/// <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);
// 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;
// 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);
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.Insert<AppModel>(parsedApp))
{
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 = 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);
}
}
}
}
/// <summary>
/// Executes a Search using the searchField as the search parameter,
/// paginates / scrolls the search results to the end adding all the url of apps
/// it finds to a AWS SQS queue
/// </summary>
/// <param name="searchField"></param>
private static void CrawlStore(string searchField)
{
// Console Feedback
Console.WriteLine("Crawling Search Term : [ " + searchField + " ]");
// Compiling Regular Expression used to parse the "pagToken" out of the Play Store
Regex pagTokenRegex = new Regex(@"GAEi+.+\:S\:.{11}\\42", RegexOptions.Compiled);
// HTML Response
string response;
// MongoDB Helper
// 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);
// Ensuring the database has the proper indexe
mongoDB.EnsureIndex("Url");
// Response Parser
PlayStoreParser parser = new PlayStoreParser();
// Executing Web Requests
using (WebRequests server = new WebRequests())
{
// Creating Request Object
server.Host = Consts.HOST;
// Executing Initial Request
response = server.Post(String.Format(Consts.CRAWL_URL, searchField), Consts.INITIAL_POST_DATA);
// Parsing Links out of Html Page (Initial Request)
foreach (string url in parser.ParseAppUrls(response))
{
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed(Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued(url)))
{
// Console Feedback
Console.WriteLine(" . Queued App");
// Than, queue it :)
mongoDB.AddToQueue(url);
Thread.Sleep(250); // Hiccup
}
else
{
// Console Feedback
Console.WriteLine(" . Duplicated App. Skipped");
}
}
// Executing Requests for more Play Store Links
int initialSkip = 48;
int currentMultiplier = 1;
int errorsCount = 0;
do
{
// Finding pagToken from HTML
var rgxMatch = pagTokenRegex.Match(response);
// If there's no match, skips it
if (!rgxMatch.Success)
{
break;
}
// Reading Match from Regex, and applying needed replacements
string pagToken = rgxMatch.Value.Replace(":S:", "%3AS%3A").Replace("\\42", String.Empty).Replace(@"\\u003d", String.Empty);
// Assembling new PostData with paging values
string postData = String.Format(Consts.POST_DATA, pagToken);
// Executing request for values
response = server.Post(String.Format(Consts.CRAWL_URL, searchField), postData);
// Checking Server Status
if (server.StatusCode != System.Net.HttpStatusCode.OK)
{
LogWriter.Error("Http Error", "Status Code [ " + server.StatusCode + " ]");
errorsCount++;
continue;
}
// Parsing Links
foreach (string url in parser.ParseAppUrls(response))
{
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed(Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued(url)))
{
// Console Feedback
Console.WriteLine(" . Queued App");
// Than, queue it :)
mongoDB.AddToQueue(url);
Thread.Sleep(250); // Hiccup
}
else
{
// Console Feedback
Console.WriteLine(" . Duplicated App. Skipped");
}
}
// Incrementing Paging Multiplier
currentMultiplier++;
} while (parser.AnyResultFound(response) && errorsCount <= Consts.MAX_REQUEST_ERRORS);
}
}
/// <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);
}
}
}
}
/// <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 Threshold
LogWriter.Threshold = TLogEventLevel.Information;
LogWriter.Info("Worker Started");
// 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);
// Creating Instance of Web Requests Server
WebRequests server = new WebRequests();
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 = 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
Console.WriteLine("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.Encoding = "utf-8";
server.EncodingDetection = WebRequests.CharsetDetection.DefaultCharset;
