Welcome to mirror list, hosted at ThFree Co, Russian Federation.

git.blender.org/blender.git - Unnamed repository; edit this file 'description' to name the repository.
summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat (limited to 'source/blender/modifiers/intern/MOD_weighted_normal.c')
-rw-r--r--source/blender/modifiers/intern/MOD_weighted_normal.c671
1 files changed, 671 insertions, 0 deletions
diff --git a/source/blender/modifiers/intern/MOD_weighted_normal.c b/source/blender/modifiers/intern/MOD_weighted_normal.c
new file mode 100644
index 00000000000..78a994a7107
--- /dev/null
+++ b/source/blender/modifiers/intern/MOD_weighted_normal.c
@@ -0,0 +1,671 @@
+/*
+ * ***** 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.
+ *
+ * ***** END GPL LICENSE BLOCK *****
+ *
+ */
+
+/** \file blender/modifiers/intern/MOD_weighted_normal.c
+ * \ingroup modifiers
+ */
+
+#include "MEM_guardedalloc.h"
+
+#include "DNA_mesh_types.h"
+#include "DNA_object_types.h"
+#include "DNA_scene_types.h"
+
+#include "BKE_cdderivedmesh.h"
+#include "BKE_deform.h"
+#include "BKE_library.h"
+#include "BKE_library_query.h"
+#include "BKE_mesh.h"
+
+#include "BLI_math.h"
+#include "BLI_linklist.h"
+
+#include "MOD_modifiertypes.h"
+#include "MOD_util.h"
+
+#define CLNORS_VALID_VEC_LEN (1e-6f)
+
+typedef struct ModePair {
+ float val; /* Contains mode based value (face area / corner angle). */
+ int index; /* Index value per poly or per loop. */
+} ModePair;
+
+/* Sorting function used in modifier, sorts in decreasing order. */
+static int modepair_cmp_by_val_inverse(const void *p1, const void *p2)
+{
+ ModePair *r1 = (ModePair *)p1;
+ ModePair *r2 = (ModePair *)p2;
+
+ return (r1->val < r2->val) ? 1 : ((r1->val > r2->val) ? -1 : 0);
+}
+
+/* There will be one of those per vertex (simple case, computing one normal per vertex), or per smooth fan. */
+typedef struct WeightedNormalDataAggregateItem {
+ float normal[3];
+
+ int num_loops; /* Count number of loops using this item so far. */
+ float curr_val; /* Current max val for this item. */
+ int curr_strength; /* Current max strength encountered for this item. */
+} WeightedNormalDataAggregateItem;
+
+#define NUM_CACHED_INVERSE_POWERS_OF_WEIGHT 128
+
+typedef struct WeightedNormalData {
+ const int numVerts;
+ const int numEdges;
+ const int numLoops;
+ const int numPolys;
+
+ MVert *mvert;
+ MEdge *medge;
+
+ MLoop *mloop;
+ short (*clnors)[2];
+ const bool has_clnors; /* True if clnors already existed, false if we had to create them. */
+ const float split_angle;
+
+ MPoly *mpoly;
+ float (*polynors)[3];
+ int *poly_strength;
+
+ MDeformVert *dvert;
+ const int defgrp_index;
+ const bool use_invert_vgroup;
+
+ const float weight;
+ const short mode;
+
+ /* Lower-level, internal processing data. */
+ float cached_inverse_powers_of_weight[NUM_CACHED_INVERSE_POWERS_OF_WEIGHT];
+
+ WeightedNormalDataAggregateItem *items_data;
+
+ ModePair *mode_pair;
+
+ int *loop_to_poly;
+} WeightedNormalData;
+
+/* Check strength of given poly compared to those found so far for that given item (vertex or smooth fan),
+ * and reset matching item_data in case we get a stronger new strength. */
+static bool check_item_poly_strength(
+ WeightedNormalData *wn_data, WeightedNormalDataAggregateItem *item_data, const int mp_index)
