mirror of
https://github.com/davidjohnbarton/crawler-google-places.git
synced 2025-12-12 16:38:45 +00:00
Version 0.1
This commit is contained in:
JakubDrobnik
parent
86a82b1612
commit
1b963d9ac5
|
|
@ -4,7 +4,8 @@
|
|||
"description": "",
|
||||
"main": "main.js",
|
||||
"scripts": {
|
||||
"test": "echo \"Error: no test specified\" && exit 1"
|
||||
"test": "echo \"Error: no test specified\" && exit 1",
|
||||
"start": "node ./src/main.js"
|
||||
},
|
||||
"repository": {
|
||||
"type": "git",
|
||||
|
|
@ -20,5 +21,8 @@
|
|||
"bugs": {
|
||||
"url": "https://github.com/drobnikj/crawler-google-places/issues"
|
||||
},
|
||||
"homepage": "https://github.com/drobnikj/crawler-google-places#readme"
|
||||
"homepage": "https://github.com/drobnikj/crawler-google-places#readme",
|
||||
"dependencies": {
|
||||
"apify": "^0.8.18"
|
||||
}
|
||||
}
|
||||
|
|
|
|||
225
src/main.js
225
src/main.js
|
|
@ -1 +1,224 @@
|
|||
tets
|
||||
/**
|
||||
* Run the following example to perform a recursive crawl of a website using Puppeteer.
|
||||
*/
|
||||
|
||||
const Apify = require('apify');
|
||||
|
||||
const { sleep } = Apify.utils;
|
||||
const { injectJQuery } = Apify.utils.puppeteer;
|
||||
|
||||
const sleepPromised = ms => new Promise(resolve => setTimeout(resolve, ms));
|
||||
|
||||
const logError = (msg, e) => {
|
||||
console.log(`ERROR: ${msg}`);
|
||||
console.error(e);
|
||||
};
|
||||
const logInfo = (msg) => console.log(`INFO: ${msg}`);
|
||||
const logDebug = (msg) => console.log(`DEBUG: ${msg}`);
|
||||
|
||||
/**
|
||||
* Method scrolls page to xpos, ypos.
|
||||
*/
|
||||
const scrollTo = (page, xpos, ypos) => page.evaluate((x, y) => window.scrollTo(x, y), xpos, ypos);
|
||||
|
||||
/**
|
||||
* Method returns info about page scroll
|
||||
*/
|
||||
const getPageScrollInfo = page => page.evaluate(() => {
|
||||
return {
|
||||
scrollHeight: document.documentElement.scrollHeight,
|
||||
scrollTop: document.documentElement.scrollTop,
|
||||
clientHeight: document.documentElement.clientHeight,
|
||||
};
|
||||
});
|
||||
|
||||
/**
|
||||
* Scroll to down page until infinite scroll ends or reaches maxHeight
|
||||
* @param page - instance of crawled page
|
||||
* @param maxHeight - max height of document to scroll
|
||||
* @return {Promise.<void>}
|
||||
*/
|
||||
const infiniteScroll = async (page, maxHeight) => {
|
||||
const maybeResourceTypesInfiniteScroll = ['xhr', 'fetch', 'websocket', 'other'];
|
||||
const stringifyScrollInfo = (scrollInfo) => {
|
||||
return `scrollTop=${scrollInfo.scrollTop}, ` +
|
||||
`clientHeight=${scrollInfo.clientHeight}, ` +
|
||||
`scrollHeight=${scrollInfo.scrollHeight}, ` +
|
||||
`maxHeight=${maxHeight}`;
|
||||
};
|
||||
const defaultScrollDelay = 500;
|
||||
|
||||
// Catch and count all pages request for resources
|
||||
const resourcesStats = {
|
||||
requested: 0,
|
||||
finished: 0,
|
||||
failed: 0,
|
||||
forgotten: 0,
|
||||
};
|
||||
const pendingRequests = {};
|
||||
page.on('request', (msg) => {
|
||||
if (maybeResourceTypesInfiniteScroll.includes(msg.resourceType)) {
|
||||
pendingRequests[msg._requestId] = Date.now();
|
||||
resourcesStats.requested++;
|
||||
}
|
||||
});
|
||||
page.on('requestfailed', (msg) => {
|
||||
if (maybeResourceTypesInfiniteScroll.includes(msg.resourceType)) {
|
||||
if (pendingRequests[msg._requestId]) {
|
||||
delete pendingRequests[msg._requestId];
|
||||
resourcesStats.failed++;
|
||||
}
|
||||
|
||||
}
|
||||
});
|
||||
page.on('requestfinished', (msg) => {
|
||||
if (maybeResourceTypesInfiniteScroll.includes(msg.resourceType)) {
|
||||
if (pendingRequests[msg._requestId]) {
|
||||
delete pendingRequests[msg._requestId];
|
||||
resourcesStats.finished++;
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
try {
|
||||
let scrollInfo = await getPageScrollInfo(page);
|
||||
logInfo(`Infinite scroll started (${stringifyScrollInfo(scrollInfo)}).`);
|
||||
|
||||
while (true) {
|
||||
scrollInfo = await getPageScrollInfo(page);
|
||||
|
||||
// Forget pending resources that didn't finish loading in time
|
||||
const now = Date.now();
|
||||
const timeout = 30000; // TODO: use resourceTimeout
|
||||
Object.keys(pendingRequests)
|
||||
.forEach((requestId) => {
|
||||
if (pendingRequests[requestId] + timeout < now) {
|
||||
delete pendingRequests[requestId];
|
||||
resourcesStats.forgotten++;
|
||||
}
|
||||
});
|
||||
|
||||
logDebug(`Infinite scroll stats (${stringifyScrollInfo(scrollInfo)} resourcesStats=${JSON.stringify(resourcesStats)}).`);
|
||||
|
||||
