working on shared docMap and docMapLookup

crawler/crawler.cpp

@@ -3,13 +3,13 @@
 //
 
 #include "crawler.h"
+#include <unordered_map>
 
-
-void Crawler::SpawnSpiders( size_t num_spiders )
+void Crawler::SpawnSpiders( size_t num_spiders, unordered_map<string, int> *docMapLookup  )
 	{
 	for ( size_t i = 0; i < num_spiders; i++ )
 		{
-		Spider *temp = new Spider( this->mode, this->urlFrontier);
+		Spider *temp = new Spider( this->mode, this->urlFrontier, docMapLookup);
 		temp->StartThread( );
 		this->spiders.push_back( temp );
 		}

crawler/crawler.h

@@ -4,6 +4,8 @@
 #include "spider.h"
 #include<string>
 #include "../shared/ProducerConsumerQueue.h"
+#include <unordered_map>
+
 //#include "CrawlerStatistics.h"
 /*
  *
@@ -19,7 +21,7 @@ public:
 		{ };
 
 	//spawns a number of works
-	void SpawnSpiders( size_t num_spiders );
+	void SpawnSpiders( size_t num_spiders, unordered_map<string, int> *docMapLookup);
 
 	//Creates a housekeeping thread
 	void houseKeeper();

crawler/spider.cpp

@@ -28,6 +28,9 @@ void Spider::FuncToRun()
 		string currentUrl = getUrl( );
 		char *fileMap;
 
+		shouldURLbeCrawled( currentUrl );
+
+
 		if ( request( currentUrl, fileMap ))
 			{
 			// markURLSeen( currentUrl );
@@ -41,7 +44,65 @@ void Spider::FuncToRun()
 		}
 	}
 
+/*
+Takes a URL. Hashes it. Checks if the url is in the docMapLookup. If it is, check file on disk to see if its been crawled successfully
+ if it has been indexed, (check last time index, maybe reindex?) return false (ignore this url)
+
+ if its not in the docMapLookup, get the current size of the docMap on disk, then calculate new location for this docObjec
+ create a doc object, find a new place, write the document contents to spot. Write the spot to the cache. Return true
+
+*/
+
+bool Spider::shouldURLbeCrawled( string url )
+	{
+
+	auto locationOnDisk = this->docMapLookup->find(url);
+
+	if ( locationOnDisk == this->docMapLookup->end() )
+		{
+		cerr << "Url Not Found In Cache Lookup";
+		int file = getFileDescriptor( "/Users/jakeclose/Desktop/398/project/eecs398-search/docMap.txt" );
+		if ( file )
+			{
+			size_t seekPosition = FileSize( file );
+			off_t resultPosition = lseek( file, seekPosition, SEEK_SET );
+
+			if ( resultPosition == -1 )
+				{
+				cerr << "Could not seek to " << seekPosition <<
+					  ", error = " << errno;
+				return errno;
+				}
+
+			cout << "Current docMap position on disk" << endl;
+			cout << resultPosition << endl;
+
+
+
+
+			}
+		else
+			cerr << "Error opening docMap" << endl;
+
+
+		}
+
+
+	else
+		std::cout << locationOnDisk->first << " is " << locationOnDisk->second;
+
+
+
+
+	}
+
+
+
 
+/*
+returns true if fileMap was created, otherwise false
+ Modifies the filemap to be a char* of the file of the url passed
+*/
 bool Spider::request( string url, char *fileMap )
 	{
 	string localFile;

crawler/spider.h

@@ -7,7 +7,7 @@
 #include "../shared/ProducerConsumerQueue.h"
 #include "../shared/ThreadClass.h"
 #include<iostream>
-
+#include <unordered_map>
 using namespace std;
 
 class Spider : public ThreadClass
@@ -15,8 +15,8 @@ class Spider : public ThreadClass
 
 public:
 
-	Spider( string mode_in, ProducerConsumerQueue < string > *url_q_in )
-			: mode( mode_in ), urlFrontier( url_q_in )
+	Spider( string mode_in, ProducerConsumerQueue < string > *url_q_in , unordered_map<string, int> *doc_map_lookup_in)
+			: mode( mode_in ), urlFrontier( url_q_in ) , docMapLookup(doc_map_lookup_in)
 		{ };
 
 
@@ -39,11 +39,15 @@ public:
 
 	void markURLSeen( string URL );
 
+
+	bool shouldURLbeCrawled( string URL );
+
 private:
 
 	int locationOnDisk;
 	ProducerConsumerQueue < string > *urlFrontier;
 	string mode;
+	unordered_map<string, int> *docMapLookup;
 
 
 	};
\ No newline at end of file

main.cpp

@@ -12,7 +12,7 @@
 #include "crawler/crawler.h"
 #include <string>
 //#include "crawler/CrawlerStatistics.h"
-
+#include <unordered_map>
 
 #define PATH_TO_BLACKLIST = '/bin/blacklist.txt'
 #define PATH_TO_VISITED_URL = 'bin/urls.txt'
@@ -58,10 +58,13 @@ int main(int argc, const char * argv[])
     cout << "Pushed File\n";
     urlFrontier.Push("tests/cats.html");
 
+	unordered_map<string, int>* docMapLookUp = new unordered_map<string, int>();
+
+
 
     Crawler crawler(mode, &urlFrontier);
 
-    crawler.SpawnSpiders(1);
+    crawler.SpawnSpiders(1 , docMapLookUp);
 
     crawler.WaitOnAllSpiders();
 

shared/documentMap.h

@@ -12,8 +12,8 @@ class Document
 	{
 	public:
 		Url url;
-		long docID;
 		string domain;
+		long docID;
 		bool lastCrawlStatus;
 		int lastCrawlDate;
 		int lastCrawlPageCount;

util/util.h

@@ -30,6 +30,15 @@ size_t FileSize( int f )
  *
  */
 
+int getFileDescriptor( string fileName )
+	{
+	return open( fileName.c_str( ), O_RDONLY );
+
+	}
+
+
+
+
 char *getFileMap( string fileName )
 	{
 
@@ -92,4 +101,7 @@ int writeToNewFileToLocation( char *fileContents, string locationOnDisk )
 
 	return fd;
 
-	}
+	}
\ No newline at end of file
+
+
+