+{
+ BLI_assert (wn_data->poly_strength != NULL);
+
+ const int mp_strength = wn_data->poly_strength[mp_index];
+
+ if (mp_strength > item_data->curr_strength) {
+ item_data->curr_strength = mp_strength;
+ item_data->curr_val = 0.0f;
+ item_data->num_loops = 0;
+ zero_v3(item_data->normal);
+ }
+
+ return mp_strength == item_data->curr_strength;
+}
+
+static void aggregate_item_normal(
+ WeightedNormalModifierData *wnmd, WeightedNormalData *wn_data,
+ WeightedNormalDataAggregateItem *item_data,
+ const int mv_index, const int mp_index,
+ const float curr_val, const bool use_face_influence)
+{
+ float (*polynors)[3] = wn_data->polynors;
+
+ MDeformVert *dvert = wn_data->dvert;
+ const int defgrp_index = wn_data->defgrp_index;
+ const bool use_invert_vgroup = wn_data->use_invert_vgroup;
+
+ const float weight = wn_data->weight;
+
+ float *cached_inverse_powers_of_weight = wn_data->cached_inverse_powers_of_weight;
+
+ const bool has_vgroup = dvert != NULL;
+ const bool vert_of_group = has_vgroup && defvert_find_index(&dvert[mv_index], defgrp_index) != NULL;
+
+ if (has_vgroup && ((vert_of_group && use_invert_vgroup) || (!vert_of_group && !use_invert_vgroup))) {
+ return;
+ }
+
+ if (use_face_influence && !check_item_poly_strength(wn_data, item_data, mp_index)) {
+ return;
+ }
+
+ /* If item's curr_val is 0 init it to present value. */
+ if (item_data->curr_val == 0.0f) {
+ item_data->curr_val = curr_val;
+ }
+ if (!compare_ff(item_data->curr_val, curr_val, wnmd->thresh)) {
+ /* item's curr_val and present value differ more than threshold, update. */
+ item_data->num_loops++;
+ item_data->curr_val = curr_val;
+ }
+
+ /* Exponentially divided weight for each normal (since a few values will be used by most cases, we cache those). */
+ const int num_loops = item_data->num_loops;
+ if (num_loops < NUM_CACHED_INVERSE_POWERS_OF_WEIGHT && cached_inverse_powers_of_weight[num_loops] == 0.0f) {
+ cached_inverse_powers_of_weight[num_loops] = 1.0f / powf(weight, num_loops);
+ }
+ const float inverted_n_weight = num_loops < NUM_CACHED_INVERSE_POWERS_OF_WEIGHT ?
+ cached_inverse_powers_of_weight[num_loops] : 1.0f / powf(weight, num_loops);
+
+ madd_v3_v3fl(item_data->normal, polynors[mp_index], curr_val * inverted_n_weight);
+}
+
+static void apply_weights_vertex_normal(WeightedNormalModifierData *wnmd, WeightedNormalData *wn_data)
+{
+ const int numVerts = wn_data->numVerts;
+ const int numEdges = wn_data->numEdges;
+ const int numLoops = wn_data->numLoops;
+ const int numPolys = wn_data->numPolys;
+
+ MVert *mvert = wn_data->mvert;
+ MEdge *medge = wn_data->medge;
+
+ MLoop *mloop = wn_data->mloop;
+ short (*clnors)[2] = wn_data->clnors;
+ int *loop_to_poly = wn_data->loop_to_poly;
+
+ MPoly *mpoly = wn_data->mpoly;
+ float (*polynors)[3] = wn_data->polynors;
+ int *poly_strength = wn_data->poly_strength;
+
+ MDeformVert *dvert = wn_data->dvert;
+
+ const short mode = wn_data->mode;
+ ModePair *mode_pair = wn_data->mode_pair;
+
+ const bool has_clnors = wn_data->has_clnors;
+ const float split_angle = wn_data->split_angle;
+ MLoopNorSpaceArray lnors_spacearr = {NULL};
+
+ const bool keep_sharp = (wnmd->flag & MOD_WEIGHTEDNORMAL_KEEP_SHARP) != 0;
+ const bool use_face_influence = (wnmd->flag & MOD_WEIGHTEDNORMAL_FACE_INFLUENCE) != 0 && poly_strength != NULL;
