public string GetRootPage (bool useProxies)
{
string htmlResponse = String.Empty;
int currentRetry = 0, maxRetries = 100;
using (WebRequests httpClient = new WebRequests())
{
// (Re) Trying to reach Root Page
do
{
// Should this request use HTTP Proxies ?
if (useProxies)
{
httpClient.Proxy = ProxiesLoader.GetWebProxy ();
}
htmlResponse = httpClient.Get (Consts.ROOT_STORE_URL);
currentRetry++;
} while (String.IsNullOrEmpty (htmlResponse) && currentRetry <= maxRetries);
}
return htmlResponse;
}
private static void SearchForContentOnIMDB(WebRequests client, string searchTerm)
{
// First thing we have to do is to define the actual "Target" for our demo, in this case, it will be "www.imdb.com"
// To start Simple, let's try to get the HTML of the Home Page of the site. Since we are RETRIEVING a resource (the home page), the request method needs to be a GET
// But first, let's setup the headers for this request
client.Host = "imdb.com";
client.UserAgent = "Web Crawling 101 Book - Used for educational purposes only";
client.Timeout = 18000; // 18 Seconds
// Issuing the request itself
string normalizedSearchTerm = searchTerm.Replace(' ', '+');
string imdbSearchUrl = String.Format("http://www.imdb.com/find?ref_=nv_sr_fn&q={0}&s=all", normalizedSearchTerm);
Console.WriteLine("GET Request - Searching For '{0}'", searchTerm);
string htmlResponse = client.Get(imdbSearchUrl);
// Sanity Check - Checking if the "Status Code" was "200 - OK", which is what a regular succesful HTTP request should return
Console.WriteLine(" => Request Status Code : " + client.StatusCode);
// Printing Length of the Response (which should be the HTML of the home page)
Console.WriteLine(" => Response Length : " + htmlResponse.Length);
// Writing Response to a file (TIP: Double click on the file to open it with your browser to see if the content is the same as IMDB's home page)
string exeDirectory = Path.GetDirectoryName(System.Reflection.Assembly.GetEntryAssembly().Location);
Console.WriteLine(" => Writing HTML Page to file in : " + exeDirectory);
File.WriteAllText(normalizedSearchTerm + ".html", htmlResponse);
}
public void GetUserPersonalRecord(
string user,
Action <LeaderBoardScoreModel> onSuccess,
Action <string> onError
)
{
string url = USER_URL.Replace("<user>", user);
WebRequests.Get(
$"{url}?order=desc",
(response) => {
var personalRecord = JsonConvert
.DeserializeObject <List <LeaderBoardScoreModel> >(response);
if (personalRecord.Count == 0)
{
onSuccess(null);
return;
}
onSuccess(personalRecord[0]);
},
(error) => {
Debug.Log(error);
onError(error);
}
);
}
public static string Get(ref WebRequests client, Logger logger, string url)
{
int retry = 10;
string htmlResponse = String.Empty;
do
{
// Get html of the current category main page
try
{
htmlResponse = client.Get(url, true);
}
catch (Exception ex)
{
logger.Error(ex);
}
// Sanity check
if (!String.IsNullOrWhiteSpace(htmlResponse) && client.StatusCode == HttpStatusCode.OK)
{
break;
}
retry -= 1;
logger.Debug(String.Format("Status Code not OK. Retries left: {0}", retry));
logger.Debug("StatusCode = " + client.StatusCode + " Message = " + client.Error);
logger.Debug("Html Response = " + htmlResponse);
// Polite Sleeping
Thread.Sleep(TimeSpan.FromSeconds(_rnd.Next(2, 5)));
} while (retry >= 0);
return(htmlResponse);
}
public string Get(string url)
{
using (WebRequests httpClient = new WebRequests())
{
httpClient.UserAgent = Consts.USER_AGENT;
string htmlResponse = httpClient.Get(url);
return(htmlResponse);
}
}
public string Get(string url, bool useProxies)
{
using (WebRequests httpClient = new WebRequests())
{
// Should this request use HTTP Proxies ?
