X-Git-Url: https://git.llucax.com/z.facultad/75.52/treemulator.git/blobdiff_plain/2c1d3ea3cf9a25e8f957587d4a1581fc292c831b..d1cfb7971a17f744717953a0466ee2efbd53f442:/src/btree.cpp

diff --git a/src/btree.cpp b/src/btree.cpp
index cf01b2c..a063bef 100644
--- a/src/btree.cpp
+++ b/src/btree.cpp
@@ -1,12 +1,13 @@
 
 #include "btree.h"
 
-BTree::BTree (const std::string &name, unsigned int block_size, bool create_new_file)
+BTree::BTree (const std::string &name, unsigned int block_size, int kt, bool create_new_file)
 {
+	key_type = kt;
 	uchar *node;
 	BTreeNodeHeader nh;
 
-	fp = fopen (name.c_str(), "wb");
+	fp = fopen (name.c_str(), "wb+");
 	if (!fp) {
 		/* TODO : mandar una exception ? */
 		return;
@@ -44,19 +45,197 @@ void BTree::WriteFileHeader ()
 
 void BTree::WriteBlock (uchar *block, uint num)
 {
-	fseek (fp, num*header.block_size + sizeof (BTreeFileHeader), SEEK_SET);
+	num++;
+	fseek (fp, num*header.block_size, SEEK_SET);
 	fwrite (block, 1, header.block_size, fp);
 }
 
 void BTree::AddKey (const Clave &k)
 {
-	AddKeyR (&k, 0);
+	uint left, right;
+	Clave *kout = AddKeyR (k.Clone (), 0, left, right);
+
+	if (kout) {
+		unsigned short level;
+		/* Debo dejar la raiz en el nodo 0, por lo que paso el nodo
+		 * que esta usando el hijo izquierdo a un nuevo nodo */
+		std::list<BTreeData *> node_keys;
+		BTreeNodeHeader node_header;
+		uchar *node = ReadBlock (left);
+		ReadNodoHeader (node, &node_header);
+		node_keys = ReadKeys (node, node_header);
+		level = node_header.level + 1;
+
+		uchar *new_node = NewBlock (left);
+		delete [] new_node; /* No me interesa, voy a usar lo leio antes */
+		
+		WriteKeys (node, node_header, node_keys);
+		WriteNodoHeader (node, &node_header);
+		WriteBlock (node, left);
+		DeleteKeys (node_keys);
+		delete [] node;
+
+		/* Leo y actualizo la Raiz */
+		node = ReadBlock (0);
+		ReadNodoHeader (node, &node_header);
+		node_keys = std::list<BTreeData *>();
+
+		node_keys.push_back (new BTreeChildData (left));
+		node_keys.push_back (new BTreeData (kout, right));
+
+		node_header.level = level;
+		node_header.item_count = 1;
+
+		WriteKeys (node, node_header, node_keys);
+		WriteNodoHeader (node, &node_header);
+		WriteBlock (node, 0);
+		delete [] node;
+		DeleteKeys (node_keys);
+		PrintNode (0);
+	}
+}
+
+Clave* BTree::AddKeyR (const Clave *k, uint node_num, uint &left_child, uint &right_child)
+{
+	uchar *node = ReadBlock (node_num);
+	BTreeNodeHeader node_header;
+	ReadNodoHeader (node, &node_header);
+	delete [] node;
+
+	if (node_header.level == 0)
+		return AddKeyLeafR (k, node_num, left_child, right_child);
+
+	return AddKeyOtherR (k, node_num, left_child, right_child);
 }
 