+ const bool has_vgroup = dvert != NULL;
+
+ float (*loop_normals)[3] = NULL;
+
+ WeightedNormalDataAggregateItem *items_data = NULL;
+ int num_items = 0;
+ if (keep_sharp) {
+ BLI_bitmap *done_loops = BLI_BITMAP_NEW(numLoops, __func__);
+
+ /* This will give us loop normal spaces, we do not actually care about computed loop_normals for now... */
+ loop_normals = MEM_calloc_arrayN((size_t)numLoops, sizeof(*loop_normals), __func__);
+ BKE_mesh_normals_loop_split(mvert, numVerts, medge, numEdges,
+ mloop, loop_normals, numLoops, mpoly, polynors, numPolys,
+ true, split_angle, &lnors_spacearr, has_clnors ? clnors : NULL, loop_to_poly);
+
+ num_items = lnors_spacearr.num_spaces;
+ items_data = MEM_calloc_arrayN((size_t)num_items, sizeof(*items_data), __func__);
+
+ /* In this first loop, we assign each WeightedNormalDataAggregateItem
+ * to its smooth fan of loops (aka lnor space). */
+ MPoly *mp;
+ int mp_index;
+ int item_index;
+ for (mp = mpoly, mp_index = 0, item_index = 0; mp_index < numPolys; mp++, mp_index++) {
+ int ml_index = mp->loopstart;
+ const int ml_end_index = ml_index + mp->totloop;
+
+ for (; ml_index < ml_end_index; ml_index++) {
+ if (BLI_BITMAP_TEST(done_loops, ml_index)) {
+ /* Smooth fan of this loop has already been processed, skip it. */
+ continue;
+ }
+ BLI_assert(item_index < num_items);
+
+ WeightedNormalDataAggregateItem *itdt = &items_data[item_index];
+ itdt->curr_strength = FACE_STRENGTH_WEAK;
+
+ MLoopNorSpace *lnor_space = lnors_spacearr.lspacearr[ml_index];
+ lnor_space->user_data = itdt;
+
+ if (!(lnor_space->flags & MLNOR_SPACE_IS_SINGLE)) {
+ for (LinkNode *lnode = lnor_space->loops; lnode; lnode = lnode->next) {
+ const int ml_fan_index = POINTER_AS_INT(lnode->link);
+ BLI_BITMAP_ENABLE(done_loops, ml_fan_index);
+ }
+ }
+ else {
+ BLI_BITMAP_ENABLE(done_loops, ml_index);
+ }
+
+ item_index++;
+ }
+ }
+
+ MEM_freeN(done_loops);
+ }
+ else {
+ num_items = numVerts;
+ items_data = MEM_calloc_arrayN((size_t)num_items, sizeof(*items_data), __func__);
+ if (use_face_influence) {
+ for (int item_index = 0; item_index < num_items; item_index++) {
+ items_data[item_index].curr_strength = FACE_STRENGTH_WEAK;
+ }
+ }
+ }
+ wn_data->items_data = items_data;
+
+ switch (mode) {
+ case MOD_WEIGHTEDNORMAL_MODE_FACE:
+ for (int i = 0; i < numPolys; i++) {
+ const int mp_index = mode_pair[i].index;
+ const float mp_val = mode_pair[i].val;
+
+ int ml_index = mpoly[mp_index].loopstart;
+ const int ml_index_end = ml_index + mpoly[mp_index].totloop;
+ for (; ml_index < ml_index_end; ml_index++) {
+ const int mv_index = mloop[ml_index].v;
+ WeightedNormalDataAggregateItem *item_data = keep_sharp ?
+ lnors_spacearr.lspacearr[ml_index]->user_data:
+ &items_data[mv_index];
+
+ aggregate_item_normal(wnmd, wn_data, item_data, mv_index, mp_index, mp_val, use_face_influence);
+ }
+ }
+ break;
+ case MOD_WEIGHTEDNORMAL_MODE_ANGLE:
+ case MOD_WEIGHTEDNORMAL_MODE_FACE_ANGLE:
+ BLI_assert(loop_to_poly != NULL);
+
+ for (int i = 0; i < numLoops; i++) {
+ const int ml_index = mode_pair[i].index;
+ const float ml_val = mode_pair[i].val;
+
+ const int mp_index = loop_to_poly[ml_index];
+ const int mv_index = mloop[ml_index].v;
+ WeightedNormalDataAggregateItem *item_data = keep_sharp ?