if (useProxies)
{
httpClient.Proxy = ProxiesLoader.GetWebProxy();
}
httpClient.UserAgent = Consts.USER_AGENT;
string htmlResponse = httpClient.Get(url);
return(htmlResponse);
}
}
public string Get (string url, bool useProxies)
{
using (WebRequests httpClient = new WebRequests ())
{
// Should this request use HTTP Proxies ?
if (useProxies)
{
httpClient.Proxy = ProxiesLoader.GetWebProxy ();
}
httpClient.UserAgent = Consts.USER_AGENT;
string htmlResponse = httpClient.Get (url);
return htmlResponse;
}
}
void Start()
{
var url = "http://localhost:3000/";
WebRequests.Get(url,
(string response) => { text.text = response; },
(string error) => { Debug.LogError(error); }
);
var imageUrl = "http://localhost:3000/image.jpg";
WebRequests.GetImage(imageUrl,
(Texture2D texture) => {
sprite.sprite = Sprite.Create(texture, new Rect(0, 0, texture.width, texture.height), new Vector2(.5f, .5f));
},
(string error) => { Debug.LogError(error); }
);
}
public string GetRootPage()
{
string htmlResponse = String.Empty;
int currentRetry = 0, maxRetries = 100;
using (WebRequests client = new WebRequests())
{
// (Re) Trying to reach Root Page
do
{
htmlResponse = client.Get(Consts.ROOT_STORE_URL);
currentRetry++;
} while (String.IsNullOrEmpty(htmlResponse) && currentRetry <= maxRetries);
}
return(htmlResponse);
}
public void GetHighestScores(
int limit,
Action <List <LeaderBoardScoreModel> > onSuccess,
Action <string> onError
)
{
WebRequests.Get(
$"{URL}?order=desc&limit={limit}",
(response) => {
var highestScores = JsonConvert
.DeserializeObject <List <LeaderBoardScoreModel> >(response);
onSuccess(highestScores);
},
(error) => {
Debug.Log(error);
onError(error);
}
);
}
private static void PerformLoginOnPocket(WebRequests client, string username, string password)
{
// Explanation of the requests below:
// In order to perform the LOGIN on Pocket website, we have to issue a POST request with a few parameters in it's BODY.
// Aside from the obvious "Login" and "Password", there's also a variable parameter called "Form_Check".
// This "form_check" parameter can be found on the login page, hidden in it's HTML, so what we are going to do is "Extract" this information
// from there, and use it on the login request to automate the whole login flow.
// First thing we have to do is to define the actual "Target" for our demo, in this case, it will be "www.imdb.com"
// To start Simple, let's try to get the HTML of the Home Page of the site. Since we are RETRIEVING a resource (the home page), the request method needs to be a GET
// But first, let's setup the headers for this request
client.ClearCookies();
client.Host = "getpocket.com";
client.UserAgent = "Web Crawling 101 Book - Used for educational purposes only";
client.Referer = "https://getpocket.com/login?e=4";
client.Origin = "https://getpocket.com";
client.Timeout = 18000; // 18 Seconds
// Reaching Home Page for the "Form Check" parameter
Console.WriteLine(" => Executing GET Request for Home Page");
string homePageHTML = client.Get("https://getpocket.com/login?e=4");
// Parsing "Form Check" parameter
string formCheck = ExtractFormCheckParameter(homePageHTML);
Console.WriteLine(" => Extracted FormCheck parameter hidden within the HTML '{0}'", formCheck);
// Formatting the HTTP POST BODY of the LOGIN Request
string postData = String.Format("feed_id={0}&password={1}&route=&form_check={2}&src=&source=email&source_page=%2Flogin&is_ajax=1",
username, password, formCheck);
// HTTP Post URl for "Login" on Pocket
Console.WriteLine(" => Performing Login");
string pocketLoginUrl = "https://getpocket.com/login_process.php";
string loginResponse = client.Post(pocketLoginUrl, postData);
Console.WriteLine(" => Login Status Code : {0}", client.StatusCode);
Console.WriteLine(" => Login Response Text : {0}", loginResponse);
}
public string GetRootPage(bool useProxies)
{
string htmlResponse = String.Empty;
int currentRetry = 0, maxRetries = 100;
using (WebRequests httpClient = new WebRequests())
{
// (Re) Trying to reach Root Page
do
{
// Should this request use HTTP Proxies ?
