This repository has been archived on 2023-10-09.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/modifiers/intern/MOD_array.c
Campbell Barton 54310b8086 bmesh: internal api change, remove BMOElemMapping, its not needed since the values can be stored in the ghash directly.
saves 24 bytes per element (per vertex/edge/face on a 64bit system), for bmesh operators.
2013-09-02 03:13:51 +00:00

628 lines
18 KiB
C

/*
* ***** BEGIN GPL 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.
*
* 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.
*
* The Original Code is Copyright (C) 2005 by the Blender Foundation.
* All rights reserved.
*
* Contributor(s): Daniel Dunbar
* Ton Roosendaal,
* Ben Batt,
* Brecht Van Lommel,
* Campbell Barton
*
* ***** END GPL LICENSE BLOCK *****
*
*/
/** \file blender/modifiers/intern/MOD_array.c
* \ingroup modifiers
*/
/* Array modifier: duplicates the object multiple times along an axis */
#include "MEM_guardedalloc.h"
#include "BLI_math.h"
#include "BLI_utildefines.h"
#include "BLI_string.h"
#include "BLI_ghash.h"
#include "BLI_edgehash.h"
#include "DNA_curve_types.h"
#include "DNA_meshdata_types.h"
#include "DNA_object_types.h"
#include "DNA_scene_types.h"
#include "BKE_cdderivedmesh.h"
#include "BKE_displist.h"
#include "BKE_curve.h"
#include "BKE_mesh.h"
#include "BKE_modifier.h"
#include "BKE_object.h"
#include "bmesh.h"
#include "depsgraph_private.h"
#include <ctype.h>
#include <stdlib.h>
#include <string.h>
static void initData(ModifierData *md)
{
ArrayModifierData *amd = (ArrayModifierData *) md;
/* default to 2 duplicates distributed along the x-axis by an
* offset of 1 object-width
*/
amd->start_cap = amd->end_cap = amd->curve_ob = amd->offset_ob = NULL;
amd->count = 2;
zero_v3(amd->offset);
amd->scale[0] = 1;
amd->scale[1] = amd->scale[2] = 0;
amd->length = 0;
amd->merge_dist = 0.01;
amd->fit_type = MOD_ARR_FIXEDCOUNT;
amd->offset_type = MOD_ARR_OFF_RELATIVE;
amd->flags = 0;
}
static void copyData(ModifierData *md, ModifierData *target)
{
ArrayModifierData *amd = (ArrayModifierData *) md;
ArrayModifierData *tamd = (ArrayModifierData *) target;
tamd->start_cap = amd->start_cap;
tamd->end_cap = amd->end_cap;
tamd->curve_ob = amd->curve_ob;
tamd->offset_ob = amd->offset_ob;
tamd->count = amd->count;
copy_v3_v3(tamd->offset, amd->offset);
copy_v3_v3(tamd->scale, amd->scale);
tamd->length = amd->length;
tamd->merge_dist = amd->merge_dist;
tamd->fit_type = amd->fit_type;
tamd->offset_type = amd->offset_type;
tamd->flags = amd->flags;
}
static void foreachObjectLink(
ModifierData *md, Object *ob,
void (*walk)(void *userData, Object *ob, Object **obpoin),
void *userData)
{
ArrayModifierData *amd = (ArrayModifierData *) md;
walk(userData, ob, &amd->start_cap);
walk(userData, ob, &amd->end_cap);
walk(userData, ob, &amd->curve_ob);
walk(userData, ob, &amd->offset_ob);
}
static void updateDepgraph(ModifierData *md, DagForest *forest,
struct Scene *UNUSED(scene), Object *UNUSED(ob), DagNode *obNode)
{
ArrayModifierData *amd = (ArrayModifierData *) md;
if (amd->start_cap) {
DagNode *curNode = dag_get_node(forest, amd->start_cap);
dag_add_relation(forest, curNode, obNode,
DAG_RL_DATA_DATA | DAG_RL_OB_DATA, "Array Modifier");
}
if (amd->end_cap) {
DagNode *curNode = dag_get_node(forest, amd->end_cap);
