/// <summary>
/// Entry point of the worker piece of the process
/// Notice that you can run as many workers as you want to in order to make the crawling faster
/// </summary>
/// <param name="args"></param>
static void Main(string[] args)
{
// Configuring Log Object
LogSetup.InitializeLog("PlayStoreWorker.log", "info");
Logger logger = LogManager.GetCurrentClassLogger();
logger.Info("Worker Started");
// Control Variable (Bool - Should the process use proxies? )
bool isUsingProxies = false;
// Checking for the need to use proxies
if (args != null && args.Length == 1)
{
// Setting flag to true
isUsingProxies = true;
// Loading proxies from .txt received as argument
String fPath = args[0];
// Sanity Check
if (!File.Exists(fPath))
{
logger.Fatal("Couldnt find proxies on path : " + fPath);
System.Environment.Exit(-100);
}
// Reading Proxies from File
string[] fLines = File.ReadAllLines(fPath, Encoding.GetEncoding("UTF-8"));
try
{
// Actual Load of Proxies
ProxiesLoader.Load(fLines.ToList());
}
catch (Exception ex)
{
logger.Fatal(ex);
System.Environment.Exit(-101);
}
}
// Parser
PlayStoreParser parser = new PlayStoreParser();
// Configuring MongoDB Wrapper
MongoDBWrapper mongoDB = new MongoDBWrapper();
string fullServerAddress = String.Join(":", Consts.MONGO_SERVER, Consts.MONGO_PORT);
mongoDB.ConfigureDatabase(Consts.MONGO_USER, Consts.MONGO_PASS, Consts.MONGO_AUTH_DB, fullServerAddress, Consts.MONGO_TIMEOUT, Consts.MONGO_DATABASE, Consts.MONGO_COLLECTION);
/*
* // populate min downloaded & max downloaded
* int count = 0;
* var apps = mongoDB.FindAll<AppModel>();
* foreach(var a in apps)
* {
* a.FillMinAndMaxInstalls();
++count;
*
* if((count % 100) == 0)
* {
* Console.WriteLine("updated {0}", count);
* }
*
* if (!mongoDB.UpsertKeyEq<AppModel>(a, "Url", a.Url))
* {
* Console.WriteLine("UpsertKey failed");
* }
* }
*/
// Creating Instance of Web Requests Server
WebRequests server = new WebRequests();
// Queued App Model
QueuedApp app;
// Retry Counter (Used for exponential wait increasing logic)
int retryCounter = 0;
// Iterating Over MongoDB Records while no document is found to be processed
while ((app = mongoDB.FindAndModify()) != null)
{
try
{
// Building APP URL
string appUrl = app.Url;
// Sanity check of app page url
if (app.Url.IndexOf("http", StringComparison.OrdinalIgnoreCase) < 0)
{
appUrl = Consts.APP_URL_PREFIX + app.Url;
}
// Checking if this app is on the database already
if (mongoDB.AppProcessed(appUrl))
{
// Console Feedback, Comment this line to disable if you want to
logger.Info("Duplicated App, skipped.");
// Delete it from the queue and continues the loop
mongoDB.RemoveFromQueue(app.Url);
continue;
}
// Configuring server and Issuing Request
server.Headers.Add(Consts.ACCEPT_LANGUAGE);
server.Host = Consts.HOST;
server.UserAgent = Consts.GITHUBURL;
server.Encoding = "utf-8";
server.EncodingDetection = WebRequests.CharsetDetection.DefaultCharset;
// Checking for the need to use "HTTP Proxies"
if (isUsingProxies)
{
server.Proxy = ProxiesLoader.GetWebProxy();
}
// Issuing HTTP Request
string response = server.Get(appUrl);
// Flag Indicating Success while processing and parsing this app
bool ProcessingWorked = true;
// Sanity Check
if (String.IsNullOrEmpty(response) || server.StatusCode != System.Net.HttpStatusCode.OK)
{
logger.Info("Error opening app page : " + appUrl);
ProcessingWorked = false;
if (isUsingProxies)
{
ProxiesLoader.IncrementCurrentProxy();
}
// Renewing WebRequest Object to get rid of Cookies
server = new WebRequests();
// Fallback time variable
double waitTime;
// Checking which "Waiting Logic" to use - If there are proxies being used, there's no need to wait too much
// If there are no proxies in use, on the other hand, the process must wait more time
if (isUsingProxies)
{
// Waits two seconds everytime
waitTime = TimeSpan.FromSeconds(2).TotalMilliseconds;
}
else
{
// Increments retry counter
retryCounter++;
// Checking for maximum retry count
if (retryCounter >= 8)
{
waitTime = TimeSpan.FromMinutes(20).TotalMilliseconds;
}
else
{
// Calculating next wait time ( 2 ^ retryCounter seconds)
waitTime = TimeSpan.FromSeconds(Math.Pow(2, retryCounter)).TotalMilliseconds;
}
}
// Hiccup to avoid google blocking connections in case of heavy traffic from the same IP
logger.Info("======================================================");
logger.Info("\n\tFallback : " + waitTime + " Seconds");
Thread.Sleep(Convert.ToInt32(waitTime));
// If The Status code is "ZERO" (it means 404) - App must be removed from "Queue"
if (server.StatusCode == 0)
{
// Console Feedback
logger.Info("\tApp Not Found (404) - " + app.Url);