if (useProxies)
{
httpClient.Proxy = ProxiesLoader.GetWebProxy();
}
htmlResponse = httpClient.Get(Consts.ROOT_STORE_URL);
currentRetry++;
} while (String.IsNullOrEmpty(htmlResponse) && currentRetry <= maxRetries);
}
return(htmlResponse);
}
/// <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;
}
// Vu
// Check if the app does not meet criteria
if (app.NotMeetCrit)
{
Console.WriteLine("App Not meet Criteria, Skipped.");
}
// 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 >= 7)
{
waitTime = TimeSpan.FromMinutes (35).TotalMilliseconds;
// Removing App from the database (this the app page may have expired)
mongoDB.RemoveFromQueue (app.Url);
Process.Start ("PlayStoreWorker.exe");
Process.GetCurrentProcess ().Kill ();
}
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);
// Vu
// Here is where insert the app into the ProcessedApps Database.
// Attemp to check for the condition base on number of instalation and rating
// First split the string into the string array
string[] installations;
string[] separators = new string[] { " - " };
// Getting the Installation number for the current app
installations = parsedApp.Instalations.Split(separators, StringSplitOptions.RemoveEmptyEntries);
installations[0] = installations[0].Replace(",", ""); // replace the "," in the number of installations
installations[1] = installations[1].Replace(",", "");
long install_num = 0;
try {
install_num = Convert.ToInt64(installations[0]);
}
catch (OverflowException) {
Console.WriteLine("{0} is outside the range of the Int64 type.");
}
catch (FormatException) {
Console.WriteLine("The {0} value '{1}' is not recognizable");
}
bool removed = false;
// Getting the rating for the current app
double rating = parsedApp.Score.Total;
// Getting the developer name ( company name)
string developer = parsedApp.Developer;
// if the installation number is less than 1000,000
// OR rating less than 3 stars
// OR appName is empty
// -> skip the app
string appName = parsedApp.Name;
if (install_num < 1000000 || rating < 3.5 || appName == "" || appName == null)
{
Console.WriteLine("Cannot add app <" + appName + "> -- NOT MEET CRITERIA");
// TODO: Update the NotMeetCriteria
// Removing App from the database
mongoDB.RemoveFromQueue(app.Url);
removed = true;
}
// Inserting App into MONGO_COLLECTION collection
// if the Insert func return false, then print a message indicates that
if (ProcessingWorked && !mongoDB.Insert<AppModel>(parsedApp) && !removed)
{
Console.WriteLine("Cannot add app <" + appName + "> -- FAIL TO ADD TO Database");
ProcessingWorked = false;
}
// If 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
if (!removed)
{
Console.WriteLine("Inserted App : " + parsedApp.Name);
mongoDB.RemoveFromQueue(app.Url);
}
else
{
Console.WriteLine("Removed App : " + parsedApp.Name);
}
}
// Vu
// TRY TO NOT DOWNLOAD THE RELATED APPS
/*
// 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");
*/
// Hiccup (used to minimize blocking issues)
Thread.Sleep (300);
}
}
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);
}
}
}
}
static void Main(string[] args)
{
// Checking for Input Parameters
if (args == null || args.Length != 1)
{
Console.WriteLine("Incorrect number of arguments received. Expected One");
System.Environment.Exit(-100);
}
// Human Readable Variable
string inputFile = args[0];
// Checking if the Input file received exists
if (!File.Exists(inputFile))
{
Console.WriteLine(String.Format("Received input file does not exist : {0}", inputFile));
System.Environment.Exit(-101);
}
// App Status
_appStatus = new Dictionary <String, AppStatusModel> ();
// Creating Instance of Database Manager
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 Parser
PlayStoreParser dataParser = new PlayStoreParser();
goto PeopleData;
using (WebRequests httpClient = new WebRequests())
{
// Minor Configuration of the Http Client - Ensures that the requests response will be in english
// By doing so, we have no problems parsing the dates to their proper formats
httpClient.Headers.Add(Consts.ACCEPT_LANGUAGE);
httpClient.Host = Consts.HOST;
httpClient.Encoding = "utf-8";
httpClient.EncodingDetection = WebRequests.CharsetDetection.DefaultCharset;
// Iterating over File Lines (App Urls) - To Extract Data, Not The Reviews Yet.