dag_add_relation(forest, curNode, obNode,
DAG_RL_DATA_DATA | DAG_RL_OB_DATA, "Array Modifier");
}
if (amd->curve_ob) {
DagNode *curNode = dag_get_node(forest, amd->curve_ob);
dag_add_relation(forest, curNode, obNode,
DAG_RL_DATA_DATA | DAG_RL_OB_DATA, "Array Modifier");
}
if (amd->offset_ob) {
DagNode *curNode = dag_get_node(forest, amd->offset_ob);
dag_add_relation(forest, curNode, obNode,
DAG_RL_DATA_DATA | DAG_RL_OB_DATA, "Array Modifier");
}
}
static float vertarray_size(MVert *mvert, int numVerts, int axis)
{
int i;
float min_co, max_co;
/* if there are no vertices, width is 0 */
if (numVerts == 0) return 0;
/* find the minimum and maximum coordinates on the desired axis */
min_co = max_co = mvert->co[axis];
mvert++;
for (i = 1; i < numVerts; ++i, ++mvert) {
if (mvert->co[axis] < min_co) min_co = mvert->co[axis];
if (mvert->co[axis] > max_co) max_co = mvert->co[axis];
}
return max_co - min_co;
}
static int *find_doubles_index_map(BMesh *bm, BMOperator *dupe_op,
const ArrayModifierData *amd,
int *index_map_length)
{
BMOperator find_op;
BMOIter oiter;
BMVert *v, *v2;
BMElem *ele;
int *index_map, i;
BMO_op_initf(bm, &find_op, (BMO_FLAG_DEFAULTS & ~BMO_FLAG_RESPECT_HIDE),
"find_doubles verts=%av dist=%f keep_verts=%s",
amd->merge_dist, dupe_op, "geom");
BMO_op_exec(bm, &find_op);
i = 0;
BMO_ITER (ele, &oiter, dupe_op->slots_in, "geom", BM_ALL) {
BM_elem_index_set(ele, i); /* set_dirty */
i++;
}
BMO_ITER (ele, &oiter, dupe_op->slots_out, "geom.out", BM_ALL) {
BM_elem_index_set(ele, i); /* set_dirty */
i++;
}
/* above loops over all, so set all to dirty, if this is somehow
* setting valid values, this line can be removed - campbell */
bm->elem_index_dirty |= BM_VERT | BM_EDGE | BM_FACE;
(*index_map_length) = i;
index_map = MEM_callocN(sizeof(int) * (*index_map_length), "index_map");
/*element type argument doesn't do anything here*/
BMO_ITER (v, &oiter, find_op.slots_out, "targetmap.out", 0) {
v2 = BMO_iter_map_value_ptr(&oiter);
index_map[BM_elem_index_get(v)] = BM_elem_index_get(v2) + 1;
}
BMO_op_finish(bm, &find_op);
return index_map;
}
/* Used for start/end cap.
*
* this function expects all existing vertices to be tagged,
* so we can know new verts are not tagged.
*
* All verts will be tagged on exit.
*/
static void bm_merge_dm_transform(BMesh *bm, DerivedMesh *dm, float mat[4][4],
const ArrayModifierData *amd,
BMOperator *dupe_op,
BMOpSlot dupe_op_slot_args[BMO_OP_MAX_SLOTS], const char *dupe_slot_name,
BMOperator *weld_op)
{
const int is_input = (dupe_op->slots_in == dupe_op_slot_args);
BMVert *v, *v2, *v3;
BMIter iter;
/* Add the DerivedMesh's elements to the BMesh. The pre-existing
* elements were already tagged, so the new elements can be
* identified by not having the BM_ELEM_TAG flag set. */
DM_to_bmesh_ex(dm, bm, false);
if (amd->flags & MOD_ARR_MERGE) {
/* if merging is enabled, find doubles */
BMOIter oiter;
BMOperator find_op;
BMOpSlot *slot_targetmap;
BMO_op_initf(bm, &find_op, (BMO_FLAG_DEFAULTS & ~BMO_FLAG_RESPECT_HIDE),
is_input ? /* ugh */
"find_doubles verts=%Hv dist=%f keep_verts=%s" :
"find_doubles verts=%Hv dist=%f keep_verts=%S",
BM_ELEM_TAG, amd->merge_dist,
dupe_op, dupe_slot_name);
/* append the dupe's geom to the findop input verts */
if (is_input) {
BMO_slot_buffer_append(&find_op, slots_in, "verts",
dupe_op, slots_in, dupe_slot_name);