+ lnors_spacearr.lspacearr[ml_index]->user_data:
+ &items_data[mv_index];
+
+ aggregate_item_normal(wnmd, wn_data, item_data, mv_index, mp_index, ml_val, use_face_influence);
+ }
+ break;
+ default:
+ BLI_assert(0);
+ }
+
+ /* Validate computed weighted normals. */
+ for (int item_index = 0; item_index < num_items; item_index++) {
+ if (normalize_v3(items_data[item_index].normal) < CLNORS_VALID_VEC_LEN) {
+ zero_v3(items_data[item_index].normal);
+ }
+ }
+
+ if (keep_sharp) {
+ /* Set loop normals for normal computed for each lnor space (smooth fan).
+ * Note that loop_normals is already populated with clnors (before this modifier is applied, at start of
+ * this function), so no need to recompute them here. */
+ for (int ml_index = 0; ml_index < numLoops; ml_index++) {
+ WeightedNormalDataAggregateItem *item_data = lnors_spacearr.lspacearr[ml_index]->user_data;
+ if (!is_zero_v3(item_data->normal)) {
+ copy_v3_v3(loop_normals[ml_index], item_data->normal);
+ }
+ }
+
+ BKE_mesh_normals_loop_custom_set(mvert, numVerts, medge, numEdges,
+ mloop, loop_normals, numLoops, mpoly, polynors, numPolys, clnors);
+ }
+ else {
+ /* TODO: Ideally, we could add an option to BKE_mesh_normals_loop_custom_[from_vertices_]set() to keep current
+ * clnors instead of resetting them to default autocomputed ones, when given new custom normal is zero-vec.
+ * But this is not exactly trivial change, better to keep this optimization for later...
+ */
+ if (!has_vgroup) {
+ /* Note: in theory, we could avoid this extra allocation & copying... But think we can live with it for now,
+ * and it makes code simpler & cleaner. */
+ float (*vert_normals)[3] = MEM_calloc_arrayN((size_t)numVerts, sizeof(*loop_normals), __func__);
+
+ for (int ml_index = 0; ml_index < numLoops; ml_index++) {
+ const int mv_index = mloop[ml_index].v;
+ copy_v3_v3(vert_normals[mv_index], items_data[mv_index].normal);
+ }
+
+ BKE_mesh_normals_loop_custom_from_vertices_set(mvert, vert_normals, numVerts, medge, numEdges,
+ mloop, numLoops, mpoly, polynors, numPolys, clnors);
+
+ MEM_freeN(vert_normals);
+ }
+ else {
+ loop_normals = MEM_calloc_arrayN((size_t)numLoops, sizeof(*loop_normals), __func__);
+
+ BKE_mesh_normals_loop_split(mvert, numVerts, medge, numEdges,
+ mloop, loop_normals, numLoops, mpoly, polynors, numPolys,
+ true, split_angle, NULL, has_clnors ? clnors : NULL, loop_to_poly);
+
+ for (int ml_index = 0; ml_index < numLoops; ml_index++) {
+ const int item_index = mloop[ml_index].v;
+ if (!is_zero_v3(items_data[item_index].normal)) {
+ copy_v3_v3(loop_normals[ml_index], items_data[item_index].normal);
+ }
+ }
+
+ BKE_mesh_normals_loop_custom_set(mvert, numVerts, medge, numEdges,
+ mloop, loop_normals, numLoops, mpoly, polynors, numPolys, clnors);
+ }
+ }
+
+ if (keep_sharp) {
+ BKE_lnor_spacearr_free(&lnors_spacearr);
+ }
+ MEM_SAFE_FREE(loop_normals);
+}
+
+static void wn_face_area(WeightedNormalModifierData *wnmd, WeightedNormalData *wn_data)
+{
+ const int numPolys = wn_data->numPolys;
+
+ MVert *mvert = wn_data->mvert;
+ MLoop *mloop = wn_data->mloop;
+ MPoly *mpoly = wn_data->mpoly;
+
+ MPoly *mp;
+ int mp_index;
+
+ ModePair *face_area = MEM_malloc_arrayN((size_t)numPolys, sizeof(*face_area), __func__);
+
+ ModePair *f_area = face_area;
+ for (mp_index = 0, mp = mpoly; mp_index < numPolys; mp_index++, mp++, f_area++) {
