made minhash base class and LSHBanding a derivate class

Makefile

@@ -1,6 +1,6 @@
 CC = g++
 CFLAGS = -O2 -DLOCAL -g -fsanitize=undefined,address -Wall -Wshadow -std=c++14
-OBJ = bin/main.o bin/parser.o bin/AhoCorasick.o bin/PolyHash.o bin/MinHash.o bin/Jaccard.o bin/JaccardAhoCorasick.o
+OBJ = bin/main.o bin/parser.o bin/AhoCorasick.o bin/PolyHash.o bin/MinHash.o bin/Jaccard.o bin/JaccardAhoCorasick.o bin/LSHBanding.o
 EXE = bin/LSH.exe
 
 bin/LSH.exe: $(OBJ)
@@ -21,6 +21,9 @@ bin/PolyHash.o: src/PolyHash.cc inc/PolyHash.h
 bin/MinHash.o: src/MinHash.cc inc/MinHash.h inc/PolyHash.h
 	$(CC) -o bin/MinHash.o -c src/MinHash.cc $(CFLAGS) -I ./inc
 
+bin/LSHBanding.o: src/LSHBanding.cc inc/LSHBanding.h inc/MinHash.h
+	$(CC) -o bin/LSHBanding.o -c src/LSHBanding.cc $(CFLAGS) -I ./inc
+
 bin/Jaccard.o: src/Jaccard.cc inc/Jaccard.h
 	$(CC) -o bin/Jaccard.o -c src/Jaccard.cc $(CFLAGS) -I ./inc
 

inc/LSHBanding.h

@@ -0,0 +1,16 @@
+#ifndef LSHBANDING_H
+#define LSHBANDING_H
+#include "MinHash.h"
+#include <set>
+
+class LSHBanding : MinHash {
+public:
+  LSHBanding(vector<vector<string>> shinglesMatrix, int r);
+  vector <pair<int,int>> getSimilarDocuments(double threshold);
+private:
+  vector<pair<int, int> > getCandidatesLSH(int bandWidth);
+  int calculateBandWidth(double threshold);
+  int hashVector(int docIdx, int init, int fin, int value, int mod);
+};
+
+#endif

inc/MinHash.h

@@ -4,21 +4,18 @@
 #include <vector>
 #include <string>
 #include <unordered_set>
+#include <math.h>
 #include "PolyHash.h"
 using namespace std;
 
 class MinHash {
 public:
   MinHash(vector<vector<string>> shinglesMatrix, int r);
-  vector < pair < int, int > > getSimilarDocuments(double threshold);
   double getJaccard(int docIdx1, int docIdx2);
-private:
+protected:
   int numDoc;
   int numPerm;
   vector<vector<int> > signatureMatrix;
-  vector<pair<int, int> > getCandidatesLSH(int bandWidth);
-  int calculateBandWidth(double threshold);
-  int hashVector(int docIdx, int init, int fin, int value, int mod);
 };
 
 #endif

src/LSHBanding.cc

@@ -0,0 +1,51 @@
+#include "LSHBanding.h"
+
+LSHBanding::LSHBanding(vector<vector<string>> shinglesMatrix, int r) : MinHash(shinglesMatrix,r) {}
+
+int LSHBanding::calculateBandWidth(double threshold){
+	double margin = 0.1;
+	if (threshold < margin) {
+		return -1;
+	}
+	for (int bandWidth = 1; bandWidth < numPerm; bandWidth++){
+		double value = pow(1.0/double(numPerm/bandWidth),1.0/double(bandWidth));
+		if (value > threshold - margin) return bandWidth;
+	}
+	return 1;
+}
+
+int LSHBanding::hashVector(int docIdx, int init, int fin, int value, int mod){
+	vector<int> v(fin-init);
+	for (int i = init; i < fin; i++) v[i-init] = signatureMatrix[i][docIdx];
+	PolyHash P = PolyHash(v,mod);
+	return P.evaluate(value);
+}
+
+vector<pair<int,int>> LSHBanding::getCandidatesLSH(int bandWidth){
+	set<pair<int,int>> candidates1;
+	int numBands = numPerm/bandWidth;
+	for (int i = 0; i < numBands; ++i){
+		vector <vector<int>> v (7919);
+		for (int j = 0; j < numDoc; ++j){
+			int hash = hashVector(j,i*bandWidth, (i+1)*bandWidth, 107, 7919);
+			for (int t = 0; t<(int)v[hash].size(); ++t){
+				candidates1.insert({min(v[hash][t], j), max(v[hash][t], j)});
+			}
+			v[hash].push_back(j);
+		}
+	}
+	vector <pair<int,int>> candidates;
+	for (auto t : candidates1) candidates.push_back(t);
+	return candidates;
+}
+
+vector <pair<int,int>> LSHBanding::getSimilarDocuments(double threshold){
+	int bandWidth = calculateBandWidth(threshold);
+	auto candidates = getCandidatesLSH(bandWidth);
+	vector <pair<int,int>> similar;
+	for (auto s : candidates){
+		if (getJaccard(s.first, s.second)) similar.push_back(s);
+	}
+	return similar;
+}
+

