This repository has been archived on 2023-10-09. You can view files and clone it, but cannot push or open issues or pull requests.
Files
blender-archive/source/blender/python/api2_2x/matrix.c
Stephen Swaney a4ca8267cf bugfix: [ #3009 ] possible memory leak in Mathutils
fixed by patch [ #3013 ] patch for memleak in vector
Submitted By: Ken Hughes (khughes)
2005-09-12 06:18:45 +00:00

842 lines
28 KiB
C

/*
* $Id$
*
* ***** BEGIN GPL/BL DUAL LICENSE BLOCK *****
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version. The Blender
* Foundation also sells licenses for use in proprietary software under
* the Blender License. See http://www.blender.org/BL/ for information
* about this.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*
* The Original Code is Copyright (C) 2001-2002 by NaN Holding BV.
* All rights reserved.
*
* Contributor(s): Michel Selten & Joseph Gilbert
*
* ***** END GPL/BL DUAL LICENSE BLOCK *****
*/
#include "Mathutils.h"
#include "BKE_utildefines.h"
#include "BLI_arithb.h"
#include "BLI_blenlib.h"
#include "gen_utils.h"
//-------------------------DOC STRINGS ---------------------------
char Matrix_Zero_doc[] = "() - set all values in the matrix to 0";
char Matrix_Identity_doc[] = "() - set the square matrix to it's identity matrix";
char Matrix_Transpose_doc[] = "() - set the matrix to it's transpose";
char Matrix_Determinant_doc[] = "() - return the determinant of the matrix";
char Matrix_Invert_doc[] = "() - set the matrix to it's inverse if an inverse is possible";
char Matrix_TranslationPart_doc[] = "() - return a vector encompassing the translation of the matrix";
char Matrix_RotationPart_doc[] = "() - return a vector encompassing the rotation of the matrix";
char Matrix_Resize4x4_doc[] = "() - resize the matrix to a 4x4 square matrix";
char Matrix_toEuler_doc[] = "() - convert matrix to a euler angle rotation";
char Matrix_toQuat_doc[] = "() - convert matrix to a quaternion rotation";
//-----------------------METHOD DEFINITIONS ----------------------
struct PyMethodDef Matrix_methods[] = {
{"zero", (PyCFunction) Matrix_Zero, METH_NOARGS, Matrix_Zero_doc},
{"identity", (PyCFunction) Matrix_Identity, METH_NOARGS, Matrix_Identity_doc},
{"transpose", (PyCFunction) Matrix_Transpose, METH_NOARGS, Matrix_Transpose_doc},
{"determinant", (PyCFunction) Matrix_Determinant, METH_NOARGS, Matrix_Determinant_doc},
{"invert", (PyCFunction) Matrix_Invert, METH_NOARGS, Matrix_Invert_doc},
{"translationPart", (PyCFunction) Matrix_TranslationPart, METH_NOARGS, Matrix_TranslationPart_doc},
{"rotationPart", (PyCFunction) Matrix_RotationPart, METH_NOARGS, Matrix_RotationPart_doc},
{"resize4x4", (PyCFunction) Matrix_Resize4x4, METH_NOARGS, Matrix_Resize4x4_doc},
{"toEuler", (PyCFunction) Matrix_toEuler, METH_NOARGS, Matrix_toEuler_doc},
{"toQuat", (PyCFunction) Matrix_toQuat, METH_NOARGS, Matrix_toQuat_doc},
{NULL, NULL, 0, NULL}
};
//-----------------------------METHODS----------------------------
//---------------------------Matrix.toQuat() ---------------------
PyObject *Matrix_toQuat(MatrixObject * self)
{
float quat[4];
//must be 3-4 cols, 3-4 rows, square matrix
if(self->colSize < 3 || self->rowSize < 3 || (self->colSize != self->rowSize)) {
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix.toQuat(): inappropriate matrix size - expects 3x3 or 4x4 matrix\n");
}
if(self->colSize == 3){
Mat3ToQuat((float (*)[3])*self->matrix, quat);
}else{
Mat4ToQuat((float (*)[4])*self->matrix, quat);
}
