412 lines
11 KiB
C
412 lines
11 KiB
C
/*
|
|
* 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.
|
|
*
|
|
* 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
|
|
*/
|
|
|
|
/** \file
|
|
* \ingroup bli
|
|
* \brief Generic array manipulation API.
|
|
*
|
|
* \warning Some array operations here are inherently inefficient,
|
|
* and only included for the cases where the performance is acceptable.
|
|
* Use with care.
|
|
*/
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
|
|
#include "MEM_guardedalloc.h"
|
|
|
|
#include "BLI_alloca.h"
|
|
#include "BLI_math_base.h"
|
|
#include "BLI_strict_flags.h"
|
|
#include "BLI_sys_types.h"
|
|
#include "BLI_utildefines.h"
|
|
|
|
#include "BLI_array_utils.h"
|
|
|
|
/**
|
|
*In-place array reverse.
|
|
*
|
|
* Access via #BLI_array_reverse
|
|
*/
|
|
void _bli_array_reverse(void *arr_v, unsigned int arr_len, size_t arr_stride)
|
|
{
|
|
const unsigned int arr_stride_uint = (unsigned int)arr_stride;
|
|
const unsigned int arr_half_stride = (arr_len / 2) * arr_stride_uint;
|
|
unsigned int i, i_end;
|
|
char *arr = arr_v;
|
|
char *buf = BLI_array_alloca(buf, arr_stride);
|
|
|
|
for (i = 0, i_end = (arr_len - 1) * arr_stride_uint; i < arr_half_stride;
|
|
i += arr_stride_uint, i_end -= arr_stride_uint) {
|
|
memcpy(buf, &arr[i], arr_stride);
|
|
memcpy(&arr[i], &arr[i_end], arr_stride);
|
|
memcpy(&arr[i_end], buf, arr_stride);
|
|
}
|
|
}
|
|
|
|
/**
|
|
* In-place array wrap.
|
|
* (rotate the array one step forward or backwards).
|
|
*
|
|
* Access via #BLI_array_wrap
|
|
*/
|
|
void _bli_array_wrap(void *arr_v, unsigned int arr_len, size_t arr_stride, int dir)
|
|
{
|
|
char *arr = arr_v;
|
|
char *buf = BLI_array_alloca(buf, arr_stride);
|
|
|
|
if (dir == -1) {
|
|
memcpy(buf, arr, arr_stride);
|
|
memmove(arr, arr + arr_stride, arr_stride * (arr_len - 1));
|
|
memcpy(arr + (arr_stride * (arr_len - 1)), buf, arr_stride);
|
|
}
|
|
else if (dir == 1) {
|
|
memcpy(buf, arr + (arr_stride * (arr_len - 1)), arr_stride);
|
|
memmove(arr + arr_stride, arr, arr_stride * (arr_len - 1));
|
|
memcpy(arr, buf, arr_stride);
|
|
}
|
|
else {
|
|
BLI_assert(0);
|
|
}
|
|
}
|
|
|
|
/**
|
|
*In-place array permute.
|
|
* (re-arrange elements based on an array of indices).
|
|
*
|
|
* Access via #BLI_array_wrap
|
|
*/
|
|
void _bli_array_permute(void *arr,
|
|
const unsigned int arr_len,
|
|
const size_t arr_stride,
|
|
const unsigned int *order,
|
|
void *arr_temp)
|
|
{
|
|
const size_t len = arr_len * arr_stride;
|
|
const unsigned int arr_stride_uint = (unsigned int)arr_stride;
|
|
void *arr_orig;
|
|
unsigned int i;
|
|
|
|
if (arr_temp == NULL) {
|
|
arr_orig = MEM_mallocN(len, __func__);
|
|
}
|
|
else {
|
|
arr_orig = arr_temp;
|
|
}
|
|
|
|
memcpy(arr_orig, arr, len);
|
|
|
|
for (i = 0; i < arr_len; i++) {
|
|
BLI_assert(order[i] < arr_len);
|
|
memcpy(POINTER_OFFSET(arr, arr_stride_uint * i),
|
|
POINTER_OFFSET(arr_orig, arr_stride_uint * order[i]),
|
|
arr_stride);
|
|
}
|
|
|
|
if (arr_temp == NULL) {
|
|
MEM_freeN(arr_orig);
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Find the first index of an item in an array.
|
|
*
|
|
* Access via #BLI_array_findindex
|
|
*
|
|
* \note Not efficient, use for error checks/asserts.
|
|
*/
|
|
int _bli_array_findindex(const void *arr, unsigned int arr_len, size_t arr_stride, const void *p)
|
|
{
|
|
const char *arr_step = (const char *)arr;
|
|
for (unsigned int i = 0; i < arr_len; i++, arr_step += arr_stride) {
|
|
if (memcmp(arr_step, p, arr_stride) == 0) {
|
|
return (int)i;
|
|
}
|
|
}
|
|
return -1;
|
|
}
|
|
|
|
/**
|
|
* A version of #BLI_array_findindex that searches from the end of the list.