mongoDB.RemoveFromQueue(app.Url);
}
logger.Info("======================================================");
}
else
{
// Reseting retry counter
retryCounter = 0;
// Parsing Useful App Data
AppModel parsedApp = parser.ParseAppPage(response, appUrl);
// Normalizing URLs
if (!String.IsNullOrWhiteSpace(parsedApp.DeveloperPrivacyPolicy))
{
parsedApp.DeveloperPrivacyPolicy = parsedApp.DeveloperPrivacyPolicy.Replace("https://www.google.com/url?q=", String.Empty);
}
if (!String.IsNullOrWhiteSpace(parsedApp.DeveloperWebsite))
{
parsedApp.DeveloperNormalizedDomain = parser.NormalizeDomainName(parsedApp.DeveloperWebsite);
}
List <String> relatedApps = new List <String> ();
// Avoiding Exceptions caused by "No Related Apps" situations - Must be treated differently
try
{
// Parsing "Related Apps" and "More From Developer" Apps (URLS Only)
foreach (string extraAppUrl in parser.ParseExtraApps(response))
{
relatedApps.Add(Consts.APP_URL_PREFIX + extraAppUrl);
}
// Adding "Related Apps" to Apps Model
parsedApp.RelatedUrls = relatedApps.Distinct().ToArray();
}
catch
{
logger.Info("\tNo Related Apps Found. Skipping");
}
// Inserting App into Mongo DB Database
if (!mongoDB.UpsertKeyEq <AppModel>(parsedApp, "Url", appUrl))
{
ProcessingWorked = false;
}
// If the processing failed, do not remove the app from the database, instead, keep it and flag it as not busy
// so that other workers can try to process it later
if (!ProcessingWorked)
{
mongoDB.ToggleBusyApp(app, false);
}
else // On the other hand, if processing worked, removes it from the database
{
// Console Feedback, Comment this line to disable if you want to
Console.ForegroundColor = ConsoleColor.Red;
logger.Info("Inserted App : " + parsedApp.Name);
Console.ForegroundColor = ConsoleColor.White;
mongoDB.RemoveFromQueue(app.Url);
}
// Counters for console feedback only
int extraAppsCounter = 0, newExtraApps = 0;
// Parsing "Related Apps" and "More From Developer" Apps (URLS Only)
foreach (string extraAppUrl in relatedApps)
{
// Incrementing counter of extra apps
extraAppsCounter++;
// Assembling Full app Url to check with database
string fullExtraAppUrl;
if (extraAppUrl.IndexOf("https://play.google.com/") >= 0)
{
fullExtraAppUrl = extraAppUrl;
}
else
{
fullExtraAppUrl = Consts.APP_URL_PREFIX + extraAppUrl;
}
// Checking if the app was either processed or queued to be processed already
if ((!mongoDB.AppProcessed(fullExtraAppUrl)) && (!mongoDB.IsAppOnQueue(extraAppUrl)))
{
// Incrementing counter of inserted apps
newExtraApps++;
// Adds it to the queue of apps to be processed
mongoDB.AddToQueue(extraAppUrl);
}
}
// Console Feedback
logger.Info("Queued " + newExtraApps + " / " + extraAppsCounter + " related apps");
}
}
catch (Exception ex)
{
logger.Error(ex);
}
finally
{
try
{
// Toggles Busy status back to false
mongoDB.ToggleBusyApp(app, false);
}
catch (Exception ex)
{
// Toggle Busy App may raise an exception in case of lack of internet connection, so, i must use this
// "inner catch" to avoid it from happenning
logger.Error(ex);
}
}
}
}
static void Main(string[] args)
{
// Creating Needed Instances
RequestsHandler httpClient = new RequestsHandler();
AppStoreParser parser = new AppStoreParser();
// Loading Configuration
LogSetup.InitializeLog("Apple_Store_Numerics_Worker.log", "info");
_logger = LogManager.GetCurrentClassLogger();
// Loading Config
_logger.Info("Loading Configurations from App.config");
LoadConfiguration();
// Control Variable (Bool - Should the process use proxies? )
bool shouldUseProxies = false;
// Checking for the need to use proxies
if (args != null && args.Length == 1)
{
// Setting flag to true
shouldUseProxies = true;
// Loading proxies from .txt received as argument
String fPath = args[0];
// Sanity Check
if (!File.Exists(fPath))
{
_logger.Fatal("Couldnt find proxies on path : " + fPath);
System.Environment.Exit(-100);
}
// Reading Proxies from File
string[] fLines = File.ReadAllLines(fPath, Encoding.GetEncoding("UTF-8"));
try
{
// Actual Load of Proxies
ProxiesLoader.Load(fLines.ToList());
}
catch (Exception ex)
{
_logger.Fatal(ex);
System.Environment.Exit(-101);
}
}
// AWS Queue Handler
_logger.Info("Initializing Queues");
AWSSQSHelper numericUrlQueue = new AWSSQSHelper(_numericUrlsQueueName, _maxMessagesPerDequeue, _awsKey, _awsKeySecret);
AWSSQSHelper appsUrlQueue = new AWSSQSHelper(_appUrlsQueueName, _maxMessagesPerDequeue, _awsKey, _awsKeySecret);
// Setting Error Flag to No Error ( 0 )
System.Environment.ExitCode = 0;
// Initialiazing Control Variables
int fallbackWaitTime = 1;
_logger.Info("Started Processing Numeric Urls");
do
{
try
{
