new way of scraping

data/cities.csv

@@ -1,27 +1,4 @@
 rank,Latitude,Longitude,Name of City,Country,2021 Population,2020 Population,Growth,Population Difference,Population Change
-1,35.6828387,139.7594549,Tokyo,Japan,37339804,37393128,-0.0014,53324,declined
-2,28.6517178,77.2219388,Delhi,India,31181376,30290936,0.0294,890440,grew
-3,31.2322758,121.4692071,Shanghai,China,27795702,27058480,0.0272,737222,grew
-4,-23.5506507,-46.6333824,Sao Paulo,Brazil,22237472,22043028,0.0088,194444,grew
-5,19.4326296,-99.1331785,Mexico City,Mexico,21918936,21782378,0.0063,136558,grew
-6,23.7861979,90.4026151,Dhaka,Bangladesh,21741090,21005860,0.035,735230,grew
-7,30.0443879,31.2357257,Cairo,Egypt,21322750,20900604,0.0202,422146,grew
-8,39.906217,116.3912757,Beijing,China,20896820,20462610,0.0212,434210,grew
-9,19.0759899,72.8773928,Mumbai,India,20667656,20411274,0.0126,256382,grew
-10,34.6198813,135.490357,Osaka,Japan,19110616,19165340,-0.0029,54724,declined
-11,24.8546842,67.0207055,Karachi,Pakistan,16459472,16093786,0.0227,365686,grew
-12,29.5647398,106.5478767,Chongqing,China,16382376,15872179,0.0321,510197,grew
-13,41.0096334,28.9651646,Istanbul,Turkey,15415197,15190336,0.0148,224861,grew
-14,-34.6075682,-58.4370894,Buenos Aires,Argentina,15257673,15153729,0.0069,103944,grew
-15,22.5414185,88.3576912,Kolkata,India,14974073,14850066,0.0084,124007,grew
-16,-4.3217055,15.3125974,Kinshasa,DR Congo,14970460,14342439,0.0438,628021,grew
-17,6.4550575,3.3941795,Lagos,Nigeria,14862111,14368332,0.0344,493779,grew
-18,14.5907332,120.9809674,Manila,Philippines,14158573,13923452,0.0169,235121,grew
-19,39.0856735,117.1951073,Tianjin,China,13794450,13589078,0.0151,205372,grew
-20,23.1301964,113.2592945,Guangzhou,China,13635397,13301532,0.0251,333865,grew
-21,-22.9110137,-43.2093727,Rio de Janeiro,Brazil,13544462,13458075,0.0064,86387,grew
-22,31.5656822,74.3141829,Lahore,Pakistan,13095166,12642423,0.0358,452743,grew
-23,12.9767936,77.590082,Bangalore,India,12764935,12326532,0.0356,438403,grew
 24,55.7504461,37.6174943,Moscow,Russia,12593252,12537954,0.0044,55298,grew
 25,22.555454,114.0543297,Shenzhen,China,12591696,12356820,0.019,234876,grew
 26,13.0836939,80.270186,Chennai,India,11235018,10971108,0.0241,263910,grew

src/index.ts

@@ -9,65 +9,27 @@ import chrome, { ServiceBuilder } from 'selenium-webdriver/chrome';
 import * as fs from 'fs';
 import * as path from 'path';
 import { getCities } from './lib/cities';
-import { WebDriverUtils, saveContactInfoToCSV } from './lib/utils';
+import { WebDriverUtils, saveContactInfoToCSV, useExistingChrome, disableCookiesInChrome, useChrome } from './lib/utils';
 import * as UIActions from './lib/UIActions';
 import { randomUUID } from 'crypto';
-import os from 'os';
-import edge from 'selenium-webdriver/edge';
+
 
 /**
  * Function to visit TripAdvisor pages for each city
  */
 async function visitCityPages(): Promise<void> {
 
-    const userHomeDir = os.homedir(); // gets C:\Users\<YourName>
-    const driverPath = path.join(userHomeDir, 'Documents', 'edgedriver_win64', 'msedgedriver.exe');
-  
-    // Configure Edge service to use your custom driver path
-    const service = new edge.ServiceBuilder(driverPath);
-  
-    const options = new edge.Options();
-    options.addArguments('--inprivate');
-    options.addArguments('--start-maximized');
-  
-    let driver: WebDriver;
-    driver = await new Builder()
-    .forBrowser('MicrosoftEdge')
-    .setEdgeOptions(options)
-    .setEdgeService(service)
-    .build();
-
-    await driver.get('https://www.tripadvisor.com');
-    await WebDriverUtils.wait(5);
-
-    /*
-
-
-    // Connect to an existing Chrome browser running in debug mode on port 9222
-    const options = new chrome.Options();
-
-    // Set the debugger address to connect to the existing Chrome instance
-    options.debuggerAddress('localhost:9222');
-
-    // Create WebDriver instance that connects to the existing browser
-    const driver: WebDriver = await new Builder()
-        .forBrowser('chrome')
-        .setChromeOptions(options)
-        .build();
-    */
-
     const cities = getCities(path.join(__dirname, '../data/cities.csv'));
 
     console.log('Connecting to existing Chrome browser...');
 