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)
{
LogWriter.Info("Error opening app page : " + appUrl);
ProcessingWorked = false;
// Renewing WebRequest Object to get rid of Cookies
server = new WebRequests();
// Inc. retry counter
retryCounter++;
Console.WriteLine("Retrying:" + retryCounter);
// Checking for maximum retry count
double waitTime;
if (retryCounter >= 11)
{
waitTime = TimeSpan.FromMinutes(35).TotalMilliseconds;
// Removing App from the database (this the app page may have expired)
mongoDB.RemoveFromQueue(app.Url);
}
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
Thread.Sleep(Convert.ToInt32(waitTime));
}
else
{
// Reseting retry counter
retryCounter = 0;
// Parsing Useful App Data
AppModel parsedApp = parser.ParseAppPage(response, appUrl);
// Inserting App into Mongo DB Database
if (!mongoDB.Insert <AppModel>(parsedApp))
{
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.WriteLine("Inserted App : " + parsedApp.Name);
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 parser.ParseExtraApps(response))
{
// Incrementing counter of extra apps
extraAppsCounter++;
// Assembling Full app Url to check with database
string 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
Console.WriteLine("Queued " + newExtraApps + " / " + extraAppsCounter + " related apps");
}
}
catch (Exception ex)
{
LogWriter.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
LogWriter.Error(ex);
}
}
}
}
/// <summary>
///
/// </summary>
/// <param name="categoryUrl"></param>
private static void CrawlCategory(string categoryUrl, string categoryName, bool shouldUseProxies)
{
// Console Feedback
_logger.Warn("Crawling Category : [ " + categoryName + " ]");
// Hashset of urls used to keep track of what's been parsed already
HashSet <String> foundUrls = new HashSet <String> ();
// Control variable to avoid "Loop" on pagging
bool isDonePagging = false;
// HTML Response
string response;
// MongoDB Helper
// 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);
// Ensuring the database has the proper indexe
mongoDB.EnsureIndex("Url");
// Response Parser
PlayStoreParser parser = new PlayStoreParser();
// Executing Web Requests
using (WebRequests server = new WebRequests())
{
// Creating Request Object
server.Headers.Add(Consts.ACCEPT_LANGUAGE);
server.Host = Consts.HOST;
server.UserAgent = Consts.GITHUBURL;
server.Encoding = "utf-8";
// Executing Initial Request
response = server.Get(categoryUrl);
// Parsing Links out of Html Page (Initial Request)
foreach (string url in parser.ParseAppUrls(response))
{
// Saving found url on local hashset
foundUrls.Add(url);
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed(Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued(url)))
{
// Than, queue it :)
mongoDB.AddToQueue(url);
}
}
// Executing Requests for more Play Store Links
int baseSkip = 60;
int currentMultiplier = 1;
int errorsCount = 0;
do
{
// Assembling new PostData with paging values
string postData = String.Format(Consts.CATEGORIES_POST_DATA, (currentMultiplier * baseSkip), baseSkip);
// Executing request for values
response = server.Post(String.Format(categoryUrl + "?authuser=0"), postData);
// Checking Server Status
if (server.StatusCode != System.Net.HttpStatusCode.OK)
{
_logger.Error("Http Error" + " - Status Code [ " + server.StatusCode + " ]");
errorsCount++;
continue;
}
// Parsing Links
foreach (string url in parser.ParseAppUrls(response))
{
// If a certain app is found twice, it means that the "pagging" logic got stuck into a
// Loop, so the all the apps for this category were parsed already
if (foundUrls.Contains(url))
{
isDonePagging = true;
break;
}
// Saving found url on local hashset
foundUrls.Add(url);
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed(Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued(url)))
{
// Than, queue it :)
mongoDB.AddToQueue(url);
}
}
// Incrementing Paging Multiplier