// Dequeueing messages from the Queue
if (!numericUrlQueue.DeQueueMessages())
{
Thread.Sleep(_hiccupTime); // Hiccup
continue;
}
// Checking for no message received, and false positives situations
if (!numericUrlQueue.AnyMessageReceived())
{
// If no message was found, increases the wait time
int waitTime;
if (fallbackWaitTime <= 12)
{
// Exponential increase on the wait time, truncated after 12 retries
waitTime = Convert.ToInt32(Math.Pow(2, fallbackWaitTime) * 1000);
}
else // Reseting Wait after 12 fallbacks
{
waitTime = 2000;
fallbackWaitTime = 0;
}
fallbackWaitTime++;
// Sleeping before next try
Console.WriteLine("Fallback (seconds) => " + waitTime);
Thread.Sleep(waitTime);
continue;
}
// Reseting fallback time
fallbackWaitTime = 1;
// Iterating over dequeued Messages
foreach (var numericUrl in numericUrlQueue.GetDequeuedMessages())
{
try
{
// Retries Counter
int retries = 0;
string htmlResponse;
// Retrying if necessary
do
{
// Executing Http Request for the Category Url
htmlResponse = httpClient.Get(numericUrl.Body, shouldUseProxies);
if (String.IsNullOrEmpty(htmlResponse))
{
_logger.Info("Retrying Request for Category Page");
retries++;
}
} while (String.IsNullOrWhiteSpace(htmlResponse) && retries <= _maxRetries);
// Checking if retries failed
if (String.IsNullOrWhiteSpace(htmlResponse))
{
// Deletes Message and moves on
numericUrlQueue.DeleteMessage(numericUrl);
continue;
}
// Feedback
_logger.Info("Current page " + numericUrl.Body);
foreach (var parsedAppUrl in parser.ParseAppsUrls(htmlResponse))
{
// Enqueueing App Urls
appsUrlQueue.EnqueueMessage(HttpUtility.HtmlDecode(parsedAppUrl));
}
}
catch (Exception ex)
{
_logger.Info(ex);
}
finally
{
// Deleting the message
numericUrlQueue.DeleteMessage(numericUrl);
}
}
}
catch (Exception ex)
{
_logger.Error(ex);
}
} while (true);
}
static void Main(string[] args)
{
// Creating Needed Instances
RequestsHandler httpClient = new RequestsHandler();
AppStoreParser parser = new AppStoreParser();
// Setting Up Log
LogSetup.InitializeLog("Apple_Store_Crawler.log", "info");
_logger = LogManager.GetCurrentClassLogger();
// Starting Flow
_logger.Info("Worker Started");
// Loading Configuration
_logger.Info("Reading Configuration");
LoadConfiguration();
// Control Variable (Bool - Should the process use proxies? )
bool shouldUseProxies = false;
// Checking for the need to use proxies
if (args != null && args.Length == 1)
{
// Setting flag to true
shouldUseProxies = true;
// Loading proxies from .txt received as argument
String fPath = args[0];
// Sanity Check
if (!File.Exists(fPath))
{
_logger.Fatal("Couldnt find proxies on path : " + fPath);
System.Environment.Exit(-100);
}
// Reading Proxies from File
string[] fLines = File.ReadAllLines(fPath, Encoding.GetEncoding("UTF-8"));
try
{
// Actual Load of Proxies
ProxiesLoader.Load(fLines.ToList());
}
catch (Exception ex)
{
_logger.Fatal(ex);
System.Environment.Exit(-101);
}
}
// AWS Queue Handler
_logger.Info("Initializing Queues");
AWSSQSHelper sqsWrapper = new AWSSQSHelper(_categoriesQueueName, 10, _awsKey, _awsKeySecret);
// Step 1 - Trying to obtain the root page html (source of all the apps)
var rootPageResponse = httpClient.GetRootPage(shouldUseProxies);
// Sanity Check
if (String.IsNullOrWhiteSpace(rootPageResponse))
{
_logger.Info("Error obtaining Root Page HTMl - Aborting", "Timeout Error");
return;
}
// Step 2 - Extracting Category Urls from the Root Page and queueing their Urls
foreach (var categoryUrl in parser.ParseCategoryUrls(rootPageResponse))
{
// Logging Feedback
_logger.Info("Queueing Category : " + categoryUrl);
// Queueing Category Urls
sqsWrapper.EnqueueMessage(categoryUrl);
}
_logger.Info("End of Bootstrapping phase");
}
static void Main(string[] args)
{
// Creating Needed Instances
RequestsHandler httpClient = new RequestsHandler();
AppStoreParser parser = new AppStoreParser();
// Loading Configuration
LogSetup.InitializeLog("Apple_Store_Urls_Worker.log", "info");
_logger = LogManager.GetCurrentClassLogger();
// Loading Config
_logger.Info("Loading Configurations from App.config");
LoadConfiguration();
// Control Variable (Bool - Should the process use proxies? )
bool shouldUseProxies = false;
// Checking for the need to use proxies
if (args != null && args.Length == 1)
{
// Setting flag to true
shouldUseProxies = true;
// Loading proxies from .txt received as argument
String fPath = args[0];
// Sanity Check
if (!File.Exists(fPath))
{
_logger.Fatal("Couldnt find proxies on path : " + fPath);
System.Environment.Exit(-100);
}
// Reading Proxies from File
string[] fLines = File.ReadAllLines(fPath, Encoding.GetEncoding("UTF-8"));