+    const driver = await useExistingChrome();
+    if (!driver) return;
+
     // Visit each city's TripAdvisor page
     for (let i = 0; i < cities.length; i++) {
         const city = cities[i];
         console.log(`[${i + 1}/${cities.length}] Visiting TripAdvisor page for ${city}...`);
-        let originalWindow;
-        let cityTopWindow;
-        let attactionsWindow;
-        let museumWindow;
 
         try {
 
@@ -75,107 +37,65 @@ async function visitCityPages(): Promise<void> {
 
             console.log("Logo click")
             if (!await UIActions.gotoHome(driver)) throw `${city} failed`;
-            await WebDriverUtils.wait(5);
+            await WebDriverUtils.wait(driver);
 
             console.log("Exec Search")
-            if (!await UIActions.execSearch(driver, city)) throw `${city} failed`;
-            await WebDriverUtils.wait(5);
-
-            console.log("Click See all")
-            if (!await UIActions.clickSeeAll(driver)) {
-                if (!await UIActions.clickTourismLink(driver)) throw `${city} failed`;
-                if (!await UIActions.clickSeeAll(driver)) throw `${city} failed`;
-            }
-            await WebDriverUtils.wait(5);
-
-            console.log("Switch tab")
-            let windows = await driver.getAllWindowHandles();
-            // Switch to the newly opened window/tab
-            for (const handle of windows) {
-                if (handle !== originalWindow) {
-                    cityTopWindow = handle;
-                    await driver.switchTo().window(handle);
-                }
-            }
-
-            console.log("Click See all attractions")
-            if (!await UIActions.clickSeeAllAttractions(driver)) throw `${city} failed`;
-            await WebDriverUtils.wait(5);
-
-            console.log("Switch tab to Attraction")
-            windows = await driver.getAllWindowHandles();
-            // Switch to the newly opened window/tab
-            for (const handle of windows) {
-                if (handle !== originalWindow && handle !== cityTopWindow) {
-                    attactionsWindow = handle;
-                    await driver.switchTo().window(attactionsWindow);
-                }
-            }
-
-            // click museum
-            console.log("Click Museum link");
-            if (!await UIActions.clickMuseumsLink(driver)) throw `${city} failed`;
-            await WebDriverUtils.wait(5);
+            if (!await UIActions.execSearch(driver, `${city} museum`)) throw `${city} failed`;
+            await WebDriverUtils.wait(driver);
 
             let page = 1;
+            let museumNames: string[] = [];
+
             while (1) {
 
                 // get list of museums
                 console.log("Get list of museums");
-                const museumElms = await UIActions.getMusiums(driver);
-                await WebDriverUtils.wait(1);
-
-                for (const listItem of museumElms) {
-
-                    await listItem.click();
-                    await WebDriverUtils.wait(3);
-
-                    windows = await driver.getAllWindowHandles();
-                    for (const handle of windows) {
-                        if (handle !== originalWindow && handle !== cityTopWindow && handle !== attactionsWindow) {
-                            museumWindow = handle;
-                            await driver.switchTo().window(museumWindow);
-                        }
-                    }
-
-                    const { websiteUrl, email } = await UIActions.getWebsiteAndEmail(driver);
-
-                    console.log(`${websiteUrl} / ${email}`);
-                    saveContactInfoToCSV(city, { websiteUrl: websiteUrl, email: email }, path.join(__dirname, '../data/contact_info.csv'));
-
-                    museumWindow && await driver.switchTo().window(museumWindow);
-                    await driver.close();
-                    await WebDriverUtils.wait(1);
-
-                    attactionsWindow && await driver.switchTo().window(attactionsWindow);
-                    await WebDriverUtils.wait(1);
-
-                }
+                const tmpMuseumNames = await UIActions.getAttractionNames(driver);
+                await WebDriverUtils.wait(driver);
+                museumNames = [...museumNames, ...tmpMuseumNames]
 
                 page++;
 
                 if (page > 10) break;
 