foreach (string appUrl in File.ReadAllLines(inputFile))
{
// Logging Progress
Console.WriteLine("\n => Processing App : " + appUrl);
// Executing Http Get Request for the Apps's Data - With max of 5 Retries
String appDataResponse = String.Empty;
int currentRetry = 0;
do
{
// Http Get
appDataResponse = httpClient.Get(appUrl);
} while (String.IsNullOrWhiteSpace(appDataResponse) || ++currentRetry <= _maxRetries);
// Sanity Check
if (String.IsNullOrWhiteSpace(appDataResponse))
{
Console.WriteLine("\t\t.Error - Failed to find page of app : " + appUrl + ". Skipping it");
continue;
}
Console.WriteLine("\t\t.Page Found. Firing Parser");
// Parsing App Data
AppModel appData = dataParser.ParseAppPage(appDataResponse, appUrl);
// Checking If this app is on the database already
if (mongoDB.AppProcessed(appUrl))
{
Console.WriteLine("\t\t.Previous Version of App Found. Updating It");
mongoDB.UpdateRecord(appData, "Url", appData.Url);
// Updating App Status
_appStatus.Add
(
appData.Url,
new AppStatusModel()
{
appId = appData.Url.Replace(Consts.PLAY_STORE_PREFIX, String.Empty),
appUrl = appData.Url,
appName = appData.Name,
status = "Updated"
}
);
}
else
{
Console.WriteLine("\t\t.No Previous Version of the App Found. Adding to Database");
mongoDB.Insert <AppModel> (appData);
// Updating App Status
_appStatus.Add
(
appData.Url,
new AppStatusModel()
{
appId = appData.Url.Replace(Consts.PLAY_STORE_PREFIX, String.Empty),
appUrl = appData.Url,
appName = appData.Name,
status = "Inserted"
}
);
}
}
}
Reviews:
// Next Phase: Parse Reviews of those Apps
Console.WriteLine("\n => Parsing Complete. Obtaining Reviews");
// Iterating again over app urls to parse the reviews from this app
foreach (string appUrl in File.ReadAllLines(inputFile))
{
// Reaching App Id
string appID = _appStatus[appUrl].appId;
// Reviews-Break-Parsing Flag
bool shouldContinueParsing = true;
// Parsing Review Pages from the apps
for (int currentPage = 1; /* no stop condition */; currentPage++)
{
// Getting Reviews Data Bundle
string reviewsData = ReviewsWrapper.GetAppReviews(appID, currentPage);
// Checking for Blocking Situation
if (String.IsNullOrEmpty(reviewsData))
{
Console.WriteLine("Blocked by Play Store. Sleeping process for 10 minutes before retrying.");
// Thread Wait for 10 Minutes
Thread.Sleep(10 * 60 * 1000);
}
// Checking for "No Reviews" app
if (reviewsData.Length < 50)
{
Console.WriteLine("No Reviews left for this app. Skipping");
break;
}
// Normalizing Response to Proper HTML
reviewsData = ReviewsWrapper.NormalizeResponse(reviewsData);
// Iterating over Parsed Reviews
foreach (var review in dataParser.ParseReviews(reviewsData))
{
// Adding App Data to the review
review.appID = _appStatus[appUrl].appId;
review.appName = _appStatus[appUrl].appName;
review.appURL = _appStatus[appUrl].appUrl;
// Incrementing Reviews Count for this app
_appStatus[appUrl].reviews++;
// Adding Review Object to Database
review.timestamp = DateTime.Now;
// Building Query to check for duplicated review
var duplicatedReviewQuery = Query.EQ("permalink", review.permalink);
// Checking for duplicated review before inserting it
if (mongoDB.FindMatch <AppReview> (duplicatedReviewQuery, 1, 0, Consts.REVIEWS_COLLECTION).Count() == 0)
{
// Inserting Review into MongoDB
mongoDB.Insert <AppReview> (review, Consts.REVIEWS_COLLECTION);
}
else
{
Console.WriteLine("Duplicated Review. Skipping App");
// When this happens, there are no more reviews to be parsed
shouldContinueParsing = false; // Skipping this apps processing
}
}