currentMultiplier++;
} while (!isDonePagging && errorsCount <= Consts.MAX_REQUEST_ERRORS);
}
}
/// <summary>
/// Executes a Search using the searchField as the search parameter,
/// paginates / scrolls the search results to the end adding all the url of apps
/// it finds to a AWS SQS queue
/// </summary>
/// <param name="searchField"></param>
private static void CrawlStore (string searchField)
{
// Console Feedback
Console.WriteLine ("Crawling Search Term : [ " + searchField + " ]");
// HTML Response
string response;
// MongoDB Helper
// 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);
// Response Parser
PlayStoreParser parser = new PlayStoreParser ();
// Executing Web Requests
using (WebRequests server = new WebRequests ())
{
// Creating Request Object
server.Host = Consts.HOST;
// Executing Initial Request
response = server.Post (Consts.CRAWL_URL, Consts.INITIAL_POST_DATA);
// Parsing Links out of Html Page (Initial Request)
foreach (string url in parser.ParseAppUrls (response))
{
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed (Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued (url)))
{
// Console Feedback
Console.WriteLine (" . Queued App");
// Than, queue it :)
mongoDB.AddToQueue (url);
}
else
{
// Console Feedback
Console.WriteLine (" . Duplicated App. Skipped");
}
}
// Executing Requests for more Play Store Links
int initialSkip = 48;
int currentMultiplier = 1;
int errorsCount = 0;
do
{
// Assembling new PostData with paging values
string postData = String.Format (Consts.POST_DATA, (initialSkip * currentMultiplier));
// Executing request for values
response = server.Post (Consts.CRAWL_URL, postData);
// Checking Server Status
if (server.StatusCode != System.Net.HttpStatusCode.OK)
{
LogWriter.Error ("Http Error", "Status Code [ " + server.StatusCode + " ]");
errorsCount++;
continue;
}
// Parsing Links
foreach (string url in parser.ParseAppUrls (response))
{
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed (Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued (url)))
{
// Console Feedback
Console.WriteLine (" . Queued App");
// Than, queue it :)
mongoDB.AddToQueue (url);
}
else
{
// Console Feedback
Console.WriteLine (" . Duplicated App. Skipped");
}
}
// Incrementing Paging Multiplier
currentMultiplier++;
} while (parser.AnyResultFound (response) && errorsCount <= Consts.MAX_REQUEST_ERRORS);
}
}
/// <summary>
/// Executes a Search using the searchField as the search parameter,
/// paginates / scrolls the search results to the end adding all the url of apps
/// it finds to a AWS SQS queue
/// </summary>
/// <param name="searchField"></param>
private static void CrawlStore (string searchField)
{
// Console Feedback
Console.WriteLine ("Crawling Search Term : [ " + searchField + " ]");
// Compiling Regular Expression used to parse the "pagToken" out of the Play Store
Regex pagTokenRegex = new Regex (@"GAEi+.+\:S\:.{11}\\42", RegexOptions.Compiled);
// HTML Response
string response;
// MongoDB Helper
// 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);
// Ensuring the database has the proper indexe
mongoDB.EnsureIndex ("Url");
// Response Parser
PlayStoreParser parser = new PlayStoreParser ();
// Executing Web Requests
using (WebRequests server = new WebRequests ())
{
// Creating Request Object
server.Host = Consts.HOST;
// Executing Initial Request
response = server.Post (String.Format (Consts.CRAWL_URL, searchField), Consts.INITIAL_POST_DATA);
// Parsing Links out of Html Page (Initial Request)
foreach (string url in parser.ParseAppUrls (response))
{
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed (Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued (url)))
{
// Console Feedback
Console.WriteLine (" . Queued App");
// Than, queue it :)
mongoDB.AddToQueue (url);
Thread.Sleep (250); // Hiccup
}
else
{
// Console Feedback
Console.WriteLine (" . Duplicated App. Skipped");