                 UIActions.clickPagination(driver, page);
-                await WebDriverUtils.wait(5);
+                await WebDriverUtils.wait(driver);
 
             }
 
+            for (const museumName of museumNames) {
 
+                console.log("back to home");
+                if (!await UIActions.gotoHome(driver)) throw `${city} failed`;
+                await WebDriverUtils.wait(driver);
+
+                if (!await UIActions.typeSearch(driver, `${city} ${museumName}`)) throw `${city} failed`;
+                await WebDriverUtils.wait(driver);
+
+                if (!await UIActions.clickFirstAttractionLinkInForm(driver)) throw `${city} failed`;
+                await WebDriverUtils.wait(driver);
+
+                const { websiteUrl, email } = await UIActions.getWebsiteAndEmail(driver);
+
+                console.log(`${websiteUrl} / ${email}`);
+                saveContactInfoToCSV(city, { websiteUrl: websiteUrl, email: email }, path.join(__dirname, '../data/contact_info.csv'));
+
+            }
+
+            console.log(museumNames);
             await UIActions.closeAllTabsExceptFirst(driver);
 
 
             if (i < cities.length - 1) {
                 console.log(`Waiting for 5000 seconds before next city...`);
-                await WebDriverUtils.wait(5); // Wait 5000 seconds before next city
+                await WebDriverUtils.wait(); // Wait 5000 seconds before next city
             }
-        } catch (error) {
 
-            await UIActions.closeAllTabsExceptFirst(driver);
+        } catch (error) {
 
             // If the button is not found within the timeout, log and continue to the next city
             console.log(`No Museums button found for ${city}. Moving to next city after 5 seconds...`);
-            await WebDriverUtils.wait(5); // Wait 5 seconds before next city
+            await WebDriverUtils.wait(); // Wait 5 seconds before next city
         }
     }
 

src/lib/UIActions.ts

@@ -8,7 +8,7 @@ import { ContactInfo } from './types';
 
 
 
-export async function execSearch(driver: WebDriver, city: string): Promise<boolean> {
+export async function execSearch(driver: WebDriver, searchTerm: string): Promise<boolean> {
     try {
         // Find the search input field
         const searchSelector = 'input[name="q"][placeholder="Places to go, things to do, hotels..."]';
@@ -18,12 +18,12 @@ export async function execSearch(driver: WebDriver, city: string): Promise<boole
 
         // Clear any existing text and enter the city name
         await searchInput.clear();
-        await searchInput.sendKeys(city);
+        await searchInput.sendKeys(searchTerm);
 
         // Submit the search (press Enter)
-        await WebDriverUtils.wait(2);
+        await WebDriverUtils.wait(driver);
         await searchInput.sendKeys('\uE007'); // Unicode for Enter key
-        await WebDriverUtils.wait(5); // Wait 5 seconds before next city
+        await WebDriverUtils.wait(driver); // Wait 5 seconds before next city
 
         return true;
     } catch (e) {
@@ -31,6 +31,25 @@ export async function execSearch(driver: WebDriver, city: string): Promise<boole
     }
 }
 
+export async function typeSearch(driver: WebDriver, searchTerm: string): Promise<boolean> {
+    try {
+        // Find the search input field
+        const searchSelector = 'input[name="q"][placeholder="Places to go, things to do, hotels..."]';
+        await WebDriverUtils.waitForElement(driver, searchSelector, 10000);
+        console.log("Search box found");
+        const searchInput = await driver.findElement(By.css(searchSelector));
+
+        // Clear any existing text and enter the city name
+        await searchInput.clear();
+        await searchInput.sendKeys(searchTerm);
+
+        return true;
+    } catch (e) {
+        return false;
+    }
+}
+
+
 export async function clickSeeAll(driver: WebDriver): Promise<boolean> {
     try {
         const seeAllElement = await driver.wait(
@@ -49,6 +68,21 @@ export async function clickSeeAll(driver: WebDriver): Promise<boolean> {
     }
 }
 