+ f_area->val = BKE_mesh_calc_poly_area(mp, &mloop[mp->loopstart], mvert);
+ f_area->index = mp_index;
+ }
+
+ qsort(face_area, numPolys, sizeof(*face_area), modepair_cmp_by_val_inverse);
+
+ wn_data->mode_pair = face_area;
+ apply_weights_vertex_normal(wnmd, wn_data);
+}
+
+static void wn_corner_angle(WeightedNormalModifierData *wnmd, WeightedNormalData *wn_data)
+{
+ const int numLoops = wn_data->numLoops;
+ const int numPolys = wn_data->numPolys;
+
+ MVert *mvert = wn_data->mvert;
+ MLoop *mloop = wn_data->mloop;
+ MPoly *mpoly = wn_data->mpoly;
+
+ MPoly *mp;
+ int mp_index;
+
+ int *loop_to_poly = MEM_malloc_arrayN((size_t)numLoops, sizeof(*loop_to_poly), __func__);
+
+ ModePair *corner_angle = MEM_malloc_arrayN((size_t)numLoops, sizeof(*corner_angle), __func__);
+
+ for (mp_index = 0, mp = mpoly; mp_index < numPolys; mp_index++, mp++) {
+ MLoop *ml_start = &mloop[mp->loopstart];
+
+ float *index_angle = MEM_malloc_arrayN((size_t)mp->totloop, sizeof(*index_angle), __func__);
+ BKE_mesh_calc_poly_angles(mp, ml_start, mvert, index_angle);
+
+ ModePair *c_angl = &corner_angle[mp->loopstart];
+ float *angl = index_angle;
+ for (int ml_index = mp->loopstart; ml_index < mp->loopstart + mp->totloop; ml_index++, c_angl++, angl++) {
+ c_angl->val = (float)M_PI - *angl;
+ c_angl->index = ml_index;
+
+ loop_to_poly[ml_index] = mp_index;
+ }
+ MEM_freeN(index_angle);
+ }
+
+ qsort(corner_angle, numLoops, sizeof(*corner_angle), modepair_cmp_by_val_inverse);
+
+ wn_data->loop_to_poly = loop_to_poly;
+ wn_data->mode_pair = corner_angle;
+ apply_weights_vertex_normal(wnmd, wn_data);
+}
+
+static void wn_face_with_angle(WeightedNormalModifierData *wnmd, WeightedNormalData *wn_data)
+{
+ const int numLoops = wn_data->numLoops;
+ const int numPolys = wn_data->numPolys;
+
+ MVert *mvert = wn_data->mvert;
+ MLoop *mloop = wn_data->mloop;
+ MPoly *mpoly = wn_data->mpoly;
+
+ MPoly *mp;
+ int mp_index;
+
+ int *loop_to_poly = MEM_malloc_arrayN((size_t)numLoops, sizeof(*loop_to_poly), __func__);
+
+ ModePair *combined = MEM_malloc_arrayN((size_t)numLoops, sizeof(*combined), __func__);
+
+ for (mp_index = 0, mp = mpoly; mp_index < numPolys; mp_index++, mp++) {
+ MLoop *ml_start = &mloop[mp->loopstart];
+
+ float face_area = BKE_mesh_calc_poly_area(mp, ml_start, mvert);
+ float *index_angle = MEM_malloc_arrayN((size_t)mp->totloop, sizeof(*index_angle), __func__);
+ BKE_mesh_calc_poly_angles(mp, ml_start, mvert, index_angle);
+
+ ModePair *cmbnd = &combined[mp->loopstart];
+ float *angl = index_angle;
+ for (int ml_index = mp->loopstart; ml_index < mp->loopstart + mp->totloop; ml_index++, cmbnd++, angl++) {
+ /* In this case val is product of corner angle and face area. */
+ cmbnd->val = ((float)M_PI - *angl) * face_area;
+ cmbnd->index = ml_index;
+
+ loop_to_poly[ml_index] = mp_index;
+ }
+ MEM_freeN(index_angle);
+ }
+
+ qsort(combined, numLoops, sizeof(*combined), modepair_cmp_by_val_inverse);
+
+ wn_data->loop_to_poly = loop_to_poly;
+ wn_data->mode_pair = combined;
+ apply_weights_vertex_normal(wnmd, wn_data);
+}
+
+static Mesh *applyModifier(ModifierData *md, const ModifierEvalContext *ctx, Mesh *mesh)
+{
+ WeightedNormalModifierData *wnmd = (WeightedNormalModifierData *)md;
+ Object *ob = ctx->object;
+
+ /* XXX TODO ARG GRRR XYQWNMPRXTYY
+ * Once we fully switch to Mesh evaluation of modifiers, we can expect to get that flag from the COW copy.