// Hiccup to avoid Blocking problems
Console.WriteLine("Parsed Reviews: " + _appStatus[appUrl].reviews);
Thread.Sleep(new Random().Next(14000, 21000));
if (!shouldContinueParsing)
{
break;
}
}
}
PeopleData:
Console.WriteLine("\n\n => Processing People Data");
Console.WriteLine("\nSimulating Google Login Using Selenium.");
using (var firefoxDriver = new FirefoxDriver())
{
// Navigating to Dummy Url - One that I Know that well be asked for a login
firefoxDriver.Navigate().GoToUrl("https://play.google.com/store/people/details?id=101242565951396343093");
// Reaching Login Fields
var loginField = firefoxDriver.FindElementById("Email");
var passwordField = firefoxDriver.FindElementById("Passwd");
var btnSignIn = firefoxDriver.FindElementById("signIn");
// Sending Credentials to the browser
loginField.SendKeys("YOUREMAIL");
passwordField.SendKeys("YOURPASSWORD");
btnSignIn.Click();
string lastPeople = "https://play.google.com/store/people/details?id=115037241907660526856";
bool shouldcontinue = false;
// Processing Reviewers Data
foreach (string peopleUrl in mongoDB.FindPeopleUrls())
{
// Skipping until last link
if (peopleUrl == lastPeople)
{
shouldcontinue = true;
}
if (!shouldcontinue)
{
continue;
}
// Navigating To the Reviewer Page
firefoxDriver.Navigate().GoToUrl(peopleUrl);
// Executing Get Request for the Reviewer page on Google Play
string reviewerPage = firefoxDriver.PageSource;
// Extracting Reviewer Data from the Page
ReviewerPageData reviewerData = dataParser.ParsePeopleData(reviewerPage);
// Adding Url to the model
reviewerData.reviewerUrl = peopleUrl;
// Inserting it to the database - If no previous record of this Reviewer is found
if (!mongoDB.IsReviewerOnDatabase(peopleUrl))
{
mongoDB.Insert <ReviewerPageData> (reviewerData, "ReviewersData");
}
}
}
// End of Processing + Console Feedback
Console.WriteLine("\n\n == Processing Summary ==");
foreach (var status in _appStatus.Select(t => t.Value))
{
// Message
string cMessage = "=> App : {0} - Status {1} - Reviews : {2}";
Console.WriteLine(String.Format(cMessage, status.appName, status.status, status.reviews));
}
Console.ReadLine();
}
/// <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>
///
/// </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>
/// 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);
}
}
}
}
static void Main (string[] args)
{
// Checking for Input Parameters
if (args == null || args.Length != 1)
{
Console.WriteLine ("Incorrect number of arguments received. Expected One");
System.Environment.Exit (-100);
}
// Human Readable Variable
string inputFile = args[0];
// Checking if the Input file received exists
if (!File.Exists (inputFile))
{
Console.WriteLine (String.Format("Received input file does not exist : {0}", inputFile));
System.Environment.Exit (-101);
}
// App Status
_appStatus = new Dictionary<String, AppStatusModel> ();
// Creating Instance of Database Manager
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 Parser
PlayStoreParser dataParser = new PlayStoreParser ();
goto PeopleData;
using (WebRequests httpClient = new WebRequests ())
{
// Minor Configuration of the Http Client - Ensures that the requests response will be in english
// By doing so, we have no problems parsing the dates to their proper formats
httpClient.Headers.Add (Consts.ACCEPT_LANGUAGE);
httpClient.Host = Consts.HOST;
httpClient.Encoding = "utf-8";
httpClient.EncodingDetection = WebRequests.CharsetDetection.DefaultCharset;
// Iterating over File Lines (App Urls) - To Extract Data, Not The Reviews Yet.