try
{
// Actual Load of Proxies
ProxiesLoader.Load(fLines.ToList());
}
catch (Exception ex)
{
_logger.Fatal(ex);
System.Environment.Exit(-101);
}
}
// AWS Queue Handler
_logger.Info("Initializing Queues");
AWSSQSHelper appsUrlQueue = new AWSSQSHelper(_appUrlsQueueName, _maxMessagesPerDequeue, _awsKey, _awsKeySecret);
AWSSQSHelper appsDataQueue = new AWSSQSHelper(_appsDataQueueName, _maxMessagesPerDequeue, _awsKey, _awsKeySecret);
// Setting Error Flag to No Error ( 0 )
System.Environment.ExitCode = 0;
// Initialiazing Control Variables
int fallbackWaitTime = 1;
_logger.Info("Started Processing Individual Apps Urls");
do
{
try
{
// Dequeueing messages from the Queue
if (!appsUrlQueue.DeQueueMessages())
{
Thread.Sleep(_hiccupTime); // Hiccup
continue;
}
// Checking for no message received, and false positives situations
if (!appsUrlQueue.AnyMessageReceived())
{
// If no message was found, increases the wait time
int waitTime;
if (fallbackWaitTime <= 12)
{
// Exponential increase on the wait time, truncated after 12 retries
waitTime = Convert.ToInt32(Math.Pow(2, fallbackWaitTime) * 1000);
}
else // Reseting Wait after 12 fallbacks
{
waitTime = 2000;
fallbackWaitTime = 0;
}
fallbackWaitTime++;
// Sleeping before next try
Console.WriteLine("Fallback (seconds) => " + waitTime);
Thread.Sleep(waitTime);
continue;
}
// Reseting fallback time
fallbackWaitTime = 1;
// Iterating over dequeued Messages
foreach (var appUrl in appsUrlQueue.GetDequeuedMessages())
{
bool processingWorked = true;
try
{
// Retries Counter
int retries = 0;
string htmlResponse;
// Retrying if necessary
do
{
// Executing Http Request for the Category Url
//appUrl.Body = "https://itunes.apple.com/us/app/action-run-3d/id632371832?mt=8";
//appUrl.Body = "https://itunes.apple.com/us/app/emoji-2-free-new-emoticons/id521863802?mt=8";
//appUrl.Body = "https://itunes.apple.com/us/app/candy-crush-saga/id553834731?mt=8";
//appUrl.Body = "https://itunes.apple.com/us/app/dba-den-bla-avis/id448605988?mt=8";
htmlResponse = httpClient.Get(appUrl.Body, shouldUseProxies);
if (String.IsNullOrEmpty(htmlResponse))
{
// Extending Fallback time
retries++;
int sleepTime = retries * _hiccupTime <= 30000 ? retries * _hiccupTime : 30000;
_logger.Info("Retrying Request for App Page [ " + sleepTime / 1000 + " ]");
Thread.Sleep(sleepTime);
}
} while (String.IsNullOrWhiteSpace(htmlResponse) && retries <= _maxRetries);
// Checking if retries failed
if (String.IsNullOrWhiteSpace(htmlResponse))
{
continue;
}
// Feedback
_logger.Info("Current page " + appUrl.Body, "Parsing App Data");
// Parsing Data out of the Html Page
AppleStoreAppModel parsedApp = parser.ParseAppPage(htmlResponse);
parsedApp.url = appUrl.Body;
// Enqueueing App Data
appsDataQueue.EnqueueMessage(parsedApp.ToJson());
// Little Hiccup
Thread.Sleep(_hiccupTime);
}
catch (Exception ex)
{
_logger.Error(ex);
// Setting Flag to "False"
processingWorked = false;
}
finally
{
//Deleting the message - Only if the processing worked
if (processingWorked)
{
appsUrlQueue.DeleteMessage(appUrl);
}
}
}
}
catch (Exception ex)
{
_logger.Error(ex);
}
} while (true);
}
/// <summary>
/// Entry point of the crawler
/// </summary>
/// <param name="args"></param>
static void Main(string[] args)
{
// Setting Up Log
LogSetup.InitializeLog("PlayStoreCrawler.log", "info");
_logger = LogManager.GetCurrentClassLogger();
// Control Variable (Bool - Should the process use proxies? )
bool isUsingProxies = false;
// Checking for the need to use HTTP proxies or not
if (args != null && args.Length == 1)
{
_logger.Info("Loading Proxies from File");
// Setting flag to true
isUsingProxies = true;
// Loading proxies from .txt received as argument
String fPath = args[0];
// Sanity Check
if (!File.Exists(fPath))
{
_logger.Fatal("Couldnt find proxies on path : " + fPath);
System.Environment.Exit(-100);
}
// Reading Proxies from File
string[] fLines = File.ReadAllLines(fPath, Encoding.GetEncoding("UTF-8"));
try
{
// Actual Load of Proxies
ProxiesLoader.Load(fLines.ToList());
}
catch (Exception ex)
{
_logger.Fatal(ex);
System.Environment.Exit(-101);
}
}
// Configuring MongoDB Wrapper
_logger.Info("Setting up MongoDB Collections and Indexes");
_mongoDB = new MongoDBWrapper();
string fullServerAddress = String.Join(":", Consts.MONGO_SERVER, Consts.MONGO_PORT);
_mongoDB.ConfigureDatabase(Consts.MONGO_USER, Consts.MONGO_PASS, Consts.MONGO_AUTH_DB, fullServerAddress, Consts.MONGO_TIMEOUT, Consts.MONGO_DATABASE, Consts.MONGO_COLLECTION);
// Ensuring the database collections have the proper indexes
_mongoDB.EnsureIndex("Url");