}
}
// Executing Requests for more Play Store Links
int initialSkip = 48;
int currentMultiplier = 1;
int errorsCount = 0;
do
{
// Finding pagToken from HTML
var rgxMatch = pagTokenRegex.Match (response);
// If there's no match, skips it
if (!rgxMatch.Success)
{
break;
}
// Reading Match from Regex, and applying needed replacements
string pagToken = rgxMatch.Value.Replace (":S:", "%3AS%3A").Replace("\\42", String.Empty).Replace(@"\\u003d", String.Empty);
// Assembling new PostData with paging values
string postData = String.Format (Consts.POST_DATA, pagToken);
// Executing request for values
response = server.Post (String.Format (Consts.CRAWL_URL, searchField), postData);
// Checking Server Status
if (server.StatusCode != System.Net.HttpStatusCode.OK)
{
LogWriter.Error ("Http Error", "Status Code [ " + server.StatusCode + " ]");
errorsCount++;
continue;
}
// Parsing Links
foreach (string url in parser.ParseAppUrls (response))
{
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed (Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued (url)))
{
// Console Feedback
Console.WriteLine (" . Queued App");
// Than, queue it :)
mongoDB.AddToQueue (url);
Thread.Sleep (250); // Hiccup
}
else
{
// Console Feedback
Console.WriteLine (" . Duplicated App. Skipped");
}
}
// Incrementing Paging Multiplier
currentMultiplier++;
} while (parser.AnyResultFound (response) && errorsCount <= Consts.MAX_REQUEST_ERRORS);
}
}
/// <summary>
/// Executes a Search using the searchField as the search parameter,
/// paginates / scrolls the search results to the end adding all the url of apps
/// it finds to a Mongo "QUEUE" collection
/// </summary>
/// <param name="searchField"></param>
private static void CrawlStore(string searchField, bool shouldUseProxies)
{
// Console Feedback
_logger.Warn("Crawling Search Term : [ " + searchField + " ]");
// Hashset of urls used to keep track of what's been parsed already
HashSet <String> foundUrls = new HashSet <String> ();
// Control variable to avoid "Loop" on pagging
bool isDonePagging = false;
// Compiling Regular Expression used to parse the "pagToken" out of the Play Store
Regex pagTokenRegex = new Regex(@"GAEi+.+\:S\:.{11}\\x22", RegexOptions.Compiled);
// HTML Response
string response;
// Response Parser
PlayStoreParser parser = new PlayStoreParser();
// Executing Web Requests
using (WebRequests server = new WebRequests())
{
// Creating Request Object
server.Headers.Add(Consts.ACCEPT_LANGUAGE);
server.Host = Consts.HOST;
server.UserAgent = Consts.GITHUBURL;
server.Encoding = "utf-8";
// Checking for the need to use "HTTP Proxies"
if (shouldUseProxies)
{
server.Proxy = ProxiesLoader.GetWebProxy();
}
// Executing Initial Request
response = server.Post(String.Format(Consts.CRAWL_URL, searchField), Consts.INITIAL_POST_DATA);
// Parsing Links out of Html Page (Initial Request)
foreach (string url in parser.ParseAppUrls(response))
{
// Checks whether the app have been already processed
// or is queued to be processed
foundUrls.Add(url);
if ((!_mongoDB.AppProcessed(Consts.APP_URL_PREFIX + url)) && (!_mongoDB.AppQueued(url)))
{
// Than, queue it :)
_mongoDB.AddToQueue(url);
Thread.Sleep(250); // Hiccup
}
}
// Executing Requests for more Play Store Links
int initialSkip = 48;
int currentMultiplier = 1;
int errorsCount = 0;
do
{
// Finding pagToken from HTML
var rgxMatch = pagTokenRegex.Match(response);
// If there's no match, skips it
if (!rgxMatch.Success)
{
break;
}
// Reading Match from Regex, and applying needed replacements
string pagToken = rgxMatch.Value.Replace(":S:", "%3AS%3A").Replace("\\x22", String.Empty).Replace(@"\\u003d", String.Empty);
// Assembling new PostData with paging values
string postData = String.Format(Consts.POST_DATA, pagToken);
// Executing request for values
response = server.Post(String.Format(Consts.CRAWL_URL, searchField), postData);
// Checking Server Status
if (server.StatusCode != System.Net.HttpStatusCode.OK)