|
|
*/
|
|
int _bli_array_rfindindex(const void *arr, unsigned int arr_len, size_t arr_stride, const void *p)
|
|
{
|
|
const char *arr_step = (const char *)arr + (arr_stride * arr_len);
|
|
for (unsigned int i = arr_len; i-- != 0;) {
|
|
arr_step -= arr_stride;
|
|
if (memcmp(arr_step, p, arr_stride) == 0) {
|
|
return (int)i;
|
|
}
|
|
}
|
|
return -1;
|
|
}
|
|
|
|
void _bli_array_binary_and(
|
|
void *arr, const void *arr_a, const void *arr_b, unsigned int arr_len, size_t arr_stride)
|
|
{
|
|
char *dst = arr;
|
|
const char *src_a = arr_a;
|
|
const char *src_b = arr_b;
|
|
|
|
size_t i = arr_stride * arr_len;
|
|
while (i--) {
|
|
*(dst++) = *(src_a++) & *(src_b++);
|
|
}
|
|
}
|
|
|
|
void _bli_array_binary_or(
|
|
void *arr, const void *arr_a, const void *arr_b, unsigned int arr_len, size_t arr_stride)
|
|
{
|
|
char *dst = arr;
|
|
const char *src_a = arr_a;
|
|
const char *src_b = arr_b;
|
|
|
|
size_t i = arr_stride * arr_len;
|
|
while (i--) {
|
|
*(dst++) = *(src_a++) | *(src_b++);
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Utility function to iterate over contiguous items in an array.
|
|
*
|
|
* \param use_wrap: Detect contiguous ranges across the first/last points.
|
|
* In this case the second index of \a span_step may be lower than the first,
|
|
* which indicates the values are wrapped.
|
|
* \param use_delimit_bounds: When false,
|
|
* ranges that defined by the start/end indices are excluded.
|
|
* This option has no effect when \a use_wrap is enabled.
|
|
* \param test_fn: Function to test if the item should be included in the range.
|
|
* \param user_data: User data for \a test_fn.
|
|
* \param span_step: Indices to iterate over,
|
|
* initialize both values to the array length to initialize iteration.
|
|
* \param r_span_len: The length of the span, useful when \a use_wrap is enabled,
|
|
* where calculating the length isn't a simple subtraction.
|
|
*/
|
|
bool _bli_array_iter_span(const void *arr,
|
|
unsigned int arr_len,
|
|
size_t arr_stride,
|
|
bool use_wrap,
|
|
bool use_delimit_bounds,
|
|
bool (*test_fn)(const void *arr_item, void *user_data),
|
|
void *user_data,
|
|
unsigned int span_step[2],
|
|
unsigned int *r_span_len)
|
|
{
|
|
if (arr_len == 0) {
|
|
return false;
|
|
}
|
|
if (use_wrap && (span_step[0] != arr_len) && (span_step[0] > span_step[1])) {
|
|
return false;
|
|
}
|
|
|
|
const unsigned int arr_stride_uint = (unsigned int)arr_stride;
|
|
const void *item_prev;
|
|
bool test_prev;
|
|
|
|
unsigned int i_curr;
|
|
|
|
if ((span_step[0] == arr_len) && (span_step[1] == arr_len)) {
|
|
if (use_wrap) {
|
|
item_prev = POINTER_OFFSET(arr, (arr_len - 1) * arr_stride_uint);
|
|
i_curr = 0;
|
|
test_prev = test_fn(item_prev, user_data);
|
|
}
|
|
else if (use_delimit_bounds == false) {
|
|
item_prev = arr;
|
|
i_curr = 1;
|
|
test_prev = test_fn(item_prev, user_data);
|
|
}
|
|
else {
|
|
item_prev = NULL;
|
|
i_curr = 0;
|
|
test_prev = false;
|
|
}
|
|
}
|
|
else if ((i_curr = span_step[1] + 2) < arr_len) {
|
|
item_prev = POINTER_OFFSET(arr, (span_step[1] + 1) * arr_stride_uint);
|
|
test_prev = test_fn(item_prev, user_data);
|
|
}
|
|
else {
|
|
return false;
|
|
}
|
|
BLI_assert(i_curr < arr_len);
|
|
|
|
const void *item_curr = POINTER_OFFSET(arr, i_curr * arr_stride_uint);
|
|
|
|
while (i_curr < arr_len) {
|
|
bool test_curr = test_fn(item_curr, user_data);
|
|
if ((test_prev == false) && (test_curr == true)) {
|
|
unsigned int span_len;
|
|
unsigned int i_step_prev = i_curr;
|
|
|
|
if (use_wrap) {
|
|
unsigned int i_step = i_curr + 1;
|
|
if (UNLIKELY(i_step == arr_len)) {
|
|
i_step = 0;
|
|
}
|
|
while (test_fn(POINTER_OFFSET(arr, i_step * arr_stride_uint), user_data)) {
|
|
i_step_prev = i_step;
|
|
i_step++;
|
|
if (UNLIKELY(i_step == arr_len)) {
|
|
i_step = 0;
|
|
}
|
|
}
|
|
|
|
if (i_step_prev < i_curr) {
|
|
span_len = (i_step_prev + (arr_len - i_curr)) + 1;
|
|
}
|
|
else {
|
|
span_len = (i_step_prev - i_curr) + 1;
|
|
}
|
|
}
|
|
else {
|
|
unsigned int i_step = i_curr + 1;
|
|
while ((i_step != arr_len) &&
|
|
test_fn(POINTER_OFFSET(arr, i_step * arr_stride_uint), user_data)) {
|
|
i_step_prev = i_step;
|
|
i_step++;
|
|
}
|
|
|
|
span_len = (i_step_prev - i_curr) + 1;
|
|
|
|
if ((use_delimit_bounds == false) && (i_step_prev == arr_len - 1)) {
|
|
return false;
|
|
}
|
|
}
|
|
|
|
span_step[0] = i_curr;
|
|
span_step[1] = i_step_prev;
|
|
*r_span_len = span_len;
|
|
|
|
return true;
|
|
}
|
|
|
|
test_prev = test_curr;
|
|
|
|
item_prev = item_curr;
|
|
item_curr = POINTER_OFFSET(item_curr, arr_stride_uint);
|
|
i_curr++;
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
/**
|
|
* Simple utility to check memory is zeroed.