src/MinHash.cc

@@ -1,8 +1,5 @@
 #include "MinHash.h"
-#include "PolyHash.h"
-#include <iostream>
-#include <set>
-#include <math.h>
+
 MinHash::MinHash(vector<vector<string>> shinglesMatrix, int r) : numDoc(shinglesMatrix.size()), numPerm(r){
   vector<unordered_set<string>> foundShingles(numDoc);
   vector<string> shingles;
@@ -28,57 +25,9 @@ MinHash::MinHash(vector<vector<string>> shinglesMatrix, int r) : numDoc(shingles
   }
 }
 
-int MinHash::calculateBandWidth(double threshold){
-	double margin = 0.1;
-	if (threshold < margin) {
-		cout << "threshold too low" << endl;
-		return 1;
-	}
-	for (int bandWidth = 1; bandWidth < numPerm; bandWidth++){
-		double value = pow(1.0/double(numPerm/bandWidth),1.0/double(bandWidth));
-		if (value > threshold - margin) return bandWidth;
-	}
-	return 1;
-}
-
-vector < pair < int , int > > MinHash::getSimilarDocuments(double threshold){
-	int bandWidth = calculateBandWidth(threshold);
-	cout << "BandWidth = " << bandWidth << endl;
-	auto candidates = getCandidatesLSH(bandWidth);
-	vector < pair < int, int > > similar;
-	for (auto s : candidates){
-		if (getJaccard(s.first, s.second)) similar.push_back(s);
-	}
-	return similar;
-}
-
 double MinHash::getJaccard(int docIdx1, int docIdx2) {
   int cnt = 0;
   for(int i = 0; i < numPerm; ++i) if(signatureMatrix[i][docIdx1] == signatureMatrix[i][docIdx2]) ++cnt;
   return cnt/double(numPerm);
 }
 
-int MinHash::hashVector(int docIdx, int init, int fin, int value, int mod){
-	vector<int> v(fin-init);
-	for (int i = init; i < fin; i++) v[i-init] = signatureMatrix[i][docIdx];
-	PolyHash P = PolyHash(v,mod);
-	return P.evaluate(value);
-}
-
-vector<pair<int, int> > MinHash::getCandidatesLSH(int bandWidth){
-	set < pair < int , int > > candidates1;
-	int numBands = numPerm/bandWidth;
-	for (int i = 0; i < numBands; ++i){
-		vector < vector < int > > v (7919);
-		for (int j = 0; j < numDoc; ++j){
-			int hash = hashVector(j,i*bandWidth, (i+1)*bandWidth, 107, 7919);
-			for (int t = 0; t<(int)v[hash].size(); ++t){
-				candidates1.insert({min(v[hash][t], j), max(v[hash][t], j)});
-			}
-			v[hash].push_back(j);
-		}
-	}
-	vector < pair < int , int > > candidates;
-	for (auto t : candidates1) candidates.push_back(t);
-	return candidates;
-}

src/main.cc

@@ -5,6 +5,7 @@
 #include "JaccardAhoCorasick.h"
 #include "MinHash.h"
 #include "Jaccard.h"
+#include "LSHBanding.h"
 using namespace std;
 
 int main(int argc, char *argv[]) {
@@ -24,12 +25,14 @@ int main(int argc, char *argv[]) {
     string s2 = doc2.getDocument();
 
     JaccardAhoCorasick JAC({v1,v2},{s1,s2});
-    MinHash M({v1,v2}, 10000);
+    MinHash M({v1,v2}, 5000);
+    LSHBanding LSH({v1,v2},5000);
     Jaccard J({v1,v2});
     cout << JAC.getJaccard(0,1) << endl;
     cout << J.getJaccard(0,1) << endl;
+    cout << M.getJaccard(0,1) << endl;
     double threshold = 0.18;
-    auto candidates = M.getSimilarDocuments(threshold);
+    auto candidates = LSH.getSimilarDocuments(threshold);
     cout << "List of documents more similar than " << threshold << endl;
     for (pair<int,int>& p : candidates){
       cout << p .first << " " << p.second << ' ' << M.getJaccard(p.first, p.second) << endl;