const pendingRequestsCount = resourcesStats.requested - (resourcesStats.finished + resourcesStats.failed + resourcesStats.forgotten);
|
||||
if (pendingRequestsCount === 0) {
|
||||
// If the page is scrolled to the very bottom or beyond maximum height, we are done
|
||||
if (scrollInfo.scrollTop + scrollInfo.clientHeight >= Math.min(scrollInfo.scrollHeight, maxHeight)) break;
|
||||
// Otherwise we try to scroll down
|
||||
await scrollTo(page, 0, scrollInfo.scrollHeight);
|
||||
}
|
||||
|
||||
await sleepPromised(defaultScrollDelay);
|
||||
}
|
||||
// Scroll back up, otherwise the screenshot of the browser would only show the bottom of
|
||||
// the page
|
||||
await scrollTo(page, 0, 0);
|
||||
|
||||
logInfo(`Infinite scroll finished (${stringifyScrollInfo(scrollInfo)} resourcesStats=${JSON.stringify(resourcesStats)})`);
|
||||
} catch (err) {
|
||||
logError('An exception thrown in infiniteScroll()', err);
|
||||
}
|
||||
};
|
||||
|
||||
const enqueueAllUrlsFromPagination = async (page, requestQueue) => {
|
||||
const detailLinks = [];
|
||||
let results = await page.$$('.section-result');
|
||||
const resultsCount = results.length;
|
||||
console.log('Titles ', results.length);
|
||||
for (let resultIndex = 0; resultIndex < resultsCount; resultIndex++) {
|
||||
// Need to get results again, pupptr lost context..
|
||||
results = await page.$$('.section-result');
|
||||
const link = await results[resultIndex].$('h3');
|
||||
await link.click();
|
||||
await page.waitForSelector('.section-back-to-list-button');
|
||||
const url = page.url();
|
||||
console.log(url);
|
||||
await requestQueue.addRequest({ url, userData: { label: 'detail' } });
|
||||
await page.click('.section-back-to-list-button');
|
||||
await sleep(5000);
|
||||
}
|
||||
return detailLinks;
|
||||
};
|
||||
|
||||
Apify.main(async () => {
|
||||
const { searchString } = await Apify.getValue('INPUT');
|
||||
|
||||
if (!searchString) throw new Error('Attribute searchString missing in input.')
|
||||
|
||||
const requestQueue = await Apify.openRequestQueue();
|
||||
await requestQueue.addRequest({ url: 'https://www.google.com/maps/search/', userData: { label: 'startUrl' } });
|
||||
|
||||
const crawler = new Apify.PuppeteerCrawler({
|
||||
launchPuppeteerOptions: {
|
||||
useApifyProxy: true,
|
||||
},
|
||||
requestQueue,
|
||||
handlePageFunction: async ({ request, page }) => {
|
||||
const { label } = request.userData;
|
||||
if (label === 'startUrl') {
|
||||
// Enqueue all urls for place detail
|
||||
await page.type('#searchboxinput', 'Česká Spořitelna');
|
||||
await sleep(5000);
|
||||
await page.click('#searchbox-searchbutton');
|
||||
await sleep(5000);
|
||||
while(true) {
|
||||
await page.waitForSelector('#section-pagination-button-next');
|
||||
await enqueueAllUrlsFromPagination(page, requestQueue);
|
||||
const nextButton = await page.$('#section-pagination-button-next');
|
||||
const isNextPagination = (await nextButton.getProperty('disabled') === 'true');
|
||||
console.log('isNextPagination ', isNextPagination);
|
||||
if (isNextPagination) {
|
||||
break;
|
||||
} else {
|
||||
await nextButton.click();
|
||||
}
|
||||
await sleep(5000);
|
||||
}
|
||||
} else {
|
||||
await injectJQuery(page);
|
||||
await page.waitForSelector('h1.section-hero-header-title');
|
||||
// Get data from review
|
||||
const placeDetail = await page.evaluate(() => {
|
||||
return {
|
||||
title: $('h1.section-hero-header-title').text().trim(),
|
||||
totalScore: $('span.section-star-display').eq(0).text().trim(),
|
||||
reviewsCount: $('button.section-reviewchart-numreviews').text().trim().match(/\d+/)[0],
|
||||
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 = [];
|
||||
console.log(placeDetail)
|
||||
// Get all reviews
|
||||
await page.click('button.section-reviewchart-numreviews');
|
||||
await infiniteScroll(page, 99999999999);
|
||||
const reviewEls = await page.$$('div.section-review');
|
||||
for (const reviewEl of reviewEls) {
|
||||
const moreButton = await reviewEl.$('.section-expand-review');
|
||||
if (moreButton) {
|
||||
await moreButton.click();
|
||||
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'),
|
||||
publishAt: $review.find('.section-review-publish-date').text().trim(),
|
||||
likesCount: $review.find('.section-review-thumbs-up-count').text().trim(),
|
||||
};
|
||||
}, reviewEl);
|
||||
console.log(review);
|
||||
placeDetail.reviews.push(review);
|
||||
}
|
||||
await Apify.pushData(placeDetail);
|
||||
}
|
||||
},
|
||||
maxConcurrency: 1,
|
||||
});
|
||||
|
||||
await crawler.run();
|
||||
});
|
||||
|
|
|
|||
Loading…
Reference in New Issue
Block a user