-Clave* BTree::AddKeyR (const Clave *k, uint node_num)
+Clave* BTree::AddKeyLeafR (const Clave *k, uint node_num, uint &left_child, uint &right_child)
 {
-	Clave *kout;
-	std::list<BTreeData> node_keys;
+	Clave *kout = NULL;
+	std::list<BTreeData *> node_keys;
+
+	BTreeData *data = new BTreeLeafData (k->Clone ());
+
+	/* Leo el nodo raiz para empezar a agregar */
+	uchar *node = ReadBlock (node_num);
+	BTreeNodeHeader node_header;
+	ReadNodoHeader (node, &node_header);
+
+	if (node_header.free_space > data->Size ()) {
+		BTreeData *datait;
+		node_keys = ReadKeys (node, node_header);
+		std::list<BTreeData *>::iterator it = node_keys.begin ();
+
+		while (it != node_keys.end ()) {
+			datait = (*it);
+			if ((*data) < (*datait))
+				/* Me pase, lo agrego aca! */
+				break;
+			it++;
+		}
+		node_keys.insert (it, data);
+		WriteKeys (node, node_header, node_keys);
+		WriteNodoHeader (node, &node_header);
+		WriteBlock (node, node_num);
+		DeleteKeys (node_keys);
+		delete [] node;
+
+		PrintNode (node_num);
+	} else {
+		/* Split : Creo e inicializo el nuevo nodo */
+		std::list<BTreeData *> new_node_keys;
+		std::list<BTreeData *> old_node_keys;
+		BTreeNodeHeader new_node_header;
+		uint new_node_num;
+		uchar *new_node = NewBlock (new_node_num);
+		ReadNodoHeader (new_node, &new_node_header);
+		new_node_header.level = node_header.level;
+
+		node_keys = ReadKeys (node, node_header);
+		new_node_keys = ReadKeys (new_node, new_node_header);
+
+		/* Agrego la clave en la lista que ya tengo de manera ordenada */
+		std::list<BTreeData *>::iterator it = node_keys.begin ();
+		std::list<BTreeData *>::iterator previt = node_keys.begin ();
+
+		while (it != node_keys.end ()) {
+			BTreeData *datait;
+			datait = (*it);
+			if ((*data) < (*datait))
+				/* Me pase, lo agrego aca! */
+				break;
+			previt = it;
+			it++;
+		}
+		if (it != node_keys.end ())
+			node_keys.insert (it, data);
+		else
+			node_keys.push_back (data);
+
+		/* Tengo que guardar claves hasta ocupar nodo size/2 en cada nodo
+		 * y subir la clave del medio */
+		node_header.item_count = 0;
+		node_header.free_space = header.block_size - sizeof (BTreeNodeHeader);
+
+		uint total_size = 0;
+		it = node_keys.begin ();
+		while (it != node_keys.end ()) {
+			BTreeData *datait;
+			datait = (*it);
+			total_size += datait->Size ();
+			it++;
+			/* Hack : Si me quedo con todas las claves, en el caso de ser
+			 * del mismo tama#o se desbalancea. Hay que ver que efecto 
+			 * puede tener en el caso de claves de long. variable
+			 */
+			if (it == node_keys.end ())
+				total_size -= datait->Size ();
+		}
+
+		it = node_keys.begin ();
+		uint used = 0;
+		while (used < total_size/2) {
+			BTreeData *d = (*it);
+			old_node_keys.push_back (d);
+			used += d->Size ();
+			it++;
+		}
+		kout = (*it++)->getClave (); // Esta se retorna al "padre" para que se la agregue
+
+		while (it != node_keys.end ()) {
+			BTreeData *d = (*it);
+			new_node_keys.push_back (d);
+			it++;
+		}
+	
+		/* Guardo */
+		WriteKeys (node, node_header, old_node_keys);
+		WriteNodoHeader (node, &node_header);
+		WriteBlock (node, node_num);
+		WriteKeys (new_node, new_node_header, new_node_keys);
+		WriteNodoHeader (new_node, &new_node_header);
+		WriteBlock (new_node, new_node_num);
+		DeleteKeys (old_node_keys);
+		DeleteKeys (new_node_keys);
+
+		PrintNode (node_num);
+		PrintNode (new_node_num);
+
+		/* Paso los hijos */
+		left_child = node_num;
+		right_child = new_node_num;
+		delete [] new_node;
+		delete [] node;
+	}
+
+	return kout;
+}
+
+Clave* BTree::AddKeyOtherR (const Clave *k, uint node_num, uint &left_child, uint &right_child)
+{
+	Clave *kout = NULL;
+	std::list<BTreeData *> node_keys;
+
+	BTreeData *data = new BTreeLeafData (k->Clone ());
 