return (PyObject *) newQuaternionObject(quat, Py_NEW);
}
//---------------------------Matrix.toEuler() --------------------
PyObject *Matrix_toEuler(MatrixObject * self)
{
float eul[3];
int x;
//must be 3-4 cols, 3-4 rows, square matrix
if(self->colSize !=3 || self->rowSize != 3) {
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix.toQuat(): inappropriate matrix size - expects 3x3 matrix\n");
}
Mat3ToEul((float (*)[3])*self->matrix, eul);
//have to convert to degrees
for(x = 0; x < 3; x++) {
eul[x] *= (float) (180 / Py_PI);
}
return (PyObject *) newEulerObject(eul, Py_NEW);
}
//---------------------------Matrix.resize4x4() ------------------
PyObject *Matrix_Resize4x4(MatrixObject * self)
{
int x, first_row_elem, curr_pos, new_pos, blank_columns, blank_rows;
if(self->data.blend_data){
return EXPP_ReturnPyObjError(PyExc_TypeError,
"cannot resize wrapped data - only python matrices\n");
}
self->data.py_data = PyMem_Realloc(self->data.py_data, (sizeof(float) * 16));
if(self->data.py_data == NULL) {
return EXPP_ReturnPyObjError(PyExc_MemoryError,
"matrix.resize4x4(): problem allocating pointer space\n\n");
}
self->contigPtr = self->data.py_data; //force
self->matrix = PyMem_Realloc(self->matrix, (sizeof(float) * 4));
if(self->matrix == NULL) {
return EXPP_ReturnPyObjError(PyExc_MemoryError,
"matrix.resize4x4(): problem allocating pointer space\n\n");
}
//set row pointers
for(x = 0; x < 4; x++) {
self->matrix[x] = self->contigPtr + (x * 4);
}
//move data to new spot in array + clean
for(blank_rows = (4 - self->rowSize); blank_rows > 0; blank_rows--){
for(x = 0; x < 4; x++){
self->contigPtr[(4 * (self->rowSize + (blank_rows - 1))) + x] = 0.0f;
}
}
for(x = 1; x <= self->rowSize; x++){
first_row_elem = (self->colSize * (self->rowSize - x));
curr_pos = (first_row_elem + (self->colSize -1));
new_pos = (4 * (self->rowSize - x )) + (curr_pos - first_row_elem);
for(blank_columns = (4 - self->colSize); blank_columns > 0; blank_columns--){
self->contigPtr[new_pos + blank_columns] = 0.0f;
}
for(curr_pos = curr_pos; curr_pos >= first_row_elem; curr_pos--){
self->contigPtr[new_pos] = self->contigPtr[curr_pos];
new_pos--;
}
}
self->rowSize = 4;
self->colSize = 4;
return EXPP_incr_ret((PyObject*)self);
}
//---------------------------Matrix.translationPart() ------------
PyObject *Matrix_TranslationPart(MatrixObject * self)
{
float vec[4];
if(self->colSize < 3 && self->rowSize < 4){
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix.translationPart: inappropriate matrix size\n");
}
vec[0] = self->matrix[3][0];
vec[1] = self->matrix[3][1];
vec[2] = self->matrix[3][2];
return newVectorObject(vec, 3, Py_NEW);
}
//---------------------------Matrix.rotationPart() ---------------
PyObject *Matrix_RotationPart(MatrixObject * self)
{
float mat[16] = {0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f};
if(self->colSize < 3 && self->rowSize < 3){
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix.rotationPart: inappropriate matrix size\n");
}
mat[0] = self->matrix[0][0];
mat[1] = self->matrix[0][1];
mat[2] = self->matrix[0][2];
mat[3] = self->matrix[1][0];
mat[4] = self->matrix[1][1];
mat[5] = self->matrix[1][2];
mat[6] = self->matrix[2][0];
mat[7] = self->matrix[2][1];
mat[8] = self->matrix[2][2];
return newMatrixObject(mat, 3, 3, Py_NEW);
}
//---------------------------Matrix.invert() ---------------------
PyObject *Matrix_Invert(MatrixObject * self)
{
int x, y, z = 0;
float det = 0.0f;
PyObject *f = NULL;