}
else if (dupe_op->slots_out == dupe_op_slot_args) {
BMO_slot_buffer_append(&find_op, slots_in, "verts",
dupe_op, slots_out, dupe_slot_name);
}
else {
BLI_assert(0);
}
/* transform and tag verts */
BM_ITER_MESH (v, &iter, bm, BM_VERTS_OF_MESH) {
if (!BM_elem_flag_test(v, BM_ELEM_TAG)) {
mul_m4_v3(mat, v->co);
BM_elem_flag_enable(v, BM_ELEM_TAG);
}
}
BMO_op_exec(bm, &find_op);
slot_targetmap = BMO_slot_get(weld_op->slots_in, "targetmap");
/* add new merge targets to weld operator */
BMO_ITER (v, &oiter, find_op.slots_out, "targetmap.out", 0) {
v2 = BMO_iter_map_value_ptr(&oiter);
/* check in case the target vertex (v2) is already marked
* for merging */
while ((v3 = BMO_slot_map_elem_get(slot_targetmap, v2))) {
v2 = v3;
}
BMO_slot_map_elem_insert(weld_op, slot_targetmap, v, v2);
}
BMO_op_finish(bm, &find_op);
}
else {
/* transform and tag verts */
BM_ITER_MESH (v, &iter, bm, BM_VERTS_OF_MESH) {
if (!BM_elem_flag_test(v, BM_ELEM_TAG)) {
mul_m4_v3(mat, v->co);
BM_elem_flag_enable(v, BM_ELEM_TAG);
}
}
}
}
static void merge_first_last(BMesh *bm,
const ArrayModifierData *amd,
BMOperator *dupe_first,
BMOperator *dupe_last,
BMOperator *weld_op)
{
BMOperator find_op;
BMOIter oiter;
BMVert *v, *v2;
BMOpSlot *slot_targetmap;
BMO_op_initf(bm, &find_op, (BMO_FLAG_DEFAULTS & ~BMO_FLAG_RESPECT_HIDE),
"find_doubles verts=%s dist=%f keep_verts=%s",
dupe_first, "geom", amd->merge_dist,
dupe_first, "geom");
/* append the last dupe's geom to the findop input verts */
BMO_slot_buffer_append(&find_op, slots_in, "verts",
dupe_last, slots_out, "geom.out");
BMO_op_exec(bm, &find_op);
/* add new merge targets to weld operator */
slot_targetmap = BMO_slot_get(weld_op->slots_in, "targetmap");
BMO_ITER (v, &oiter, find_op.slots_out, "targetmap.out", 0) {
if (!BMO_slot_map_contains(slot_targetmap, v)) {
v2 = BMO_iter_map_value_ptr(&oiter);
BMO_slot_map_elem_insert(weld_op, slot_targetmap, v, v2);
}
}
BMO_op_finish(bm, &find_op);
}
static DerivedMesh *arrayModifier_doArray(ArrayModifierData *amd,
Scene *scene, Object *ob, DerivedMesh *dm,
int UNUSED(initFlags))
{
DerivedMesh *result;
BMesh *bm = DM_to_bmesh(dm, false);
BMOperator first_dupe_op, dupe_op, old_dupe_op, weld_op;
BMVert **first_geom = NULL;
int i, j;
int index_len = -1; /* initialize to an invalid value */
/* offset matrix */
float offset[4][4];
float final_offset[4][4];
float length = amd->length;
int count = amd->count, maxVerts;
int *indexMap = NULL;
DerivedMesh *start_cap = NULL, *end_cap = NULL;
MVert *src_mvert;
BMOpSlot *slot_targetmap = NULL; /* for weld_op */
/* need to avoid infinite recursion here */
if (amd->start_cap && amd->start_cap != ob && amd->start_cap->type == OB_MESH)
start_cap = mesh_get_derived_final(scene, amd->start_cap, CD_MASK_MESH);
if (amd->end_cap && amd->end_cap != ob && amd->end_cap->type == OB_MESH)
end_cap = mesh_get_derived_final(scene, amd->end_cap, CD_MASK_MESH);
unit_m4(offset);
src_mvert = dm->getVertArray(dm);
maxVerts = dm->getNumVerts(dm);
if (amd->offset_type & MOD_ARR_OFF_CONST)
add_v3_v3v3(offset[3], offset[3], amd->offset);
if (amd->offset_type & MOD_ARR_OFF_RELATIVE) {
for (j = 0; j < 3; j++)
offset[3][j] += amd->scale[j] * vertarray_size(src_mvert, maxVerts, j);
}
if ((amd->offset_type & MOD_ARR_OFF_OBJ) && (amd->offset_ob)) {
float obinv[4][4];