foreach (string appUrl in File.ReadAllLines (inputFile))
{
// Logging Progress
Console.WriteLine ("\n => Processing App : " + appUrl);
// Executing Http Get Request for the Apps's Data - With max of 5 Retries
String appDataResponse = String.Empty;
int currentRetry = 0;
do
{
// Http Get
appDataResponse = httpClient.Get (appUrl);
} while (String.IsNullOrWhiteSpace(appDataResponse) || ++currentRetry <= _maxRetries);
// Sanity Check
if (String.IsNullOrWhiteSpace (appDataResponse))
{
Console.WriteLine ("\t\t.Error - Failed to find page of app : " + appUrl + ". Skipping it");
continue;
}
Console.WriteLine("\t\t.Page Found. Firing Parser");
// Parsing App Data
AppModel appData = dataParser.ParseAppPage (appDataResponse, appUrl);
// Checking If this app is on the database already
if (mongoDB.AppProcessed (appUrl))
{
Console.WriteLine ("\t\t.Previous Version of App Found. Updating It");
mongoDB.UpdateRecord (appData, "Url", appData.Url);
// Updating App Status
_appStatus.Add
(
appData.Url,
new AppStatusModel ()
{
appId = appData.Url.Replace (Consts.PLAY_STORE_PREFIX, String.Empty),
appUrl = appData.Url,
appName = appData.Name,
status = "Updated"
}
);
}
else
{
Console.WriteLine ("\t\t.No Previous Version of the App Found. Adding to Database");
mongoDB.Insert<AppModel> (appData);
// Updating App Status
_appStatus.Add
(
appData.Url,
new AppStatusModel ()
{
appId = appData.Url.Replace (Consts.PLAY_STORE_PREFIX, String.Empty),
appUrl = appData.Url,
appName = appData.Name,
status = "Inserted"
}
);
}
}
}
Reviews:
// Next Phase: Parse Reviews of those Apps
Console.WriteLine ("\n => Parsing Complete. Obtaining Reviews");
// Iterating again over app urls to parse the reviews from this app
foreach (string appUrl in File.ReadAllLines (inputFile))
{
// Reaching App Id
string appID = _appStatus[appUrl].appId;
// Reviews-Break-Parsing Flag
bool shouldContinueParsing = true;
// Parsing Review Pages from the apps
for (int currentPage = 1; /* no stop condition */; currentPage++)
{
// Getting Reviews Data Bundle
string reviewsData = ReviewsWrapper.GetAppReviews (appID, currentPage);
// Checking for Blocking Situation
if (String.IsNullOrEmpty (reviewsData))
{
Console.WriteLine("Blocked by Play Store. Sleeping process for 10 minutes before retrying.");
// Thread Wait for 10 Minutes
Thread.Sleep (10 * 60 * 1000);
}
// Checking for "No Reviews" app
if (reviewsData.Length < 50)
{
Console.WriteLine ("No Reviews left for this app. Skipping");
break;
}
// Normalizing Response to Proper HTML
reviewsData = ReviewsWrapper.NormalizeResponse (reviewsData);
// Iterating over Parsed Reviews
foreach (var review in dataParser.ParseReviews (reviewsData))
{
// Adding App Data to the review
review.appID = _appStatus[appUrl].appId;
review.appName = _appStatus[appUrl].appName;
review.appURL = _appStatus[appUrl].appUrl;
// Incrementing Reviews Count for this app
_appStatus[appUrl].reviews++;
// Adding Review Object to Database
review.timestamp = DateTime.Now;
// Building Query to check for duplicated review
var duplicatedReviewQuery = Query.EQ ("permalink", review.permalink);
// Checking for duplicated review before inserting it
if (mongoDB.FindMatch<AppReview> (duplicatedReviewQuery, 1, 0, Consts.REVIEWS_COLLECTION).Count () == 0)
{
// Inserting Review into MongoDB
mongoDB.Insert<AppReview> (review, Consts.REVIEWS_COLLECTION);
}
else
{
Console.WriteLine ("Duplicated Review. Skipping App");
// When this happens, there are no more reviews to be parsed
shouldContinueParsing = false; // Skipping this apps processing
}
}
// Hiccup to avoid Blocking problems
Console.WriteLine ("Parsed Reviews: " + _appStatus[appUrl].reviews);