float mat[16] = {0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f};
if(self->rowSize != self->colSize){
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix.invert: only square matrices are supported\n");
}
//calculate the determinant
f = Matrix_Determinant(self);
det = (float)PyFloat_AS_DOUBLE(f);
if(det != 0) {
//calculate the classical adjoint
if(self->rowSize == 2) {
mat[0] = self->matrix[1][1];
mat[1] = -self->matrix[1][0];
mat[2] = -self->matrix[0][1];
mat[3] = self->matrix[0][0];
} else if(self->rowSize == 3) {
Mat3Adj((float (*)[3]) mat,(float (*)[3]) *self->matrix);
} else if(self->rowSize == 4) {
Mat4Adj((float (*)[4]) mat, (float (*)[4]) *self->matrix);
}
//divide by determinate
for(x = 0; x < (self->rowSize * self->colSize); x++) {
mat[x] /= det;
}
//set values
for(x = 0; x < self->rowSize; x++) {
for(y = 0; y < self->colSize; y++) {
self->matrix[x][y] = mat[z];
z++;
}
}
//transpose
//Matrix_Transpose(self);
} else {
printf("Matrix.invert: matrix does not have an inverse\n");
}
return EXPP_incr_ret((PyObject*)self);
}
//---------------------------Matrix.determinant() ----------------
PyObject *Matrix_Determinant(MatrixObject * self)
{
float det = 0.0f;
if(self->rowSize != self->colSize){
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix.determinant: only square matrices are supported\n");
}
if(self->rowSize == 2) {
det = Det2x2(self->matrix[0][0], self->matrix[0][1],
self->matrix[1][0], self->matrix[1][1]);
} else if(self->rowSize == 3) {
det = Det3x3(self->matrix[0][0], self->matrix[0][1],
self->matrix[0][2], self->matrix[1][0],
self->matrix[1][1], self->matrix[1][2],
self->matrix[2][0], self->matrix[2][1],
self->matrix[2][2]);
} else {
det = Det4x4((float (*)[4]) *self->matrix);
}
return PyFloat_FromDouble( (double) det );
}
//---------------------------Matrix.transpose() ------------------
PyObject *Matrix_Transpose(MatrixObject * self)
{
float t = 0.0f;
if(self->rowSize != self->colSize){
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix.transpose: only square matrices are supported\n");
}
if(self->rowSize == 2) {
t = self->matrix[1][0];
self->matrix[1][0] = self->matrix[0][1];
self->matrix[0][1] = t;
} else if(self->rowSize == 3) {
Mat3Transp((float (*)[3])*self->matrix);
} else {
Mat4Transp((float (*)[4])*self->matrix);
}
return EXPP_incr_ret((PyObject*)self);
}
//---------------------------Matrix.zero() -----------------------
PyObject *Matrix_Zero(MatrixObject * self)
{
int row, col;
for(row = 0; row < self->rowSize; row++) {
for(col = 0; col < self->colSize; col++) {
self->matrix[row][col] = 0.0f;
}
}
return EXPP_incr_ret((PyObject*)self);
}
//---------------------------Matrix.identity(() ------------------
PyObject *Matrix_Identity(MatrixObject * self)
{
if(self->rowSize != self->colSize){
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix.identity: only square matrices are supported\n");
}
if(self->rowSize == 2) {
self->matrix[0][0] = 1.0f;
self->matrix[0][1] = 0.0f;
self->matrix[1][0] = 0.0f;
self->matrix[1][1] = 1.0f;
} else if(self->rowSize == 3) {
Mat3One((float (*)[3]) *self->matrix);
} else {
Mat4One((float (*)[4]) *self->matrix);
}
return EXPP_incr_ret((PyObject*)self);
}
//----------------------------dealloc()(internal) ----------------
//free the py_object
static void Matrix_dealloc(MatrixObject * self)
{
Py_XDECREF(self->coerced_object);
PyMem_Free(self->matrix);
//only free py_data
if(self->data.py_data){
PyMem_Free(self->data.py_data);
}
PyObject_DEL(self);
}