 	/* Leo el nodo raiz para empezar a agregar */
 	uchar *node = ReadBlock (node_num);
@@ -64,20 +243,148 @@ Clave* BTree::AddKeyR (const Clave *k, uint node_num)
 	ReadNodoHeader (node, &node_header);
 
 	node_keys = ReadKeys (node, node_header);
+
+	std::list<BTreeData *>::iterator it = node_keys.begin ();
+	std::list<BTreeData *>::iterator posterior;
+	std::list<BTreeData *>::iterator ultima;
 	
-	if (node_header.free_space >= k->Size ()) {
-		/* TODO : Insertar ordenado */
-		//node_keys.push_back (BTreek->Clone ());
+	/* Se supone que la primera es un hijo :) */
+	BTreeData *lchild = (*it++);
+	posterior = it;
+
+	while (it != node_keys.end ()) {
+		if ((*data) < (*(*it)))
+			break;
+		ultima = it;
+		it++;
+	}
+
+	if (it == posterior) {
+		k = AddKeyR (k, lchild->getChild (), left_child, right_child);
 	} else {
-		/* Split */
+		k = AddKeyR (k, (*ultima)->getChild (), left_child, right_child);
+	}
+	DeleteKeys (node_keys);
+
+	/* Nada que hacer */
+	if (data) delete data;
+	if (!k) {
+		delete [] node;
+		return NULL;
 	}
 
+	data = new BTreeData (k->Clone (), right_child);
+
+	if (node_header.free_space > data->Size ()) {
+		BTreeData *datait;
+		node_keys = ReadKeys (node, node_header);
+		std::list<BTreeData *>::iterator it = node_keys.begin ();
+
+		while (it != node_keys.end ()) {
+			datait = (*it);
+			if ((*data) < (*datait))
+				/* Me pase, lo agrego aca! */
+				break;
+			it++;
+		}
+		node_keys.insert (it, data);
+		WriteKeys (node, node_header, node_keys);
+		WriteNodoHeader (node, &node_header);
+		WriteBlock (node, node_num);
+		DeleteKeys (node_keys);
+		delete [] node;
+
+		PrintNode (node_num);
+	} else {
+		/* Split : Creo e inicializo el nuevo nodo */
+		std::list<BTreeData *> new_node_keys;
+		std::list<BTreeData *> old_node_keys;
+		BTreeNodeHeader new_node_header;
+		uint new_node_num;
+		uchar *new_node = NewBlock (new_node_num);
+		ReadNodoHeader (new_node, &new_node_header);
+		new_node_header.level = node_header.level;
+
+		node_keys = ReadKeys (node, node_header);
+		new_node_keys = ReadKeys (new_node, new_node_header);
+
+		/* Agrego la clave en la lista que ya tengo de manera ordenada */
+		std::list<BTreeData *>::iterator it = node_keys.begin ();
+		std::list<BTreeData *>::iterator previt = node_keys.begin ();
+
+		previt = ++it;
 	
-	WriteKeys (node, node_header, node_keys);
-	WriteNodoHeader (node, &node_header);
-	WriteBlock (node, node_num);
+		while (it != node_keys.end ()) {
+			BTreeData *datait;
+			datait = (*it);
+			if ((*data) < (*datait))
+				/* Me pase, lo agrego aca! */
+				break;
+			previt = it;
+			it++;
+		}
+		if (it != node_keys.end ())
+			node_keys.insert (it, data);
+		else
+			node_keys.push_back (data);
 
