godot/core/dvector.h
Juan Linietsky c5f509f238 New Navigation & Pathfinding support for 2D
-Added Navigation & NavigationPolygon nodes
-Added corresponding visual editor
-New pathfinding algorithm is modern and fast!
-Similar API to 3D Pathfinding (more coherent)
2015-02-14 12:10:15 -03:00

430 lines
8.4 KiB
C++

/*************************************************************************/
/* dvector.h */
/*************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* http://www.godotengine.org */
/*************************************************************************/
/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */
/* */
/* Permission is hereby granted, free of charge, to any person obtaining */
/* a copy of this software and associated documentation files (the */
/* "Software"), to deal in the Software without restriction, including */
/* without limitation the rights to use, copy, modify, merge, publish, */
/* distribute, sublicense, and/or sell copies of the Software, and to */
/* permit persons to whom the Software is furnished to do so, subject to */
/* the following conditions: */
/* */
/* The above copyright notice and this permission notice shall be */
/* included in all copies or substantial portions of the Software. */
/* */
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
/*************************************************************************/
#ifndef DVECTOR_H
#define DVECTOR_H
#include "os/memory.h"
/**
@author Juan Linietsky <reduzio@gmail.com>
*/
extern Mutex* dvector_lock;
template<class T>
class DVector {
mutable MID mem;
void copy_on_write() {
if (!mem.is_valid())
return;
if (dvector_lock)
dvector_lock->lock();
MID_Lock lock( mem );
if ( *(int*)lock.data() == 1 ) {
// one reference, means no refcount changes
if (dvector_lock)
dvector_lock->unlock();
return;
}
MID new_mem= dynalloc( mem.get_size() );
if (!new_mem.is_valid()) {
if (dvector_lock)
dvector_lock->unlock();
ERR_FAIL_COND( new_mem.is_valid() ); // out of memory
}
MID_Lock dst_lock( new_mem );
int *rc = (int*)dst_lock.data();
*rc=1;
T * dst = (T*)(rc + 1 );
T * src =(T*) ((int*)lock.data() + 1 );
int count = (mem.get_size() - sizeof(int)) / sizeof(T);
for (int i=0;i<count;i++) {
memnew_placement( &dst[i], T(src[i]) );
}
(*(int*)lock.data())--;
// unlock all
dst_lock=MID_Lock();
lock=MID_Lock();
mem=new_mem;
if (dvector_lock)
dvector_lock->unlock();
}
void reference( const DVector& p_dvector ) {
unreference();
if (dvector_lock)
dvector_lock->lock();
if (!p_dvector.mem.is_valid()) {
if (dvector_lock)
dvector_lock->unlock();
return;
}
MID_Lock lock(p_dvector.mem);
int * rc = (int*)lock.data();
(*rc)++;
lock = MID_Lock();
mem=p_dvector.mem;
if (dvector_lock)
dvector_lock->unlock();
}
void unreference() {
if (dvector_lock)
dvector_lock->lock();
if (!mem.is_valid()) {
if (dvector_lock)
dvector_lock->unlock();
return;
}
MID_Lock lock(mem);
int * rc = (int*)lock.data();
(*rc)--;
if (*rc==0) {
// no one else using it, destruct
T * t= (T*)(rc+1);
int count = (mem.get_size() - sizeof(int)) / sizeof(T);
for (int i=0;i<count;i++) {
t[i].~T();
}
}
lock = MID_Lock();
mem = MID ();
if (dvector_lock)
dvector_lock->unlock();
}
public:
class Read {
friend class DVector;
MID_Lock lock;
const T * mem;
public:
_FORCE_INLINE_ const T& operator[](int p_index) const { return mem[p_index]; }
_FORCE_INLINE_ const T *ptr() const { return mem; }
Read() { mem=NULL; }
};
class Write {
friend class DVector;
MID_Lock lock;
T * mem;
public:
_FORCE_INLINE_ T& operator[](int p_index) { return mem[p_index]; }