//----------------------------getattr()(internal) ----------------
//object.attribute access (get)
static PyObject *Matrix_getattr(MatrixObject * self, char *name)
{
if(STREQ(name, "rowSize")) {
return PyInt_FromLong((long) self->rowSize);
} else if(STREQ(name, "colSize")) {
return PyInt_FromLong((long) self->colSize);
}
if(STREQ(name, "wrapped")){
if(self->wrapped == Py_WRAP)
return EXPP_incr_ret((PyObject *)Py_True);
else
return EXPP_incr_ret((PyObject *)Py_False);
}
return Py_FindMethod(Matrix_methods, (PyObject *) self, name);
}
//----------------------------setattr()(internal) ----------------
//object.attribute access (set)
static int Matrix_setattr(MatrixObject * self, char *name, PyObject * v)
{
/* This is not supported. */
return (-1);
}
//----------------------------print object (internal)-------------
//print the object to screen
static PyObject *Matrix_repr(MatrixObject * self)
{
int x, y;
char buffer[48], str[1024];
BLI_strncpy(str,"",1024);
for(x = 0; x < self->rowSize; x++){
sprintf(buffer, "[");
strcat(str,buffer);
for(y = 0; y < (self->colSize - 1); y++) {
sprintf(buffer, "%.6f, ", self->matrix[x][y]);
strcat(str,buffer);
}
if(x < (self->rowSize-1)){
sprintf(buffer, "%.6f](matrix [row %d])\n", self->matrix[x][y], x);
strcat(str,buffer);
}else{
sprintf(buffer, "%.6f](matrix [row %d])", self->matrix[x][y], x);
strcat(str,buffer);
}
}
return EXPP_incr_ret(PyString_FromString(str));
}
//---------------------SEQUENCE PROTOCOLS------------------------
//----------------------------len(object)------------------------
//sequence length
static int Matrix_len(MatrixObject * self)
{
return (self->colSize * self->rowSize);
}
//----------------------------object[]---------------------------
//sequence accessor (get)
//the wrapped vector gives direct access to the matrix data
static PyObject *Matrix_item(MatrixObject * self, int i)
{
if(i < 0 || i >= self->rowSize)
return EXPP_ReturnPyObjError(PyExc_IndexError,
"matrix[attribute]: array index out of range\n");
return newVectorObject(self->matrix[i], self->colSize, Py_WRAP);
}
//----------------------------object[]-------------------------
//sequence accessor (set)
static int Matrix_ass_item(MatrixObject * self, int i, PyObject * ob)
{
int y, x, size = 0;
float vec[4];
if(i > self->rowSize || i < 0){
return EXPP_ReturnIntError(PyExc_TypeError,
"matrix[attribute] = x: bad row\n");
}
if(PySequence_Check(ob)){
size = PySequence_Length(ob);
if(size != self->colSize){
return EXPP_ReturnIntError(PyExc_TypeError,
"matrix[attribute] = x: bad sequence size\n");
}
for (x = 0; x < size; x++) {
PyObject *m, *f;
m = PySequence_GetItem(ob, x);
if (m == NULL) { // Failed to read sequence
return EXPP_ReturnIntError(PyExc_RuntimeError,
"matrix[attribute] = x: unable to read sequence\n");
}
f = PyNumber_Float(m);
if(f == NULL) { // parsed item not a number
Py_DECREF(m);
return EXPP_ReturnIntError(PyExc_TypeError,
"matrix[attribute] = x: sequence argument not a number\n");
}
vec[x] = (float)PyFloat_AS_DOUBLE(f);
EXPP_decr2(m, f);
}
//parsed well - now set in matrix
for(y = 0; y < size; y++){
self->matrix[i][y] = vec[y];
}
return 0;
}else{
return EXPP_ReturnIntError(PyExc_TypeError,
"matrix[attribute] = x: expects a sequence of column size\n");
}
}
//----------------------------object[z:y]------------------------
//sequence slice (get)
static PyObject *Matrix_slice(MatrixObject * self, int begin, int end)
{
PyObject *list = NULL;
int count;
CLAMP(begin, 0, self->rowSize);
CLAMP(end, 0, self->rowSize);