-	delete [] node;
+		/* Tengo que guardar claves hasta ocupar nodo size/2 en cada nodo
+		 * y subir la clave del medio */
+		node_header.item_count = 0;
+		node_header.free_space = header.block_size - sizeof (BTreeNodeHeader);
+
+		uint total_size = 0;
+		it = node_keys.begin ();
+		while (it != node_keys.end ()) {
+			BTreeData *datait;
+			datait = (*it);
+			total_size += datait->Size ();
+			it++;
+			/* Hack : Si me quedo con todas las claves, en el caso de ser
+			 * del mismo tama#o se desbalancea. Hay que ver que efecto 
+			 * puede tener en el caso de claves de long. variable
+			 */
+			if (it == node_keys.end ())
+				total_size -= datait->Size ();
+		}
+
+		it = node_keys.begin ();
+		uint used = 0;
+		while (used < total_size/2) {
+			BTreeData *d = (*it);
+			old_node_keys.push_back (d);
+			used += d->Size ();
+			it++;
+		}
+		kout = (*it)->getClave (); // Esta se retorna al "padre" para que se la agregue
+
+		new_node_keys.push_back ( new BTreeChildData ((*it)->getChild ()));
+		it++;
+		while (it != node_keys.end ()) {
+			BTreeData *d = (*it);
+			new_node_keys.push_back (d);
+			it++;
+		}
+	
+		/* Guardo */
+		WriteKeys (node, node_header, old_node_keys);
+		WriteNodoHeader (node, &node_header);
+		WriteBlock (node, node_num);
+		WriteKeys (new_node, new_node_header, new_node_keys);
+		WriteNodoHeader (new_node, &new_node_header);
+		WriteBlock (new_node, new_node_num);
+		DeleteKeys (old_node_keys);
+		DeleteKeys (new_node_keys);
+
+		PrintNode (node_num);
+		PrintNode (new_node_num);
+
+		/* Paso los hijos */
+		left_child = node_num;
+		right_child = new_node_num;
+		delete [] new_node;
+		delete [] node;
+	}
 
 	return kout;
 }
@@ -96,45 +403,184 @@ void BTree::WriteNodoHeader (uchar *node, BTreeNodeHeader *header)
 
 uchar *BTree::ReadBlock (uint num)
 {
+	/* Como el bloque 0 se usa para el header, el Nodo "num"
+	 * está en el bloque "num+1"
+	 */
+	num++;
+
 	uchar *out = new uchar[header.block_size];
 
-	fseek (fp, num*header.block_size + sizeof (BTreeFileHeader), SEEK_SET);	
+	fseek (fp, num*header.block_size, SEEK_SET);	
 	fread (out, 1, header.block_size, fp);
 
 	return out;
 }
 
-std::list<BTreeData> BTree::ReadKeys (uchar *node, BTreeNodeHeader &node_header)
+std::list<BTreeData *> BTree::ReadKeys (uchar *node, BTreeNodeHeader &node_header)
 {
-	std::list<BTreeData> keys;
+	std::list<BTreeData *> keys;
 	node += sizeof (BTreeNodeHeader);
+	uint count = node_header.item_count;
+
+	if (node_header.item_count == 0) return keys;
+
+	if (node_header.level != 0) {
+		/* Si no es una hoja, lo primero que tengo es un BTreeChildData */
+		BTreeChildData *d = new BTreeChildData (node);
+		node += d->Size ();
+		keys.push_back (d);
+		count--;
+	}
 
-	for (uint i=0; i<node_header.item_count; i++) {
+	for (uint i=0; i<count; i++) {
 		/* TODO : El tipo de clave deberia ser usado 
 		 * dependiendo de algun dato en el header del
 		 * arbol
 		 */
 		/* TODO : Detectar si estoy en una hoja */
-		BTreeData data(node);
-		node += data.Size ();
+		BTreeData *data;
+		if (node_header.level == 0) {
+			data = new BTreeLeafData (node, key_type);
+		} else {
+			data = new BTreeData (node, key_type);
+		}
+		node += data->Size ();
 		keys.push_back (data);
 	}
 
 	return keys;
 }
 
-void BTree::WriteKeys (uchar *node, BTreeNodeHeader &node_header, std::list<BTreeData> &keys)
+void BTree::WriteKeys (uchar *node, BTreeNodeHeader &node_header, std::list<BTreeData *> &keys)
 {
-	std::list<BTreeData>::iterator it = keys.begin ();
+	std::list<BTreeData *>::iterator it = keys.begin ();
 
 	node += sizeof (BTreeNodeHeader);
 