+export async function getSeeAllUrl(driver: WebDriver): Promise<string> {
+    const xpath = `//h3[normalize-space(.)='Things to do']/ancestor::div[1]//a[starts-with(@href, '/Attractions')]`;
+
+    try {
+        const anchor = await driver.wait(until.elementLocated(By.xpath(xpath)), 5000);
+        const url = await anchor.getAttribute('href');
+        console.log('Found Attractions URL:', url);
+        return url;
+    } catch (err) {
+        console.warn('Could not find the Attractions link:', err);
+    }
+
+    return "";
+}
+
 export async function gotoHome(driver: WebDriver): Promise<boolean> {
     try {
         // Click on the Tripadvisor logo before searching for the city
@@ -65,7 +99,7 @@ export async function gotoHome(driver: WebDriver): Promise<boolean> {
 }
 
 export async function clickSeeAllAttractions(driver: WebDriver): Promise<boolean> {
-    const xpath = `//h2[starts-with(normalize-space(.), 'Top Attractions in')]/parent::*[1]//a[starts-with(@href, '/Attractions') and .//span[normalize-space(.)='See all']]`;
+    const xpath = `//h3[starts-with(normalize-space(.), 'Things to do')]/parent::*[1]//a[starts-with(@href, '/Attractions') and .//span[normalize-space(.)='See all']]`;
 
     try {
         const anchorElement = await driver.wait(until.elementLocated(By.xpath(xpath)), 5000);
@@ -82,6 +116,22 @@ export async function clickSeeAllAttractions(driver: WebDriver): Promise<boolean
     }
 }
 
+export async function getSeeAllAttractionsUrl(driver: WebDriver): Promise<string | null> {
+    const xpath = `//h3[starts-with(normalize-space(.), 'Top Attractions in')]/parent::*[1]//a[starts-with(@href, '/Attractions') and .//span[normalize-space(.)='See all']]`;
+
+    try {
+        const anchorElement = await driver.wait(until.elementLocated(By.xpath(xpath)), 5000);
+
+        await driver.wait(until.elementIsVisible(anchorElement), 5000);
+
+        const href = await anchorElement.getAttribute('href');
+        return href;
+    } catch (error) {
+        console.warn('Element not found or href not retrievable.', error);
+        return null;
+    }
+}
+
 export async function clickMuseumsLink(driver: WebDriver): Promise<boolean> {
     const xpath = `//a[.//*[normalize-space(.)='Museums']]`;
 
@@ -130,6 +180,49 @@ export async function getMusiums(driver: WebDriver): Promise<WebElement[]> {
     }
 }
 
+export async function getMuseumsLinks(driver: WebDriver): Promise<string[]> {
+    const xpath = `//div//section[.//a[starts-with(@href, '/Attraction')] and .//h3]//a[starts-with(@href, '/Attraction') and .//img]`;
+
+    try {
+        const links = await driver.findElements(By.xpath(xpath));
+
+        const urls: string[] = [];
+        for (const link of links) {
+            const href = await link.getAttribute('href');
+            if (href) {
+                urls.push(href);
+            }
+        }
+
+        return urls;
+    } catch (error) {
+        console.warn('Error getting attraction URLs:', error);
+        return [];
+    }
+}
+
+export async function getAttractionNames(driver: WebDriver): Promise<string[]> {
+    // XPath to find <h3> inside <a> whose href starts with /Attraction_Review
+    const xpath = `//a[starts-with(@href, '/Attraction_Review')]/h3`;
+
+    try {
+        const h3Elements = await driver.findElements(By.xpath(xpath));
+
+        const names: string[] = [];
+        for (const h3 of h3Elements) {
+            const text = await h3.getText();
+            if (text) {
+                names.push(text.trim().replace(/^\d+\.\s*/, ''));
+            }
+        }
+
+        return names;
+    } catch (error) {
+        console.warn('Error getting attraction names:', error);
+        return [];
+    }
+}
+
 export async function getWebsiteAndEmail(driver: WebDriver): Promise<ContactInfo> {
     const result: ContactInfo = { websiteUrl: null, email: null };
 
@@ -193,4 +286,24 @@ export async function closeAllTabsExceptFirst(driver: WebDriver): Promise<void>
 
     await driver.switchTo().window(originalHandle);
     console.log(`Switched back to original tab: ${originalHandle}`);
-}
+}
+
+export async function clickFirstAttractionLinkInForm(driver: WebDriver): Promise<boolean> {
+    const xpath = `//form//a[starts-with(@href, '/Attraction')]`;
+
+    try {
+        // Wait for the link to appear inside a form
+        const link = await driver.wait(until.elementLocated(By.xpath(xpath)), 5000);
+
+        await driver.wait(until.elementIsVisible(link), 5000);
+        await driver.wait(until.elementIsEnabled(link), 5000);
+
+        await link.click();
+        console.log('Clicked the first attraction link in the form.');
+
+        return true;
+    } catch (error) {
+        console.warn('Attraction link not found or not clickable.', error);
+        return false;
+    }
+}