begin = MIN2(begin,end);
list = PyList_New(end - begin);
for(count = begin; count < end; count++) {
PyList_SetItem(list, count - begin,
newVectorObject(self->matrix[count], self->colSize, Py_WRAP));
}
return EXPP_incr_ret(list);
}
//----------------------------object[z:y]------------------------
//sequence slice (set)
static int Matrix_ass_slice(MatrixObject * self, int begin, int end,
PyObject * seq)
{
int i, x, y, size, sub_size = 0;
float mat[16];
CLAMP(begin, 0, self->rowSize);
CLAMP(end, 0, self->rowSize);
begin = MIN2(begin,end);
if(PySequence_Check(seq)){
size = PySequence_Length(seq);
if(size != (end - begin)){
return EXPP_ReturnIntError(PyExc_TypeError,
"matrix[begin:end] = []: size mismatch in slice assignment\n");
}
//parse sub items
for (i = 0; i < size; i++) {
//parse each sub sequence
PyObject *subseq;
subseq = PySequence_GetItem(seq, i);
if (subseq == NULL) { // Failed to read sequence
return EXPP_ReturnIntError(PyExc_RuntimeError,
"matrix[begin:end] = []: unable to read sequence\n");
}
if(PySequence_Check(subseq)){
//subsequence is also a sequence
sub_size = PySequence_Length(subseq);
if(sub_size != self->colSize){
return EXPP_ReturnIntError(PyExc_TypeError,
"matrix[begin:end] = []: size mismatch in slice assignment\n");
}
for (y = 0; y < sub_size; y++) {
PyObject *m, *f;
m = PySequence_GetItem(subseq, y);
if (m == NULL) { // Failed to read sequence
return EXPP_ReturnIntError(PyExc_RuntimeError,
"matrix[begin:end] = []: unable to read sequence\n");
}
f = PyNumber_Float(m);
if(f == NULL) { // parsed item not a number
Py_DECREF(m);
return EXPP_ReturnIntError(PyExc_TypeError,
"matrix[begin:end] = []: sequence argument not a number\n");
}
mat[(i * self->colSize) + y] = (float)PyFloat_AS_DOUBLE(f);
EXPP_decr2(f, m);
}
}else{
Py_DECREF(subseq);
return EXPP_ReturnIntError(PyExc_TypeError,
"matrix[begin:end] = []: illegal argument type for built-in operation\n");
}
}
//parsed well - now set in matrix
for(x = 0; x < (size * sub_size); x++){
self->matrix[begin + (int)floor(x / self->colSize)][x % self->colSize] = mat[x];
}
return 0;
}else{
return EXPP_ReturnIntError(PyExc_TypeError,
"matrix[begin:end] = []: illegal argument type for built-in operation\n");
}
}
//------------------------NUMERIC PROTOCOLS----------------------
//------------------------obj + obj------------------------------
static PyObject *Matrix_add(PyObject * m1, PyObject * m2)
{
int x, y;
float mat[16] = {0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f};
MatrixObject *mat1 = NULL, *mat2 = NULL;
EXPP_incr2(m1, m2);
mat1 = (MatrixObject*)m1;
mat2 = (MatrixObject*)m2;
if(mat1->coerced_object || mat2->coerced_object){
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix addition: arguments not valid for this operation....\n");
}
if(mat1->rowSize != mat2->rowSize || mat1->colSize != mat2->colSize){
EXPP_decr2((PyObject*)mat1, (PyObject*)mat2);
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix addition: matrices must have the same dimensions for this operation\n");
}
for(x = 0; x < mat1->rowSize; x++) {
for(y = 0; y < mat1->colSize; y++) {
mat[((x * mat1->colSize) + y)] = mat1->matrix[x][y] + mat2->matrix[x][y];
}
}
EXPP_decr2((PyObject*)mat1, (PyObject*)mat2);
return newMatrixObject(mat, mat1->rowSize, mat1->colSize, Py_NEW);
}
//------------------------obj - obj------------------------------
//subtraction
static PyObject *Matrix_sub(PyObject * m1, PyObject * m2)
{
int x, y;
float mat[16] = {0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f};