_mongoDB.EnsureIndex("IsBusy", Consts.QUEUED_APPS_COLLECTION);
_mongoDB.EnsureIndex("Url", Consts.QUEUED_APPS_COLLECTION);
// Main Flow
_logger.Info("Started Bootstrapping Steps");
// Scrapping "Play Store Categories"
foreach (var categoriesKVP in BootstrapTerms.categoriesAndNames)
{
CrawlCategory(categoriesKVP.Key, categoriesKVP.Value, isUsingProxies);
}
// Queueing Apps that start with each of the characters from "A" to "Z"
foreach (var character in BootstrapTerms.charactersSearchTerms)
{
CrawlStore(character, isUsingProxies);
}
/// ... Keep Adding characters / search terms in order to increase the crawler's reach
// APP CATEGORIES
foreach (var category in BootstrapTerms.categoriesSearchTerms)
{
CrawlStore(category, isUsingProxies);
}
// Extra "Random" Search terms to increase even more the crawler's reach
foreach (var miscTerm in BootstrapTerms.miscSearchTerms)
{
CrawlStore(miscTerm, isUsingProxies);
}
// Country Names as Search terms to increase even more the crawler's reach
foreach (var countryName in BootstrapTerms.countryNames)
{
CrawlStore(countryName, isUsingProxies);
}
}
static void Main(string[] args)
{
// Loading Configuration
LogSetup.InitializeLog("Apple_Store_Recorder.log", "info");
_logger = LogManager.GetCurrentClassLogger();
// Loading Config
_logger.Info("Loading Configurations from App.config");
LoadConfiguration();
// Initializing Queue
_logger.Info("Initializing Queue");
AWSSQSHelper appsDataQueue = new AWSSQSHelper(_appsDataQueueName, _maxMessagesPerDequeue, _awsKey, _awsKeySecret);
AWSSQSHelper backup = new AWSSQSHelper("DeadLetter", _maxMessagesPerDequeue, _awsKey, _awsKeySecret);
// Creating MongoDB Instance
_logger.Info("Loading MongoDB / Creating Instances");
MongoDBWrapper mongoDB = new MongoDBWrapper();
string serverAddr = String.Join(":", Consts.MONGO_SERVER, Consts.MONGO_PORT);
mongoDB.ConfigureDatabase(Consts.MONGO_USER, Consts.MONGO_PASS, Consts.MONGO_AUTH_DB, serverAddr, 10000, Consts.MONGO_DATABASE, Consts.MONGO_COLLECTION);
// Setting Error Flag to No Error ( 0 )
System.Environment.ExitCode = 0;
// Initialiazing Control Variables
int fallbackWaitTime = 1;
// Buffer of Messages to be recorder
List <AppleStoreAppModel> recordsBuffer = new List <AppleStoreAppModel> ();
List <Message> messagesBuffer = new List <Message> ();
// Insert Batch Size
int batchSize = 1000;
_logger.Info("Started Recording App Data");
do
{
try
{
// Dequeueing messages from the Queue
if (!appsDataQueue.DeQueueMessages())
{
Thread.Sleep(_hiccupTime); // Hiccup
continue;
}
// Checking for no message received, and false positives situations
if (!appsDataQueue.AnyMessageReceived())
{
// If no message was found, increases the wait time
int waitTime;
if (fallbackWaitTime <= 12)
{
// Exponential increase on the wait time, truncated after 12 retries
waitTime = Convert.ToInt32(Math.Pow(2, fallbackWaitTime) * 1000);
}
else // Reseting Wait after 12 fallbacks
{
waitTime = 2000;
fallbackWaitTime = 0;
}
fallbackWaitTime++;
// Sleeping before next try
Console.WriteLine("Fallback (seconds) => " + waitTime);
Thread.Sleep(waitTime);
continue;
}
// Reseting fallback time
fallbackWaitTime = 1;
// Iterating over dequeued Messages
foreach (var appDataMessage in appsDataQueue.GetDequeuedMessages())
{
try
{
// Deserializing message
var appData = AppleStoreAppModel.FromJson(appDataMessage.Body);
// Dumping "Url" to "_id"
appData._id = appData.url;
// Adding it to the buffer of records to be recorded
recordsBuffer.Add(appData);
// Adding message to the buffer of messages to be deleted
messagesBuffer.Add(appDataMessage);
// Is it time to batch insert ?
if ((recordsBuffer.Count % batchSize) == 0)
{
// Batch Insertion
mongoDB.BatchInsert <AppleStoreAppModel> (recordsBuffer);
// Logging Feedback
_logger.Info("\tApps Recorded : " + recordsBuffer.Count);
// Deleting Messages
messagesBuffer.ForEach((msg) => appsDataQueue.DeleteMessage(msg));
_logger.Info("\tMessages Deleted: " + messagesBuffer.Count);
// Clearing Buffers
recordsBuffer.Clear();
messagesBuffer.Clear();
}
}
catch (Exception ex)
{
_logger.Error(ex);
}
finally
{
// Deleting the message
appsDataQueue.DeleteMessage(appDataMessage);
backup.EnqueueMessage(appDataMessage.Body);
}
}
}
catch (Exception ex)
{
_logger.Error(ex);
}
} while (true);
}
/// <summary>
/// Entry point of the crawler
/// </summary>
/// <param name="args"></param>
static void Main(string[] args)
{
// Setting Up Log
LogSetup.InitializeLog("PlayStoreCrawler.log", "info");
_logger = LogManager.GetCurrentClassLogger();
// Control Variable (Bool - Should the process use proxies? )
bool isUsingProxies = false;
// Checking for the need to use HTTP proxies or not
