|
||||||
await page.click('.section-tab-info-stats-button-flex');
|
return (numberReviewsText) ? numberReviewsText.match(/\d+/)[0] : null;
|
||||||
await sleep(1000);
|
});
|
||||||
await page.click('.section-tab-info-stats-button-flex');
|
// Get all reviews
|
||||||
await sleep(1000);
|
await page.click('button.section-reviewchart-numreviews');
|
||||||
await page.click('.section-tab-info-stats-button-flex');
|
await page.waitForSelector('.section-star-display', { timeout: DEFAULT_TIMEOUT });
|
||||||
await sleep(5000);
|
await sleep(5000);
|
||||||
await page.click('.context-menu-entry[data-index="1"]');
|
// Sort reviews by newest, one click sometimes didn't work :)
|
||||||
await infiniteScroll(page, 99999999999, '.section-scrollbox.section-listbox');
|
await page.click('.section-tab-info-stats-button-flex');
|
||||||
const reviewEls = await page.$$('div.section-review');
|
await sleep(1000);
|
||||||
for (const reviewEl of reviewEls) {
|
await page.click('.section-tab-info-stats-button-flex');
|
||||||
const moreButton = await reviewEl.$('.section-expand-review');
|
await sleep(1000);
|
||||||
if (moreButton) {
|
await page.click('.section-tab-info-stats-button-flex');
|
||||||
await moreButton.click();
|
await sleep(5000);
|
||||||
await sleep(2000);
|
await page.click('.context-menu-entry[data-index="1"]');
|
||||||
}
|
await infiniteScroll(page, 99999999999, '.section-scrollbox.section-listbox');
|
||||||
const review = await page.evaluate((reviewEl) => {
|
const reviewEls = await page.$$('div.section-review');
|
||||||
const $review = $(reviewEl);
|
for (const reviewEl of reviewEls) {
|
||||||
const reviewData = {
|
const moreButton = await reviewEl.$('.section-expand-review');
|
||||||
name: $review.find('.section-review-title').text().trim(),
|
if (moreButton) {
|
||||||
text: $review.find('.section-review-review-content .section-review-text').text(),
|
await moreButton.click();
|
||||||
stars: $review.find('.section-review-stars').attr('aria-label').trim(),
|
await sleep(2000);
|
||||||
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;
|
const review = await page.evaluate((reviewEl) => {
|
||||||
}, reviewEl);
|
const $review = $(reviewEl);
|
||||||
placeDetail.reviews.push(review);
|
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 Apify.pushData(placeDetail);
|
||||||
}
|
}
|
||||||
await Apify.pushData(placeDetail);
|
|
||||||
|
|
||||||
console.log(request.url, 'Done');
|
console.log(request.url, 'Done');
|
||||||
},
|
},
|
||||||
|
|
|
||||||
Loading…
Reference in New Issue
Block a user