float result_mat[4][4];
if (ob)
invert_m4_m4(obinv, ob->obmat);
else
unit_m4(obinv);
mul_serie_m4(result_mat, offset,
obinv, amd->offset_ob->obmat,
NULL, NULL, NULL, NULL, NULL);
copy_m4_m4(offset, result_mat);
}
if (amd->fit_type == MOD_ARR_FITCURVE && amd->curve_ob) {
Curve *cu = amd->curve_ob->data;
if (cu) {
float tmp_mat[3][3];
float scale;
BKE_object_to_mat3(amd->curve_ob, tmp_mat);
scale = mat3_to_scale(tmp_mat);
if (!amd->curve_ob->curve_cache || !amd->curve_ob->curve_cache->path) {
cu->flag |= CU_PATH; // needed for path & bevlist
BKE_displist_make_curveTypes(scene, amd->curve_ob, 0);
}
if (amd->curve_ob->curve_cache->path)
length = scale * amd->curve_ob->curve_cache->path->totdist;
}
}
/* calculate the maximum number of copies which will fit within the
* prescribed length */
if (amd->fit_type == MOD_ARR_FITLENGTH || amd->fit_type == MOD_ARR_FITCURVE) {
float dist = sqrt(dot_v3v3(offset[3], offset[3]));
if (dist > 1e-6f)
/* this gives length = first copy start to last copy end
* add a tiny offset for floating point rounding errors */
count = (length + 1e-6f) / dist;
else
/* if the offset has no translation, just make one copy */
count = 1;
}
if (count < 1)
count = 1;
/* calculate the offset matrix of the final copy (for merging) */
unit_m4(final_offset);
for (j = 0; j < count - 1; j++) {
float tmp_mat[4][4];
mul_m4_m4m4(tmp_mat, offset, final_offset);
copy_m4_m4(final_offset, tmp_mat);
}
/* BMESH_TODO: bumping up the stack level avoids computing the normals
* after every top-level operator execution (and this modifier has the
* potential to execute a *lot* of top-level BMOps. There should be a
* cleaner way to do this. One possibility: a "mirror" BMOp would
* certainly help by compressing it all into one top-level BMOp that
* executes a lot of second-level BMOps. */
BM_mesh_elem_toolflags_ensure(bm);
BMO_push(bm, NULL);
bmesh_edit_begin(bm, 0);
if (amd->flags & MOD_ARR_MERGE) {
BMO_op_init(bm, &weld_op, (BMO_FLAG_DEFAULTS & ~BMO_FLAG_RESPECT_HIDE),
"weld_verts");
slot_targetmap = BMO_slot_get(weld_op.slots_in, "targetmap");
}
BMO_op_initf(bm, &dupe_op, (BMO_FLAG_DEFAULTS & ~BMO_FLAG_RESPECT_HIDE),
"duplicate geom=%avef");
first_dupe_op = dupe_op;
for (j = 0; j < count - 1; j++) {
BMVert *v, *v2, *v3;
BMOpSlot *geom_slot;
BMOpSlot *geom_out_slot;
BMOIter oiter;
if (j != 0) {
BMO_op_initf(bm, &dupe_op,
(BMO_FLAG_DEFAULTS & ~BMO_FLAG_RESPECT_HIDE),
"duplicate geom=%S", &old_dupe_op, "geom.out");
}
BMO_op_exec(bm, &dupe_op);
geom_slot = BMO_slot_get(dupe_op.slots_in, "geom");
geom_out_slot = BMO_slot_get(dupe_op.slots_out, "geom.out");
if ((amd->flags & MOD_ARR_MERGEFINAL) && j == 0) {
int first_geom_bytes = sizeof(BMVert *) * geom_slot->len;
/* make a copy of the initial geometry ordering so the
* last duplicate can be merged into it */
first_geom = MEM_mallocN(first_geom_bytes, "first_geom");
memcpy(first_geom, geom_slot->data.buf, first_geom_bytes);
}
/* apply transformation matrix */
BMO_ITER (v, &oiter, dupe_op.slots_out, "geom.out", BM_VERT) {
mul_m4_v3(offset, v->co);
}
if (amd->flags & MOD_ARR_MERGE) {
/*calculate merge mapping*/
if (j == 0) {
indexMap = find_doubles_index_map(bm, &dupe_op,
amd, &index_len);
}
#define _E(s, i) ((BMVert **)(s)->data.buf)[i]
/* ensure this is set */
BLI_assert(index_len != -1);