if (args != null && args.Length == 1)
{
_logger.Info("Loading Proxies from File");
// Setting flag to true
isUsingProxies = true;
// Loading proxies from .txt received as argument
String fPath = args[0];
// Sanity Check
if (!File.Exists(fPath))
{
_logger.Fatal("Couldnt find proxies on path : " + fPath);
System.Environment.Exit(-100);
}
// Reading Proxies from File
string[] fLines = File.ReadAllLines(fPath, Encoding.GetEncoding("UTF-8"));
try
{
// Actual Load of Proxies
ProxiesLoader.Load(fLines.ToList());
}
catch (Exception ex)
{
_logger.Fatal(ex);
System.Environment.Exit(-101);
}
}
// Main Flow
_logger.Info("Started Bootstrapping Steps");
// Scrapping "Play Store Categories"
foreach (var categoriesKVP in BootstrapTerms.categoriesAndNames)
{
CrawlCategory(categoriesKVP.Key, categoriesKVP.Value, isUsingProxies);
}
// Queueing Apps that start with each of the characters from "A" to "Z"
foreach (var character in BootstrapTerms.charactersSearchTerms)
{
CrawlStore(character, isUsingProxies);
}
/// ... Keep Adding characters / search terms in order to increase the crawler's reach
// APP CATEGORIES
foreach (var category in BootstrapTerms.categoriesSearchTerms)
{
CrawlStore(category, isUsingProxies);
}
// Extra "Random" Search terms to increase even more the crawler's reach
foreach (var miscTerm in BootstrapTerms.miscSearchTerms)
{
CrawlStore(miscTerm, isUsingProxies);
}
// Country Names as Search terms to increase even more the crawler's reach
foreach (var countryName in BootstrapTerms.countryNames)
{
CrawlStore(countryName, isUsingProxies);
}
}
/// <summary>
/// Entry point of the crawler
/// </summary>
/// <param name="args"></param>
static void Main(string[] args)
{
// Setting Up Log
LogSetup.InitializeLog("PlayStoreCrawler.log", "info");
_logger = LogManager.GetCurrentClassLogger();
// Crawling App Store using all characters as the Search Input
CrawlStore("A");
CrawlStore("B");
CrawlStore("C");
CrawlStore("D");
CrawlStore("E");
CrawlStore("F");
CrawlStore("G");
CrawlStore("H");
CrawlStore("I");
CrawlStore("J");
CrawlStore("K");
CrawlStore("L");
CrawlStore("M");
CrawlStore("N");
CrawlStore("O");
CrawlStore("P");
CrawlStore("Q");
CrawlStore("R");
CrawlStore("S");
CrawlStore("T");
CrawlStore("U");
CrawlStore("V");
CrawlStore("X");
CrawlStore("Y");
CrawlStore("Z");
CrawlStore("W");
/// ... Keep Adding characters / search terms in order to increase the crawler's reach
// APP CATEGORIES
CrawlStore("BOOKS");
CrawlStore("BUSINESS");
CrawlStore("COMICS");
CrawlStore("COMMUNICATION");
CrawlStore("EDUCATION");
CrawlStore("ENTERTAINMENT");
CrawlStore("FINANCE");
CrawlStore("HEALTH");
CrawlStore("LIFESTYLE");
CrawlStore("LIVE WALLPAPER");
CrawlStore("MEDIA");
CrawlStore("MEDICAL");
CrawlStore("MUSIC");
CrawlStore("NEWS");
CrawlStore("PERSONALIZATION");
CrawlStore("PHOTOGRAPHY");
CrawlStore("PRODUCTIVITY");
CrawlStore("SHOPPING");
CrawlStore("SOCIAL");
CrawlStore("SPORTS");
CrawlStore("TOOLS");
CrawlStore("TRANSPORTATION");
CrawlStore("TRAVEL");
CrawlStore("WEATHER");
CrawlStore("WIDGETS");
CrawlStore("ARCADE");
CrawlStore("BRAIN");
CrawlStore("CASUAL");
CrawlStore("CARDS");
CrawlStore("RACING");
// Extra "Random" Search terms to increase even more the crawler's reach
CrawlStore("INDIE");
CrawlStore("ZOMBIE");
CrawlStore("CATS");
CrawlStore("ROOT");
CrawlStore("GPS");
CrawlStore("BLUETOOTH");
CrawlStore("COMPASS");
CrawlStore("WALLPAPER");
CrawlStore("TORRENT");
CrawlStore("P**N");
CrawlStore("PLAYER");
CrawlStore("WINE");
CrawlStore("ANTIVIRUS");
CrawlStore("P**N");
// Country Names as Search terms to increase even more the crawler's reach
CrawlStore("Afghanistan");
CrawlStore("Albania");
CrawlStore("Algeria");
CrawlStore("American");
CrawlStore("Andorra");
CrawlStore("Angola");
CrawlStore("Anguilla");
CrawlStore("Antigua");
CrawlStore("Argentina");
CrawlStore("Armenia");
CrawlStore("Aruba");
CrawlStore("Australia");
CrawlStore("Austria");
CrawlStore("Azerbaijan");
CrawlStore("Bahamas");
CrawlStore("Bahrain");
CrawlStore("Bangladesh");
CrawlStore("Barbados");
CrawlStore("Belarus");
CrawlStore("Belgium");
CrawlStore("Belize");
CrawlStore("Benin");
CrawlStore("Bermuda");
CrawlStore("Bhutan");
CrawlStore("Bolivia");
CrawlStore("Bosnia");
CrawlStore("Botswana");
CrawlStore("Bouvet");
CrawlStore("Brazil");
CrawlStore("Brunei");
CrawlStore("Bulgaria");
CrawlStore("Burkina");
CrawlStore("Burundi");
CrawlStore("Cambodia");
CrawlStore("Cameroon");
CrawlStore("Canada");
CrawlStore("Cape");
CrawlStore("Cayman");
CrawlStore("Central");
CrawlStore("Chad");
CrawlStore("Chile");
CrawlStore("China");
CrawlStore("Christmas");