MatrixObject *mat1 = NULL, *mat2 = NULL;
EXPP_incr2(m1, m2);
mat1 = (MatrixObject*)m1;
mat2 = (MatrixObject*)m2;
if(mat1->coerced_object || mat2->coerced_object){
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix addition: arguments not valid for this operation....\n");
}
if(mat1->rowSize != mat2->rowSize || mat1->colSize != mat2->colSize){
EXPP_decr2((PyObject*)mat1, (PyObject*)mat2);
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix addition: matrices must have the same dimensions for this operation\n");
}
for(x = 0; x < mat1->rowSize; x++) {
for(y = 0; y < mat1->colSize; y++) {
mat[((x * mat1->colSize) + y)] = mat1->matrix[x][y] - mat2->matrix[x][y];
}
}
EXPP_decr2((PyObject*)mat1, (PyObject*)mat2);
return newMatrixObject(mat, mat1->rowSize, mat1->colSize, Py_NEW);
}
//------------------------obj * obj------------------------------
//mulplication
static PyObject *Matrix_mul(PyObject * m1, PyObject * m2)
{
int x, y, z;
float scalar;
float mat[16] = {0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f};
double dot = 0.0f;
MatrixObject *mat1 = NULL, *mat2 = NULL;
PyObject *f = NULL, *retObj = NULL;
VectorObject *vec = NULL;
PointObject *pt = NULL;
EXPP_incr2(m1, m2);
mat1 = (MatrixObject*)m1;
mat2 = (MatrixObject*)m2;
if(mat1->coerced_object){
if (PyFloat_Check(mat1->coerced_object) ||
PyInt_Check(mat1->coerced_object)){ // FLOAT/INT * MATRIX
f = PyNumber_Float(mat1->coerced_object);
if(f == NULL) { // parsed item not a number
EXPP_decr2((PyObject*)mat1, (PyObject*)mat2);
return EXPP_ReturnPyObjError(PyExc_TypeError,
"Matrix multiplication: arguments not acceptable for this operation\n");
}
scalar = (float)PyFloat_AS_DOUBLE(f);
for(x = 0; x < mat2->rowSize; x++) {
for(y = 0; y < mat2->colSize; y++) {
mat[((x * mat2->colSize) + y)] = scalar * mat2->matrix[x][y];
}
}
EXPP_decr2((PyObject*)mat1, (PyObject*)mat2);
return newMatrixObject(mat, mat2->rowSize, mat2->colSize, Py_NEW);
}
}else{
if(mat2->coerced_object){
if(VectorObject_Check(mat2->coerced_object)){ //MATRIX * VECTOR
vec = (VectorObject*)EXPP_incr_ret(mat2->coerced_object);
retObj = column_vector_multiplication(mat1, vec);
EXPP_decr3((PyObject*)mat1, (PyObject*)mat2, (PyObject*)vec);
return retObj;
}else if(PointObject_Check(mat2->coerced_object)){ //MATRIX * POINT
pt = (PointObject*)EXPP_incr_ret(mat2->coerced_object);
retObj = column_point_multiplication(mat1, pt);
EXPP_decr3((PyObject*)mat1, (PyObject*)mat2, (PyObject*)pt);
return retObj;
}else if (PyFloat_Check(mat2->coerced_object) ||
PyInt_Check(mat2->coerced_object)){ // MATRIX * FLOAT/INT
f = PyNumber_Float(mat2->coerced_object);
if(f == NULL) { // parsed item not a number
EXPP_decr2((PyObject*)mat1, (PyObject*)mat2);
return EXPP_ReturnPyObjError(PyExc_TypeError,
"Matrix multiplication: arguments not acceptable for this operation\n");
}
scalar = (float)PyFloat_AS_DOUBLE(f);
for(x = 0; x < mat1->rowSize; x++) {
for(y = 0; y < mat1->colSize; y++) {
mat[((x * mat1->colSize) + y)] = scalar * mat1->matrix[x][y];
}
}
EXPP_decr2((PyObject*)mat1, (PyObject*)mat2);
return newMatrixObject(mat, mat1->rowSize, mat1->colSize, Py_NEW);
}
}else{ //MATRIX * MATRIX
if(mat1->colSize != mat2->rowSize){
EXPP_decr2((PyObject*)mat1, (PyObject*)mat2);
return EXPP_ReturnPyObjError(PyExc_AttributeError,
"Matrix multiplication: matrix A rowsize must equal matrix B colsize\n");
}
for(x = 0; x < mat1->rowSize; x++) {
for(y = 0; y < mat2->colSize; y++) {
for(z = 0; z < mat1->colSize; z++) {