_FORCE_INLINE_ T *ptr() { return mem; }
Write() { mem=NULL; }
};
Read read() const {
Read r;
if (mem.is_valid()) {
r.lock = MID_Lock( mem );
r.mem = (const T*)((int*)r.lock.data()+1);
}
return r;
}
Write write() {
Write w;
if (mem.is_valid()) {
copy_on_write();
w.lock = MID_Lock( mem );
w.mem = (T*)((int*)w.lock.data()+1);
}
return w;
}
template<class MC>
void fill_with(const MC& p_mc) {
int c=p_mc.size();
resize(c);
Write w=write();
int idx=0;
for(const typename MC::Element *E=p_mc.front();E;E=E->next()) {
w[idx++]=E->get();
}
}
void remove(int p_index) {
int s = size();
ERR_FAIL_INDEX(p_index, s);
Write w = write();
for (int i=p_index; i<s-1; i++) {
w[i]=w[i+1];
};
w = Write();
resize(s-1);
}
inline int size() const;
T get(int p_index) const;
void set(int p_index, const T& p_val);
void push_back(const T& p_val);
void append(const T& p_val) { push_back(p_val); }
void append_array(const DVector<T>& p_arr) {
int ds = p_arr.size();
if (ds==0)
return;
int bs = size();
resize( bs + ds);
Write w = write();
Read r = p_arr.read();
for(int i=0;i<ds;i++)
w[bs+i]=r[i];
}
Error insert(int p_pos,const T& p_val) {
int s=size();
ERR_FAIL_INDEX_V(p_pos,s+1,ERR_INVALID_PARAMETER);
resize(s+1);
{
Write w = write();
for (int i=s;i>p_pos;i--)
w[i]=w[i-1];
w[p_pos]=p_val;
}
return OK;
}
bool is_locked() const { return mem.is_locked(); }
inline const T operator[](int p_index) const;
Error resize(int p_size);
void operator=(const DVector& p_dvector) { reference(p_dvector); }
DVector() {}
DVector(const DVector& p_dvector) { reference(p_dvector); }
~DVector() { unreference(); }
};
template<class T>
int DVector<T>::size() const {
return mem.is_valid() ? ((mem.get_size() - sizeof(int)) / sizeof(T) ) : 0;
}
template<class T>
T DVector<T>::get(int p_index) const {
return operator[](p_index);
}
template<class T>
void DVector<T>::set(int p_index, const T& p_val) {
if (p_index<0 || p_index>=size()) {
ERR_FAIL_COND(p_index<0 || p_index>=size());
}
Write w = write();
w[p_index]=p_val;
}
template<class T>
void DVector<T>::push_back(const T& p_val) {
resize( size() + 1 );
set( size() -1, p_val );
}
template<class T>
const T DVector<T>::operator[](int p_index) const {
if (p_index<0 || p_index>=size()) {
T& aux=*((T*)0); //nullreturn
ERR_FAIL_COND_V(p_index<0 || p_index>=size(),aux);
}
Read r = read();
return r[p_index];
}
template<class T>
Error DVector<T>::resize(int p_size) {
if (dvector_lock)
dvector_lock->lock();
bool same = p_size==size();
if (dvector_lock)
dvector_lock->unlock();
// no further locking is necesary because we are supposed to own the only copy of this (using copy on write)
if (same)
return OK;
if (p_size == 0 ) {
unreference();
return OK;
}
copy_on_write(); // make it unique
ERR_FAIL_COND_V( mem.is_locked(), ERR_LOCKED ); // if after copy on write, memory is locked, fail.
if (p_size > size() ) {
int oldsize=size();
MID_Lock lock;
if (oldsize==0) {
mem = dynalloc( p_size * sizeof(T) + sizeof(int) );
lock=MID_Lock(mem);
int *rc = ((int*)lock.data());
*rc=1;
} else {
if (dynrealloc( mem, p_size * sizeof(T) + sizeof(int) )!=OK ) {
ERR_FAIL_V(ERR_OUT_OF_MEMORY); // out of memory
}
lock=MID_Lock(mem);
}
T *t = (T*)((int*)lock.data() + 1);
for (int i=oldsize;i<p_size;i++) {
memnew_placement(&t[i], T );
}
lock = MID_Lock(); // clear
} else {
int oldsize=size();
MID_Lock lock(mem);
T *t = (T*)((int*)lock.data() + 1);
for (int i=p_size;i<oldsize;i++) {
t[i].~T();
}
lock = MID_Lock(); // clear
if (dynrealloc( mem, p_size * sizeof(T) + sizeof(int) )!=OK ) {
ERR_FAIL_V(ERR_OUT_OF_MEMORY); // wtf error
}
}
return OK;
}
#endif