for (i = 0; i < index_len; i++) {
if (!indexMap[i]) continue;
/* merge v (from 'geom.out') into v2 (from old 'geom') */
v = _E(geom_out_slot, i - geom_slot->len);
v2 = _E(geom_slot, indexMap[i] - 1);
/* check in case the target vertex (v2) is already marked
* for merging */
while ((v3 = BMO_slot_map_elem_get(slot_targetmap, v2))) {
v2 = v3;
}
BMO_slot_map_elem_insert(&weld_op, slot_targetmap, v, v2);
}
#undef _E
}
/* already copied earlier, but after executation more slot
* memory may be allocated */
if (j == 0)
first_dupe_op = dupe_op;
if (j >= 2)
BMO_op_finish(bm, &old_dupe_op);
old_dupe_op = dupe_op;
}
if ((amd->flags & MOD_ARR_MERGE) &&
(amd->flags & MOD_ARR_MERGEFINAL) &&
(count > 1))
{
/* Merge first and last copies. Note that we can't use the
* indexMap for this because (unless the array is forming a
* loop) the offset between first and last is different from
* dupe X to dupe X+1. */
merge_first_last(bm, amd, &first_dupe_op, &dupe_op, &weld_op);
}
/* start capping */
if (start_cap || end_cap) {
BM_mesh_elem_hflag_enable_all(bm, BM_VERT, BM_ELEM_TAG, FALSE);
if (start_cap) {
float startoffset[4][4];
invert_m4_m4(startoffset, offset);
bm_merge_dm_transform(bm, start_cap, startoffset, amd,
&first_dupe_op, first_dupe_op.slots_in, "geom", &weld_op);
}
if (end_cap) {
float endoffset[4][4];
mul_m4_m4m4(endoffset, offset, final_offset);
bm_merge_dm_transform(bm, end_cap, endoffset, amd,
&dupe_op, (count == 1) ? dupe_op.slots_in : dupe_op.slots_out,
(count == 1) ? "geom" : "geom.out", &weld_op);
}
}
/* done capping */
/* free remaining dupe operators */
BMO_op_finish(bm, &first_dupe_op);
if (count > 2)
BMO_op_finish(bm, &dupe_op);
/* run merge operator */
if (amd->flags & MOD_ARR_MERGE) {
BMO_op_exec(bm, &weld_op);
BMO_op_finish(bm, &weld_op);
}
/* Bump the stack level back down to match the adjustment up above */
BMO_pop(bm);
result = CDDM_from_bmesh(bm, FALSE);
if ((dm->dirty & DM_DIRTY_NORMALS) ||
((amd->offset_type & MOD_ARR_OFF_OBJ) && (amd->offset_ob)))
{
/* Update normals in case offset object has rotation. */
result->dirty |= DM_DIRTY_NORMALS;
}
BM_mesh_free(bm);
if (indexMap)
MEM_freeN(indexMap);
if (first_geom)
MEM_freeN(first_geom);
return result;
}
static DerivedMesh *applyModifier(ModifierData *md, Object *ob,
DerivedMesh *dm,
ModifierApplyFlag UNUSED(flag))
{
DerivedMesh *result;
ArrayModifierData *amd = (ArrayModifierData *) md;
result = arrayModifier_doArray(amd, md->scene, ob, dm, 0);
return result;
}
ModifierTypeInfo modifierType_Array = {
/* name */ "Array",
/* structName */ "ArrayModifierData",
/* structSize */ sizeof(ArrayModifierData),
/* type */ eModifierTypeType_Constructive,
/* flags */ eModifierTypeFlag_AcceptsMesh |
eModifierTypeFlag_SupportsMapping |
eModifierTypeFlag_SupportsEditmode |
eModifierTypeFlag_EnableInEditmode |
eModifierTypeFlag_AcceptsCVs,
/* copyData */ copyData,
/* deformVerts */ NULL,
/* deformMatrices */ NULL,
/* deformVertsEM */ NULL,
/* deformMatricesEM */ NULL,
/* applyModifier */ applyModifier,
/* applyModifierEM */ NULL,
/* initData */ initData,
/* requiredDataMask */ NULL,
/* freeData */ NULL,
/* isDisabled */ NULL,
/* updateDepgraph */ updateDepgraph,
/* dependsOnTime */ NULL,
/* dependsOnNormals */ NULL,
/* foreachObjectLink */ foreachObjectLink,
/* foreachIDLink */ NULL,
/* foreachTexLink */ NULL,
};