+	node_header.item_count = 0;
+	node_header.free_space = header.block_size - sizeof (BTreeNodeHeader);
+
 	while (it != keys.end ()) {
-		BTreeData d = (*it);
-		memcpy (node, d.ToArray(), d.Size ());
-		node += d.Size ();
+		BTreeData *d = (*it);
+		uchar *n = d->ToArray ();
+		memcpy (node, n, d->Size ());
+		delete [] n;
+		node += d->Size ();
+		node_header.free_space -= d->Size ();
+		node_header.item_count++;
 		it++;
 	}
 
 	/* TODO : incrementar node_header.item_count aca o fuera de este metodo? */
 }
+		
+void BTree::PrintNode (uint num)
+{
+	uchar *node = ReadBlock (num);
+	BTreeNodeHeader node_header;
+	ReadNodoHeader (node, &node_header);
+		
+	std::list<BTreeData *> node_keys = ReadKeys (node, node_header);
+	std::list<BTreeData *>::iterator it = node_keys.begin ();
+
+	std::cout << "Nodo  : " << num << std::endl;
+	std::cout << "Level : " << node_header.level << std::endl;
+	std::cout << "Items : " << node_header.item_count << std::endl;
+	std::cout << "Free  : " << node_header.free_space << " (" << (header.block_size - sizeof (BTreeNodeHeader)) << ")" << std::endl;
+	while (it != node_keys.end ()) {
+		std::string s = *(*it);
+		std::cout << s << " "; 
+		it++;
+	}
+	std::cout << std::endl;
+
+	delete [] node;
+	DeleteKeys (node_keys);
+}
+
+uchar *BTree::NewBlock (uint &num)
+{
+	long filelen;
+	uchar *node;
+	BTreeNodeHeader nh;
+
+	fseek (fp, 0, SEEK_END);
+	filelen = ftell (fp);
+
+	num = filelen/header.block_size - 1;
+
+	node = new uchar[header.block_size];
+	ReadNodoHeader (node, &nh);
+	nh.level = 0;
+	nh.free_space = header.block_size - sizeof (BTreeNodeHeader);
+	nh.item_count = 0;
+	WriteNodoHeader (node, &nh);
+	WriteBlock (node, num);
+
+	return node;
+}
+
+bool BTree::FindKey (const Clave &k)
+{
+	return FindKeyR (&k, 0);
+}
+
+bool BTree::FindKeyR (const Clave *k, uint node_num)
+{
+	std::list<BTreeData *> node_keys;
+	BTreeNodeHeader node_header;
+
+	/* Leo el nodo raiz para empezar a agregar */
+	uchar *node = ReadBlock (node_num);
+	ReadNodoHeader (node, &node_header);
+	node_keys = ReadKeys (node, node_header);
+
+	std::list<BTreeData *>::iterator it = node_keys.begin ();
+	std::list<BTreeData *>::iterator posterior;
+	std::list<BTreeData *>::iterator ultima;
+	
+	/* Se supone que la primera es un hijo :) */
+	BTreeData *lchild;
+	if (node_header.level != 0) {
+		lchild = (*it++);
+	}
+	posterior = it;
+
+	BTreeData *data;
+	if (node_header.level == 0)
+		data = new BTreeLeafData ((Clave *)k);
+	else
+		data = new BTreeData ((Clave *)k, 0);
+
+	while (it != node_keys.end ()) {
+		if ((*data) == (*(*it))) {
+			/* La encontre!, retorno */
+			delete [] node;
+			DeleteKeys (node_keys);
+			return true;
+		}
+
+		if ((*data) < (*(*it)))
+			break;
+		ultima = it;
+		it++;
+	}
+
+	/* TODO: Aca faltaria liberar memoria */
+	if (it == posterior)
+		return FindKeyR (k, lchild->getChild ());
+		
+	return FindKeyR (k, (*ultima)->getChild ());
+}
+
+void BTree::DeleteKeys (std::list<BTreeData *> &keys)
+{
+	std::list<BTreeData *>::iterator it = keys.begin ();
+
+	while (it != keys.end ()) {
+		BTreeData *d = (*it);
+		delete d;
+		it++;
+	}
+}
+