|
|
*/
|
|
bool _bli_array_is_zeroed(const void *arr_v, unsigned int arr_len, size_t arr_stride)
|
|
{
|
|
const char *arr_step = (const char *)arr_v;
|
|
size_t i = arr_stride * arr_len;
|
|
while (i--) {
|
|
if (*(arr_step++)) {
|
|
return false;
|
|
}
|
|
}
|
|
return true;
|
|
}
|
|
|
|
/**
|
|
* Smart function to sample a rect spiraling outside.
|
|
* Nice for selection ID.
|
|
*
|
|
* \param arr_shape: dimensions [w, h].
|
|
* \param center: coordinates [x, y] indicating where to start traversing.
|
|
*/
|
|
bool _bli_array_iter_spiral_square(const void *arr_v,
|
|
const int arr_shape[2],
|
|
size_t elem_size,
|
|
const int center[2],
|
|
bool (*test_fn)(const void *arr_item, void *user_data),
|
|
void *user_data)
|
|
{
|
|
BLI_assert(center[0] >= 0 && center[1] >= 0 && center[0] < arr_shape[0] &&
|
|
center[1] < arr_shape[1]);
|
|
|
|
const char *arr = arr_v;
|
|
const int stride[2] = {arr_shape[0] * (int)elem_size, (int)elem_size};
|
|
|
|
/* Test center first. */
|
|
int ofs[2] = {center[0] * stride[1], center[1] * stride[0]};
|
|
if (test_fn(arr + ofs[0] + ofs[1], user_data)) {
|
|
return true;
|
|
}
|
|
|
|
/* #steps_in and #steps_out are the "diameters" of the inscribed and circumscribed squares in the
|
|
* rectangle. Each step smaller than #steps_in does not need to check bounds. */
|
|
int steps_in, steps_out;
|
|
{
|
|
int x_minus = center[0];
|
|
int x_plus = arr_shape[0] - center[0] - 1;
|
|
int y_minus = center[1];
|
|
int y_plus = arr_shape[1] - center[1] - 1;
|
|
|
|
steps_in = 2 * min_iiii(x_minus, x_plus, y_minus, y_plus);
|
|
steps_out = 2 * max_iiii(x_minus, x_plus, y_minus, y_plus);
|
|
}
|
|
|
|
/* For check_bounds. */
|
|
int limits[2] = {(arr_shape[0] - 1) * stride[0], stride[0] - stride[1]};
|
|
|
|
int steps = 0;
|
|
while (steps < steps_out) {
|
|
steps += 2;
|
|
|
|
/* Move one step to the diagonal of the negative quadrant. */
|
|
ofs[0] -= stride[0];
|
|
ofs[1] -= stride[1];
|
|
|
|
bool check_bounds = steps > steps_in;
|
|
|
|
/* sign: 0 neg; 1 pos; */
|
|
for (int sign = 2; sign--;) {
|
|
/* axis: 0 x; 1 y; */
|
|
for (int axis = 2; axis--;) {
|
|
int ofs_step = stride[axis];
|
|
if (!sign) {
|
|
ofs_step *= -1;
|
|
}
|
|
|
|
int ofs_iter = ofs[axis] + ofs_step;
|
|
int ofs_dest = ofs[axis] + steps * ofs_step;
|
|
int ofs_other = ofs[!axis];
|
|
|
|
ofs[axis] = ofs_dest;
|
|
if (check_bounds) {
|
|
if (ofs_other < 0 || ofs_other > limits[!axis]) {
|
|
/* Out of bounds. */
|
|
continue;
|
|
}
|
|
|
|
CLAMP(ofs_iter, 0, limits[axis]);
|
|
CLAMP(ofs_dest, 0, limits[axis]);
|
|
}
|
|
|
|
while (true) {
|
|
if (test_fn(arr + ofs_other + ofs_iter, user_data)) {
|
|
return true;
|
|
}
|
|
if (ofs_iter == ofs_dest) {
|
|
break;
|
|
}
|
|
ofs_iter += ofs_step;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
return false;
|
|
}
|