CrawlStore("Cocos");
CrawlStore("Colombia");
CrawlStore("Comoros");
CrawlStore("Congo");
CrawlStore("Congo");
CrawlStore("Cook");
CrawlStore("Costa");
CrawlStore("Croatia");
CrawlStore("Cuba");
CrawlStore("Cyprus");
CrawlStore("Czech");
CrawlStore("Denmark");
CrawlStore("Djibouti");
CrawlStore("Dominica");
CrawlStore("Dominican");
CrawlStore("Ecuador");
CrawlStore("Egypt");
CrawlStore("El");
CrawlStore("Equatorial");
CrawlStore("Eritrea");
CrawlStore("Estonia");
CrawlStore("Ethiopia");
CrawlStore("Falkland");
CrawlStore("Faroe");
CrawlStore("Fiji");
CrawlStore("Finland");
CrawlStore("France");
CrawlStore("French");
CrawlStore("Gabon");
CrawlStore("Gambia");
CrawlStore("Georgia");
CrawlStore("Germany");
CrawlStore("Ghana");
CrawlStore("Gibraltar");
CrawlStore("Greece");
CrawlStore("Greenland");
CrawlStore("Grenada");
CrawlStore("Guadeloupe");
CrawlStore("Guam");
CrawlStore("Guatemala");
CrawlStore("Guinea");
CrawlStore("Guinea");
CrawlStore("Guyana");
CrawlStore("Haiti");
CrawlStore("Holy");
CrawlStore("Honduras");
CrawlStore("Hong");
CrawlStore("Hungary");
CrawlStore("Iceland");
CrawlStore("India");
CrawlStore("Indonesia");
CrawlStore("Iran");
CrawlStore("Iraq");
CrawlStore("Ireland");
CrawlStore("Israel");
CrawlStore("Italy");
CrawlStore("Ivory");
CrawlStore("Jamaica");
CrawlStore("Japan");
CrawlStore("Jordan");
CrawlStore("Kazakhstan");
CrawlStore("Kenya");
CrawlStore("Kiribati");
CrawlStore("Kuwait");
CrawlStore("Kyrgyzstan");
CrawlStore("Laos");
CrawlStore("Latvia");
CrawlStore("Lebanon");
CrawlStore("Lesotho");
CrawlStore("Liberia");
CrawlStore("Libya");
CrawlStore("Liechtenstein");
CrawlStore("Lithuania");
CrawlStore("Luxembourg");
CrawlStore("Macau");
CrawlStore("Macedonia");
CrawlStore("Madagascar");
CrawlStore("Malawi");
CrawlStore("Malaysia");
CrawlStore("Maldives");
CrawlStore("Mali");
CrawlStore("Malta");
CrawlStore("Marshall");
CrawlStore("Martinique");
CrawlStore("Mauritania");
CrawlStore("Mauritius");
CrawlStore("Mayotte");
CrawlStore("Mexico");
CrawlStore("Micronesia");
CrawlStore("Moldova");
CrawlStore("Monaco");
CrawlStore("Mongolia");
CrawlStore("Montenegro");
CrawlStore("Montserrat");
CrawlStore("Morocco");
CrawlStore("Mozambique");
CrawlStore("Myanmar");
CrawlStore("Namibia");
CrawlStore("Nauru");
CrawlStore("Nepal");
CrawlStore("Netherlands");
CrawlStore("Netherlands");
CrawlStore("New");
CrawlStore("New");
CrawlStore("Nicaragua");
CrawlStore("Niger");
CrawlStore("Nigeria");
CrawlStore("Niue");
CrawlStore("Norfolk");
CrawlStore("North");
CrawlStore("Northern");
CrawlStore("Norway");
CrawlStore("Oman");
CrawlStore("Pakistan");
CrawlStore("Palau");
CrawlStore("Panama");
CrawlStore("Papua");
CrawlStore("Paraguay");
CrawlStore("Peru");
CrawlStore("Philippines");
CrawlStore("Pitcairn");
CrawlStore("Poland");
CrawlStore("Polynesia");
CrawlStore("Portugal");
CrawlStore("Puerto");
CrawlStore("Qatar");
CrawlStore("Reunion");
CrawlStore("Romania");
CrawlStore("Russia");
CrawlStore("Rwanda");
CrawlStore("Saint");
CrawlStore("Saint");
CrawlStore("Saint");
CrawlStore("Saint");
CrawlStore("Saint");
CrawlStore("Samoa");
CrawlStore("San");
CrawlStore("Sao");
CrawlStore("Saudi");
CrawlStore("Senegal");
CrawlStore("Serbia");
CrawlStore("Seychelles");
CrawlStore("Sierra");
CrawlStore("Singapore");
CrawlStore("Slovakia");
CrawlStore("Slovenia");
CrawlStore("Solomon");
CrawlStore("Somalia");
CrawlStore("South");
CrawlStore("South");
CrawlStore("South");
CrawlStore("South");
CrawlStore("Spain");
CrawlStore("Sri");
CrawlStore("Sudan");
CrawlStore("Suriname");
CrawlStore("Svalbard");
CrawlStore("Swaziland");
CrawlStore("Sweden");
CrawlStore("Switzerland");
CrawlStore("Syria");
CrawlStore("Taiwan");
CrawlStore("Tajikistan");
CrawlStore("Tanzania");
CrawlStore("Thailand");
CrawlStore("Timor");
CrawlStore("Togo");
CrawlStore("Tokelau");
CrawlStore("Tonga");
CrawlStore("Trinidad");
CrawlStore("Tunisia");
CrawlStore("Turkey");
CrawlStore("Turkmenistan");
CrawlStore("Turks");
CrawlStore("Tuvalu");
CrawlStore("Uganda");
CrawlStore("Ukraine");
CrawlStore("United");
CrawlStore("United");
CrawlStore("United");
CrawlStore("Uruguay");
CrawlStore("Uzbekistan");
CrawlStore("Vanuatu");
CrawlStore("Venezuela");
CrawlStore("Vietnam");
CrawlStore("Virgin");
CrawlStore("Wallis");
CrawlStore("Yemen");
CrawlStore("Zambia");
CrawlStore("Zimbabwe");
}
static void Main(string[] args)
{
// Creating Needed Instances
RequestsHandler httpClient = new RequestsHandler();
AppStoreParser parser = new AppStoreParser();
// Loading Configuration
LogSetup.InitializeLog("Apple_Store_Urls_Worker.log", "info");
_logger = LogManager.GetCurrentClassLogger();