{
_logger.Error("Http Error" + " - Status Code [ " + server.StatusCode + " ]");
errorsCount++;
continue;
}
// Parsing Links
foreach (string url in parser.ParseAppUrls(response))
{
if (foundUrls.Contains(url))
{
isDonePagging = true;
break;
}
// Checks whether the app have been already processed
foundUrls.Add(url);
if ((!_mongoDB.AppProcessed(Consts.APP_URL_PREFIX + url)) && (!_mongoDB.AppQueued(url)))
{
// Than, queue it :)
_mongoDB.AddToQueue(url);
Thread.Sleep(250); // Hiccup
}
}
// Incrementing Paging Multiplier
currentMultiplier++;
} while (!isDonePagging && parser.AnyResultFound(response) && errorsCount <= Consts.MAX_REQUEST_ERRORS);
}
}
/// <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 Threshold
LogWriter.Threshold = TLogEventLevel.Information;
LogWriter.Info ("Worker Started");
// 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);
// Creating Instance of Web Requests Server
WebRequests server = new WebRequests ();
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 = 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
Console.WriteLine("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.Encoding = "utf-8";
server.EncodingDetection = WebRequests.CharsetDetection.DefaultCharset;
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)
{
LogWriter.Info ("Error opening app page : " + appUrl);
ProcessingWorked = false;
// Renewing WebRequest Object to get rid of Cookies
server = new WebRequests ();
// Inc. retry counter
retryCounter++;
Console.WriteLine ("Retrying:" + retryCounter);
// Checking for maximum retry count
double waitTime;
if (retryCounter >= 11)
{
waitTime = TimeSpan.FromMinutes (35).TotalMilliseconds;
// Removing App from the database (this the app page may have expired)
mongoDB.RemoveFromQueue (appUrl);
}
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
Thread.Sleep (Convert.ToInt32 (waitTime));
}
else
{
// Reseting retry counter
retryCounter = 0;
// Parsing Useful App Data
AppModel parsedApp = parser.ParseAppPage (response, appUrl);
// Inserting App into Mongo DB Database
if (!mongoDB.Insert<AppModel>(parsedApp))
{
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.WriteLine("Inserted App : " + parsedApp.Name);
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 parser.ParseExtraApps (response))
{
// Incrementing counter of extra apps
extraAppsCounter++;
// Assembling Full app Url to check with database
string 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
Console.WriteLine ("Queued " + newExtraApps + " / " + extraAppsCounter + " related apps");
}
}
catch (Exception ex)
{
LogWriter.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
LogWriter.Error (ex);
}
}
}
}
/// <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);
}
_logger.Info("\n\nBootstrapping Apps of Past Collections");
// Iterating over past collections
HashSet <String> appUrls = new HashSet <String> ();
foreach (string collection in _mongoDB.GetHistoryOfCollections())
{
_logger.Info("Reading Collection [{0}]", collection);
foreach (string app in _mongoDB.FindAllFromCollectionAs <AppModel> (collection).Select(t => t.Url))
{
if (!appUrls.Contains(app))
{
appUrls.Add(app);
}
}
_logger.Info("\t=> Distinct Apps Found {0}", appUrls.Count);
}
// Adding Distinct Apps to the collection of apps to be processed
int appsQueued = 0;
foreach (string appUrl in appUrls)
{
_mongoDB.AddToQueue(appUrl);
if (appsQueued++ % 10000 == 0)
{
_logger.Info("[Progress] Apps Queued From Past Collections [{0}]", appsQueued);
}
}
}
/// <summary>
///
/// </summary>
/// <param name="categoryUrl"></param>
private static void CrawlCategory (string categoryUrl, string categoryName, bool shouldUseProxies)
{
// Console Feedback
_logger.Warn ("Crawling Category : [ " + categoryName + " ]");
// Hashset of urls used to keep track of what's been parsed already
HashSet<String> foundUrls = new HashSet<String> ();
// Control variable to avoid "Loop" on pagging