dot += (mat1->matrix[x][z] * mat2->matrix[z][y]);
}
mat[((x * mat1->rowSize) + y)] = (float)dot;
dot = 0.0f;
}
}
return newMatrixObject(mat, mat1->rowSize, mat2->colSize, Py_NEW);
}
}
EXPP_decr2((PyObject*)mat1, (PyObject*)mat2);
return EXPP_ReturnPyObjError(PyExc_TypeError,
"Matrix multiplication: arguments not acceptable for this operation\n");
}
PyObject* Matrix_inv(MatrixObject *self)
{
return Matrix_Invert(self);
}
//------------------------coerce(obj, obj)-----------------------
//coercion of unknown types to type MatrixObject for numeric protocols
/*Coercion() is called whenever a math operation has 2 operands that
it doesn't understand how to evaluate. 2+Matrix for example. We want to
evaluate some of these operations like: (vector * 2), however, for math
to proceed, the unknown operand must be cast to a type that python math will
understand. (e.g. in the case above case, 2 must be cast to a vector and
then call vector.multiply(vector, scalar_cast_as_vector)*/
static int Matrix_coerce(PyObject ** m1, PyObject ** m2)
{
PyObject *coerced = NULL;
if(!MatrixObject_Check(*m2)) {
if(VectorObject_Check(*m2) || PyFloat_Check(*m2) || PyInt_Check(*m2) ||
PointObject_Check(*m2)) {
coerced = EXPP_incr_ret(*m2);
*m2 = newMatrixObject(NULL,3,3,Py_NEW);
((MatrixObject*)*m2)->coerced_object = coerced;
}else{
return EXPP_ReturnIntError(PyExc_TypeError,
"matrix.coerce(): unknown operand - can't coerce for numeric protocols\n");
}
}
Py_INCREF(*m2);
Py_INCREF(*m1);
return 0;
}
//-----------------PROTCOL DECLARATIONS--------------------------
static PySequenceMethods Matrix_SeqMethods = {
(inquiry) Matrix_len, /* sq_length */
(binaryfunc) 0, /* sq_concat */
(intargfunc) 0, /* sq_repeat */
(intargfunc) Matrix_item, /* sq_item */
(intintargfunc) Matrix_slice, /* sq_slice */
(intobjargproc) Matrix_ass_item, /* sq_ass_item */
(intintobjargproc) Matrix_ass_slice, /* sq_ass_slice */
};
static PyNumberMethods Matrix_NumMethods = {
(binaryfunc) Matrix_add, /* __add__ */
(binaryfunc) Matrix_sub, /* __sub__ */
(binaryfunc) Matrix_mul, /* __mul__ */
(binaryfunc) 0, /* __div__ */
(binaryfunc) 0, /* __mod__ */
(binaryfunc) 0, /* __divmod__ */
(ternaryfunc) 0, /* __pow__ */
(unaryfunc) 0, /* __neg__ */
(unaryfunc) 0, /* __pos__ */
(unaryfunc) 0, /* __abs__ */
(inquiry) 0, /* __nonzero__ */
(unaryfunc) Matrix_inv, /* __invert__ */
(binaryfunc) 0, /* __lshift__ */
(binaryfunc) 0, /* __rshift__ */
(binaryfunc) 0, /* __and__ */
(binaryfunc) 0, /* __xor__ */
(binaryfunc) 0, /* __or__ */
(coercion) Matrix_coerce, /* __coerce__ */
(unaryfunc) 0, /* __int__ */
(unaryfunc) 0, /* __long__ */
(unaryfunc) 0, /* __float__ */
(unaryfunc) 0, /* __oct__ */
(unaryfunc) 0, /* __hex__ */
};
//------------------PY_OBECT DEFINITION--------------------------
PyTypeObject matrix_Type = {
PyObject_HEAD_INIT(NULL) /* required python macro */
0, /*ob_size */
"Matrix", /*tp_name */
sizeof(MatrixObject), /*tp_basicsize */
0, /*tp_itemsize */
(destructor) Matrix_dealloc, /*tp_dealloc */
(printfunc) 0, /*tp_print */
(getattrfunc) Matrix_getattr, /*tp_getattr */
(setattrfunc) Matrix_setattr, /*tp_setattr */
0, /*tp_compare */
(reprfunc) Matrix_repr, /*tp_repr */
&Matrix_NumMethods, /*tp_as_number */
&Matrix_SeqMethods, /*tp_as_sequence */
};
//------------------------newMatrixObject (internal)-------------
//creates a new matrix object
//self->matrix self->contiguous_ptr (reference to data.xxx)
// [0]------------->[0]
// [1]
// [2]
// [1]------------->[3]
// [4]
// [5]
// ....