// Loading Config
_logger.Info("Loading Configurations from App.config");
LoadConfiguration();
// Control Variable (Bool - Should the process use proxies? )
bool shouldUseProxies = false;
// Checking for the need to use proxies
if (args != null && args.Length == 1)
{
// Setting flag to true
shouldUseProxies = true;
// Loading proxies from .txt received as argument
String fPath = args[0];
// Sanity Check
if (!File.Exists(fPath))
{
_logger.Fatal("Couldnt find proxies on path : " + fPath);
System.Environment.Exit(-100);
}
// Reading Proxies from File
string[] fLines = File.ReadAllLines(fPath, Encoding.GetEncoding("UTF-8"));
try
{
// Actual Load of Proxies
ProxiesLoader.Load(fLines.ToList());
}
catch (Exception ex)
{
_logger.Fatal(ex);
System.Environment.Exit(-101);
}
}
// AWS Queue Handler
_logger.Info("Initializing Queues");
AWSSQSHelper charactersUrlQueue = new AWSSQSHelper(_characterUrlsQueueName, _maxMessagesPerDequeue, _awsKey, _awsKeySecret);
AWSSQSHelper numericUrlQueue = new AWSSQSHelper(_numericUrlsQueueName, _maxMessagesPerDequeue, _awsKey, _awsKeySecret);
// Setting Error Flag to No Error ( 0 )
System.Environment.ExitCode = 0;
// Initialiazing Control Variables
int fallbackWaitTime = 1;
_logger.Info("Started Processing Character Urls");
do
{
try
{
// Dequeueing messages from the Queue
if (!charactersUrlQueue.DeQueueMessages())
{
Thread.Sleep(_hiccupTime); // Hiccup
continue;
}
// Checking for no message received, and false positives situations
if (!charactersUrlQueue.AnyMessageReceived())
{
// If no message was found, increases the wait time
int waitTime;
if (fallbackWaitTime <= 12)
{
// Exponential increase on the wait time, truncated after 12 retries
waitTime = Convert.ToInt32(Math.Pow(2, fallbackWaitTime) * 1000);
}
else // Reseting Wait after 12 fallbacks
{
waitTime = 2000;
fallbackWaitTime = 0;
}
fallbackWaitTime++;
// Sleeping before next try
Console.WriteLine("Fallback (seconds) => " + waitTime);
Thread.Sleep(waitTime);
continue;
}
// Reseting fallback time
fallbackWaitTime = 1;
// Iterating over dequeued Messages
foreach (var characterUrl in charactersUrlQueue.GetDequeuedMessages())
{
// Console Feedback
_logger.Info("Started Parsing Url : " + characterUrl.Body);
try
{
// Retries Counter
int retries = 0;
string htmlResponse;
// Retrying if necessary
do
{
// Executing Http Request for the Category Url
htmlResponse = httpClient.Get(characterUrl.Body, shouldUseProxies);
if (String.IsNullOrEmpty(htmlResponse))
{
_logger.Info("Retrying Request for Character Page");
retries++;
// Small Hiccup
Thread.Sleep(_hiccupTime);
}
} while (String.IsNullOrWhiteSpace(htmlResponse) && retries <= _maxRetries);
// Checking if retries failed
if (String.IsNullOrWhiteSpace(htmlResponse))
{
// Deletes Message and moves on
charactersUrlQueue.DeleteMessage(characterUrl);
continue;
}
// Hashset of urls processed (to avoid duplicates)
HashSet <String> urlsQueued = new HashSet <String> ();
// Executing Request and Queueing Urls until there's no other Url to be queued
do
{
// Flag to check whether any url was added after the last iteration (avoids endless loop)
bool anyNewUrl = false;
// If the request worked, parses the Urls out of the page
foreach (string numericUrls in parser.ParseNumericUrls(htmlResponse).Select(t => HttpUtility.HtmlDecode(t)))
{
// Checking if this url was previously queued
if (!urlsQueued.Contains(numericUrls))
{
// Enqueueing Urls
numericUrlQueue.EnqueueMessage(HttpUtility.HtmlDecode(numericUrls));
// Adding url to the local hashset
urlsQueued.Add(numericUrls);
anyNewUrl = true;
}
}
// Checking for the need to perform another http request for the next page
if (parser.IsLastPage(htmlResponse) || !anyNewUrl)
{
break; // Breaks "While" Loop
}
// Feedback
_logger.Info("Urls Queued For This Page : " + urlsQueued.Count, "\n\tProcessing Feedback");
// If it got to this point, it means that there are more pages to be processed
// Parsing URL of the "Last" page (the last that's visible)
string lastPageUrl = HttpUtility.HtmlDecode(parser.ParseLastPageUrl(htmlResponse));
// Executing Http Request for this Url (with retries)
retries = 0;
do
{
// HTTP Get for the Page
htmlResponse = httpClient.Get(lastPageUrl, shouldUseProxies);
if (String.IsNullOrEmpty(htmlResponse))
{
_logger.Error("Retrying Request for Last Page");
retries++;
// Small Hiccup
Thread.Sleep(_hiccupTime);
}
} while (String.IsNullOrEmpty(htmlResponse) && retries <= _maxRetries);
} while (true);
}
catch (Exception ex)
{
_logger.Error(ex);
}
finally
{
charactersUrlQueue.DeleteMessage(characterUrl);
}
}
}
catch (Exception ex)
{
_logger.Error(ex);
}
} while (true);
}