bool isDonePagging = false;
// HTML Response
string response;
// MongoDB Helper
// 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);
// Ensuring the database has the proper indexe
mongoDB.EnsureIndex ("Url");
// Response Parser
PlayStoreParser parser = new PlayStoreParser ();
// Executing Web Requests
using (WebRequests server = new WebRequests ())
{
// Creating Request Object
server.Headers.Add (Consts.ACCEPT_LANGUAGE);
server.Host = Consts.HOST;
server.UserAgent = Consts.GITHUBURL;
server.Encoding = "utf-8";
// Executing Initial Request
response = server.Get (categoryUrl);
// Parsing Links out of Html Page (Initial Request)
foreach (string url in parser.ParseAppUrls (response))
{
// Saving found url on local hashset
foundUrls.Add (url);
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed (Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued (url)))
{
// Than, queue it :)
mongoDB.AddToQueue (url);
}
}
// Executing Requests for more Play Store Links
int baseSkip = 60;
int currentMultiplier = 1;
int errorsCount = 0;
do
{
// Assembling new PostData with paging values
string postData = String.Format (Consts.CATEGORIES_POST_DATA, (currentMultiplier * baseSkip), baseSkip);
// Executing request for values
response = server.Post (String.Format (categoryUrl + "?authuser=0"), postData);
// Checking Server Status
if (server.StatusCode != System.Net.HttpStatusCode.OK)
{
_logger.Error ("Http Error" + " - Status Code [ " + server.StatusCode + " ]");
errorsCount++;
continue;
}
// Parsing Links
foreach (string url in parser.ParseAppUrls (response))
{
// If a certain app is found twice, it means that the "pagging" logic got stuck into a
// Loop, so the all the apps for this category were parsed already
if (foundUrls.Contains (url))
{
isDonePagging = true;
break;
}
// Saving found url on local hashset
foundUrls.Add (url);
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed (Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued (url)))
{
// Than, queue it :)
mongoDB.AddToQueue (url);
}
}
// Incrementing Paging Multiplier
currentMultiplier++;
} while (!isDonePagging && errorsCount <= Consts.MAX_REQUEST_ERRORS);
}
}
/// <summary>
/// Executes a Search using the searchField as the search parameter,
/// paginates / scrolls the search results to the end adding all the url of apps
/// it finds to a Mongo "QUEUE" collection
/// </summary>
/// <param name="searchField"></param>
private static void CrawlStore (string searchField, bool shouldUseProxies)
{
// Console Feedback
_logger.Warn ("Crawling Search Term : [ " + searchField + " ]");
// Hashset of urls used to keep track of what's been parsed already
HashSet<String> foundUrls = new HashSet<String> ();
// Control variable to avoid "Loop" on pagging
bool isDonePagging = false;
// Compiling Regular Expression used to parse the "pagToken" out of the Play Store
Regex pagTokenRegex = new Regex (@"GAEi+.+\:S\:.{11}\\42", RegexOptions.Compiled);
// HTML Response
string response;
// MongoDB Helper
// 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);
// Ensuring the database has the proper indexe
mongoDB.EnsureIndex ("Url");
// Response Parser
PlayStoreParser parser = new PlayStoreParser ();
// Executing Web Requests
using (WebRequests server = new WebRequests ())
{
// Creating Request Object
server.Headers.Add (Consts.ACCEPT_LANGUAGE);
server.Host = Consts.HOST;
server.UserAgent = Consts.GITHUBURL;
server.Encoding = "utf-8";
// Checking for the need to use "HTTP Proxies"
if (shouldUseProxies)
{
server.Proxy = ProxiesLoader.GetWebProxy ();
}
// Executing Initial Request
response = server.Post (String.Format (Consts.CRAWL_URL, searchField), Consts.INITIAL_POST_DATA);
// Parsing Links out of Html Page (Initial Request)
foreach (string url in parser.ParseAppUrls (response))
{
// Checks whether the app have been already processed
// or is queued to be processed
foundUrls.Add (url);