//self->matrix[1][1] = self->contiguous_ptr[4] = self->data.xxx_data[4]
/*pass Py_WRAP - if vector is a WRAPPER for data allocated by BLENDER
(i.e. it was allocated elsewhere by MEM_mallocN())
pass Py_NEW - if vector is not a WRAPPER and managed by PYTHON
(i.e. it must be created here with PyMEM_malloc())*/
PyObject *newMatrixObject(float *mat, int rowSize, int colSize, int type)
{
MatrixObject *self;
int x, row, col;
//matrix objects can be any 2-4row x 2-4col matrix
if(rowSize < 2 || rowSize > 4 || colSize < 2 || colSize > 4){
return EXPP_ReturnPyObjError(PyExc_RuntimeError,
"matrix(): row and column sizes must be between 2 and 4\n");
}
matrix_Type.ob_type = &PyType_Type;
self = PyObject_NEW(MatrixObject, &matrix_Type);
self->data.blend_data = NULL;
self->data.py_data = NULL;
self->rowSize = rowSize;
self->colSize = colSize;
self->coerced_object = NULL;
if(type == Py_WRAP){
self->data.blend_data = mat;
self->contigPtr = self->data.blend_data;
//create pointer array
self->matrix = PyMem_Malloc(rowSize * sizeof(float *));
if(self->matrix == NULL) { //allocation failure
return EXPP_ReturnPyObjError( PyExc_MemoryError,
"matrix(): problem allocating pointer space\n");
}
//pointer array points to contigous memory
for(x = 0; x < rowSize; x++) {
self->matrix[x] = self->contigPtr + (x * colSize);
}
self->wrapped = Py_WRAP;
}else if (type == Py_NEW){
self->data.py_data = PyMem_Malloc(rowSize * colSize * sizeof(float));
if(self->data.py_data == NULL) { //allocation failure
return EXPP_ReturnPyObjError( PyExc_MemoryError,
"matrix(): problem allocating pointer space\n");
}
self->contigPtr = self->data.py_data;
//create pointer array
self->matrix = PyMem_Malloc(rowSize * sizeof(float *));
if(self->matrix == NULL) { //allocation failure
PyMem_Free(self->data.py_data);
return EXPP_ReturnPyObjError( PyExc_MemoryError,
"matrix(): problem allocating pointer space\n");
}
//pointer array points to contigous memory
for(x = 0; x < rowSize; x++) {
self->matrix[x] = self->contigPtr + (x * colSize);
}
//parse
if(mat) { //if a float array passed
for(row = 0; row < rowSize; row++) {
for(col = 0; col < colSize; col++) {
self->matrix[row][col] = mat[(row * colSize) + col];
}
}
} else { //or if no arguments are passed return identity matrix
Matrix_Identity(self);
}
self->wrapped = Py_NEW;
}else{ //bad type
return NULL;
}
return (PyObject *) self;
}