src/lib/utils.ts

@@ -2,7 +2,8 @@
  * Utility class for common WebDriver operations
  */
 
-import { WebDriver, By, until } from 'selenium-webdriver';
+import { Builder, By, until, WebDriver } from 'selenium-webdriver';
+import chrome from 'selenium-webdriver/chrome';
 import { writeFileSync, existsSync, appendFileSync } from 'fs';
 import * as path from 'path';
 import { ContactInfo } from './types';
@@ -13,9 +14,33 @@ export class WebDriverUtils {
      * @param seconds Number of seconds to wait
      * @returns Promise that resolves after the specified time
      */
-    static async wait(seconds: number): Promise<void> {
-        console.log(`Waiting for ${seconds} seconds...`);
-        return new Promise(resolve => setTimeout(resolve, seconds * 1000));
+    static async wait(driver?: WebDriver): Promise<void> {
+        const seconds = Math.floor(Math.random() * 1000) % 3 + 3;
+        console.log(`Scrolling to bottom for ${seconds} seconds...`);
+
+        const endTime = Date.now() + seconds * 1000;
+
+        let scrollCounter = 0;
+
+        while (Date.now() < endTime) {
+
+            /*
+            try {
+                if (driver) {
+                    await driver.executeScript(`
+                        window.scrollBy(0, 10);
+                    `);
+
+                    scrollCounter++;
+                }
+            } catch (error) {
+                console.warn('Scroll failed:', error);
+            }
+            */
+
+            // Wait a little between scrolls
+            await new Promise(resolve => setTimeout(resolve, 500));
+        }
     }
 
     /**
@@ -43,3 +68,97 @@ export function saveContactInfoToCSV(city: string, contactInfo: ContactInfo, fil
 
     console.log(`Contact info saved to ${filePath}`);
 }
+
+export async function disableCookiesInChrome(): Promise<WebDriver | null> {
+    // Set Chrome options
+    const options = new chrome.Options();
+
+    // 1. Block all cookies
+    //options.setUserPreferences({
+    //    'profile.default_content_setting_values.cookies': 2, // 2 = Block all
+    //    'profile.block_third_party_cookies': true
+    //});
+
+    // 2. Optional: Launch in incognito for extra privacy
+    options.addArguments('--incognito');
+    options.addArguments('--start-maximized');
+
+    let driver: WebDriver | null = null;
+
+    try {
+        driver = await new Builder()
+            .forBrowser('chrome')
+            .setChromeOptions(options)
+            .build();
+
+        await driver.get('https://www.tripadvisor.com/');
+
+        console.log('Chrome launched with cookies disabled.');
+
+        // Optional: Verify cookies are blocked by trying to set/get a cookie
+        await driver.manage().addCookie({ name: 'test', value: '123' });
+        const cookies = await driver.manage().getCookies();
+        console.log('Cookies after trying to add:', cookies); // Should be empty or restricted
+
+        return driver;
+    } catch (error) {
+        console.error('Error:', error);
+        return driver;
+    }
+
+}
+
+export async function useChrome(initialUrl: string = "https://www.tripadvisor.com/"): Promise<WebDriver | null> {
+    // Set Chrome options
+    const options = new chrome.Options();
+
+    // 1. Block all cookies
+    //options.setUserPreferences({
+    //    'profile.default_content_setting_values.cookies': 2, // 2 = Block all
+    //    'profile.block_third_party_cookies': true
+    //});
+
+    // 2. Optional: Launch in incognito for extra privacy
+    options.addArguments('--incognito');
+    options.addArguments('--start-maximized');
+
+    let driver: WebDriver | null = null;
+
+    try {
+        driver = await new Builder()
+            .forBrowser('chrome')
+            .setChromeOptions(options)
+            .build();
+
+        await driver.get(initialUrl);
+
+        console.log('Chrome launched with cookies disabled.');
+
+        // Optional: Verify cookies are blocked by trying to set/get a cookie
+        await driver.manage().addCookie({ name: 'test', value: '123' });
+        const cookies = await driver.manage().getCookies();
+        console.log('Cookies after trying to add:', cookies); // Should be empty or restricted
+
+        return driver;
+    } catch (error) {
+        console.error('Error:', error);
+        return driver;
+    }
+
+}
+
+export async function useExistingChrome(): Promise<WebDriver> {
+    // Connect to an existing Chrome browser running in debug mode on port 9222
+    const options = new chrome.Options();
+
+    // Set the debugger address to connect to the existing Chrome instance
+    options.debuggerAddress('localhost:9222');
+
+    // Create WebDriver instance that connects to the existing browser
+    const driver: WebDriver = await new Builder()
+        .forBrowser('chrome')
+        .setChromeOptions(options)
+        .build();
+
+    return driver;
+}