if ((!mongoDB.AppProcessed (Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued (url)))
{
// Than, queue it :)
mongoDB.AddToQueue (url);
Thread.Sleep (250); // Hiccup
}
}
// Executing Requests for more Play Store Links
int initialSkip = 48;
int currentMultiplier = 1;
int errorsCount = 0;
do
{
// Finding pagToken from HTML
var rgxMatch = pagTokenRegex.Match (response);
// If there's no match, skips it
if (!rgxMatch.Success)
{
break;
}
// Reading Match from Regex, and applying needed replacements
string pagToken = rgxMatch.Value.Replace (":S:", "%3AS%3A").Replace("\\42", String.Empty).Replace(@"\\u003d", String.Empty);
// Assembling new PostData with paging values
string postData = String.Format (Consts.POST_DATA, pagToken);
// Executing request for values
response = server.Post (String.Format (Consts.CRAWL_URL, searchField), postData);
// Checking Server Status
if (server.StatusCode != System.Net.HttpStatusCode.OK)
{
_logger.Error ("Http Error" + " - Status Code [ " + server.StatusCode + " ]");
errorsCount++;
continue;
}
// Parsing Links
foreach (string url in parser.ParseAppUrls (response))
{
if (foundUrls.Contains (url))
{
isDonePagging = true;
break;
}
// Checks whether the app have been already processed
foundUrls.Add (url);
if ((!mongoDB.AppProcessed (Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued (url)))
{
// Than, queue it :)
mongoDB.AddToQueue (url);
Thread.Sleep (250); // Hiccup
}
}
// Incrementing Paging Multiplier
currentMultiplier++;
} while (!isDonePagging && parser.AnyResultFound (response) && errorsCount <= Consts.MAX_REQUEST_ERRORS);
}
}
/// <summary>
/// Executes a Search using the searchField as the search parameter,
/// paginates / scrolls the search results to the end adding all the url of apps
/// it finds to a AWS SQS queue
/// </summary>
/// <param name="searchField"></param>
private static void CrawlStore(string searchField)
{
// Console Feedback
Console.WriteLine("Crawling Search Term : [ " + searchField + " ]");
// HTML Response
string response;
// MongoDB Helper
// 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);
// Response Parser
PlayStoreParser parser = new PlayStoreParser();
// Executing Web Requests
using (WebRequests server = new WebRequests())
{
// Creating Request Object
server.Host = Consts.HOST;
// Executing Initial Request
response = server.Post(Consts.CRAWL_URL, Consts.INITIAL_POST_DATA);
// Parsing Links out of Html Page (Initial Request)
foreach (string url in parser.ParseAppUrls(response))
{
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed(Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued(url)))
{
// Console Feedback
Console.WriteLine(" . Queued App");
// Than, queue it :)
mongoDB.AddToQueue(url);
}
else
{
// Console Feedback
Console.WriteLine(" . Duplicated App. Skipped");
}
}
// Executing Requests for more Play Store Links
int initialSkip = 48;
int currentMultiplier = 1;
int errorsCount = 0;
do
{
// Assembling new PostData with paging values
string postData = String.Format(Consts.POST_DATA, (initialSkip * currentMultiplier));
// Executing request for values
response = server.Post(Consts.CRAWL_URL, postData);
// Checking Server Status
if (server.StatusCode != System.Net.HttpStatusCode.OK)
{
LogWriter.Error("Http Error", "Status Code [ " + server.StatusCode + " ]");
errorsCount++;
continue;
}
// Parsing Links
foreach (string url in parser.ParseAppUrls(response))
{
// Checks whether the app have been already processed
// or is queued to be processed
if ((!mongoDB.AppProcessed(Consts.APP_URL_PREFIX + url)) && (!mongoDB.AppQueued(url)))
{
// Console Feedback
Console.WriteLine(" . Queued App");
// Than, queue it :)
mongoDB.AddToQueue(url);
}
else
{
// Console Feedback
Console.WriteLine(" . Duplicated App. Skipped");
}
}
// Incrementing Paging Multiplier
currentMultiplier++;
} while (parser.AnyResultFound(response) && errorsCount <= Consts.MAX_REQUEST_ERRORS);
}
}