Thread.Sleep (new Random ().Next (14000, 21000));
if (!shouldContinueParsing)
{
break;
}
}
}
PeopleData:
Console.WriteLine ("\n\n => Processing People Data");
Console.WriteLine ("\nSimulating Google Login Using Selenium.");
using (var firefoxDriver = new FirefoxDriver ())
{
// Navigating to Dummy Url - One that I Know that well be asked for a login
firefoxDriver.Navigate ().GoToUrl ("https://play.google.com/store/people/details?id=101242565951396343093");
// Reaching Login Fields
var loginField = firefoxDriver.FindElementById ("Email");
var passwordField = firefoxDriver.FindElementById ("Passwd");
var btnSignIn = firefoxDriver.FindElementById ("signIn");
// Sending Credentials to the browser
loginField.SendKeys ("YOUREMAIL");
passwordField.SendKeys ("YOURPASSWORD");
btnSignIn.Click ();
string lastPeople = "https://play.google.com/store/people/details?id=115037241907660526856";
bool shouldcontinue = false;
// Processing Reviewers Data
foreach (string peopleUrl in mongoDB.FindPeopleUrls ())
{
// Skipping until last link
if (peopleUrl == lastPeople)
{
shouldcontinue = true;
}
if (!shouldcontinue) continue;
// Navigating To the Reviewer Page
firefoxDriver.Navigate ().GoToUrl (peopleUrl);
// Executing Get Request for the Reviewer page on Google Play
string reviewerPage = firefoxDriver.PageSource;
// Extracting Reviewer Data from the Page
ReviewerPageData reviewerData = dataParser.ParsePeopleData (reviewerPage);
// Adding Url to the model
reviewerData.reviewerUrl = peopleUrl;
// Inserting it to the database - If no previous record of this Reviewer is found
if (!mongoDB.IsReviewerOnDatabase (peopleUrl))
{
mongoDB.Insert<ReviewerPageData> (reviewerData, "ReviewersData");
}
}
}
// End of Processing + Console Feedback
Console.WriteLine ("\n\n == Processing Summary ==");
foreach (var status in _appStatus.Select (t => t.Value))
{
// Message
string cMessage = "=> App : {0} - Status {1} - Reviews : {2}";
Console.WriteLine (String.Format (cMessage, status.appName, status.status, status.reviews));
}
Console.ReadLine ();
}
/// <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);
}
}
}
}
private static void ParseAppUrls(ISet <string> urls, int downloadDelay = 0, IExporter exporter = null,
Action <AppModel> writeCallback = null)
{
log.Info("Parsing App URLs...");
int parsedAppCount = 0;
// Retry Counter (Used for exponential wait increasing logic)
int retryCounter = 0;
// Creating Instance of Web Requests Server
WebRequests server = new WebRequests();
foreach (string url in urls)
{
try
{
// Building APP URL
string appUrl = Consts.APP_URL_PREFIX + url;
// Configuring server and Issuing Request
server.Headers.Add(Consts.ACCEPT_LANGUAGE);
server.Host = Consts.HOST;
server.Encoding = "utf-8";
server.EncodingDetection = WebRequests.CharsetDetection.DefaultCharset;
// this is how we actually connect to all this shit
// the only thing left - we need to randomize it and check if 200
//WebProxy proxyObject = new WebProxy("http://" + ProxyLoader.ReturnRandomProxy(), true);
//server.Proxy = proxyObject;
string response = server.Get(appUrl);
// Sanity Check
if (String.IsNullOrEmpty(response) || server.StatusCode != System.Net.HttpStatusCode.OK)
{
log.Info("Error opening app page : " + appUrl);
// Renewing WebRequest Object to get rid of Cookies
server = new WebRequests();
// Inc. retry counter
retryCounter++;
log.Info("Retrying:" + retryCounter);
// Checking for maximum retry count
double waitTime;
if (retryCounter >= 11)
{
waitTime = TimeSpan.FromMinutes(35).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
Thread.Sleep(Convert.ToInt32(waitTime));
}
else
{
// Reseting retry counter
retryCounter = 0;
// Parsing App Data
AppModel parsedApp = parser.ParseAppPage(response, appUrl);
// Export the App Data
if (exporter != null)
{
log.Info("Parsed App: " + parsedApp.Name);
exporter.Write(parsedApp);
}
// Pass the App Data to callback method
if (writeCallback != null)
{
writeCallback(parsedApp);
}
// Default action is print to screen
if (exporter == null && writeCallback == null)
{
Console.WriteLine(parsedApp);
}
++parsedAppCount;
// Apply download delay
if (downloadDelay > 0)
{
Thread.Sleep(downloadDelay);
}
}
}
catch (Exception ex)
{
log.Error(ex);
Console.WriteLine(url);
}
}
log.Info("Finished. Parsed App count: " + parsedAppCount + "\n");
}
/// <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);
}
}
}
}
private static void Execute(FlexibleOptions options)
{
logger.Info ("Start");
// Sanity Checks
string inputPath = options.Get("inputFile");
string outputPath = options.Get("outputFile");
if (String.IsNullOrEmpty(inputPath))
{
logger.Fatal ("Parameter {--inputfile} is mandatory");
System.Environment.Exit (-101);
}
if (String.IsNullOrEmpty (inputPath))
{
logger.Fatal ("Parameter {--outputFile} is mandatory");
System.Environment.Exit (-102);
}
if (!File.Exists(inputPath))
{
logger.Fatal ("Parameter {--inputfile} should point to an existing file");
System.Environment.Exit (-103);
}
// Control Variables
int processed = 0, errors = 0;
// Processing Urls
using(StreamReader fReader = new StreamReader (inputPath))
{
using (StreamWriter fWriter = new StreamWriter (outputPath))
{
fWriter.AutoFlush = true;
string fLine;
using (WebRequests client = new WebRequests ())
{
// Configuring HTTP Client
client.Referer = "http://sortfolio.com/";
client.Host = "sortfolio.com";
client.Encoding = "gzip, deflate, sdch";
client.Headers.Add ("Upgrade-Insecure-Requests", "1");
while ((fLine = fReader.ReadLine ()) != null)
{
// Processing Only "Listings"
if (fLine.IndexOf ("/listings/") >= 0)
{
try
{
// Reaching Listing Page
string url = fLine.Split (';')[0];
string htmlReponse = client.Get (url);
// Control Variable Inc.
processed++;
// Sanity Check
if (String.IsNullOrEmpty (htmlReponse))
{
logger.Error ("Failed to read response for {0}", url);
errors++;
continue;
}
// Parsing Data
HtmlDocument map = new HtmlDocument ();
map.LoadHtml (htmlReponse);
var tmpNode = map.DocumentNode.SelectSingleNode ("//script[contains(@id,'mail_to')]");
if (tmpNode == null)
{
continue;
}
string email = tmpNode.InnerText;
// Extracting actual email data out of decoded string
email = email.Replace ("eval(decodeURIComponent('", String.Empty).Replace ("'", String.Empty).Replace (")", String.Empty);
email = System.Web.HttpUtility.UrlDecode (email);
Regex emailRegex = new Regex (@"Node\('(.+?)'\)\);");
email = emailRegex.Match (email).Groups[1].Value;
tmpNode = map.DocumentNode.SelectSingleNode ("//*[@id='content']/div/div[2]/p/strong");
string budget = tmpNode == null ? String.Empty : tmpNode.InnerText;
tmpNode = map.DocumentNode.SelectSingleNode ("//span[@itemprop='addressLocality']");
string headQuarter = tmpNode == null ? String.Empty : tmpNode.InnerText;
tmpNode = map.DocumentNode.SelectSingleNode ("//a[@class='website']");
string website = tmpNode == null ? String.Empty : tmpNode.InnerText;
// Writing to output file
fWriter.WriteLine ("{0};{1};{2};{3};{4}", url, headQuarter, website, email, budget);
logger.Info ("Processed: {0}", url);
if (processed % 100 == 0)
{
logger.Info ("\tProgress:{0} - {1}", processed, errors);
}
}
catch(Exception ex)
{
logger.Error (ex);
errors++;
}
}
}
}
}
}
logger.Info ("End");
Console.ReadLine ();
}