+ * But for now, it is lost in the DM intermediate step, so we need to directly check orig object's data. */
+#if 0
+ if (!(mesh->flag & ME_AUTOSMOOTH)) {
+#else
+ if (!(((Mesh *)ob->data)->flag & ME_AUTOSMOOTH)) {
+#endif
+ modifier_setError((ModifierData *)wnmd, "Enable 'Auto Smooth' option in mesh settings");
+ return mesh;
+ }
+
+ Mesh *result;
+ BKE_id_copy_ex(
+ NULL, &mesh->id, (ID **)&result,
+ LIB_ID_CREATE_NO_MAIN |
+ LIB_ID_CREATE_NO_USER_REFCOUNT |
+ LIB_ID_CREATE_NO_DEG_TAG |
+ LIB_ID_COPY_NO_PREVIEW,
+ false);
+
+ const int numVerts = result->totvert;
+ const int numEdges = result->totedge;
+ const int numLoops = result->totloop;
+ const int numPolys = result->totpoly;
+
+ MEdge *medge = result->medge;
+ MPoly *mpoly = result->mpoly;
+ MVert *mvert = result->mvert;
+ MLoop *mloop = result->mloop;
+
+ /* Right now:
+ * If weight = 50 then all faces are given equal weight.
+ * If weight > 50 then more weight given to faces with larger vals (face area / corner angle).
+ * If weight < 50 then more weight given to faces with lesser vals. However current calculation
+ * does not converge to min/max.
+ */
+ float weight = ((float)wnmd->weight) / 50.0f;
+ if (wnmd->weight == 100) {
+ weight = (float)SHRT_MAX;
+ }
+ else if (wnmd->weight == 1) {
+ weight = 1 / (float)SHRT_MAX;
+ }
+ else if ((weight - 1) * 25 > 1) {
+ weight = (weight - 1) * 25;
+ }
+
+ CustomData *pdata = &result->pdata;
+ float (*polynors)[3] = CustomData_get_layer(pdata, CD_NORMAL);
+ if (!polynors) {
+ polynors = CustomData_add_layer(pdata, CD_NORMAL, CD_CALLOC, NULL, numPolys);
+ }
+ BKE_mesh_calc_normals_poly(mvert, NULL, numVerts, mloop, mpoly, numLoops, numPolys, polynors, false);
+
+
+ const float split_angle = mesh->smoothresh;
+ short (*clnors)[2];
+ CustomData *ldata = &result->ldata;
+ clnors = CustomData_get_layer(ldata, CD_CUSTOMLOOPNORMAL);
+
+ /* Keep info whether we had clnors, it helps when generating clnor spaces and default normals. */
+ const bool has_clnors = clnors != NULL;
+ if (!clnors) {
+ clnors = CustomData_add_layer(ldata, CD_CUSTOMLOOPNORMAL, CD_CALLOC, NULL, numLoops);
+ clnors = CustomData_get_layer(ldata, CD_CUSTOMLOOPNORMAL);
+ }
+
+ MDeformVert *dvert;
+ int defgrp_index;
+ MOD_get_vgroup(ctx->object, mesh, wnmd->defgrp_name, &dvert, &defgrp_index);
+
+ WeightedNormalData wn_data = {
+ .numVerts = numVerts,
+ .numEdges = numEdges,
+ .numLoops = numLoops,
+ .numPolys = numPolys,
+
+ .mvert = mvert,
+ .medge = medge,
+
+ .mloop = mloop,
+ .clnors = clnors,
+ .has_clnors = has_clnors,
+ .split_angle = split_angle,
+
+ .mpoly = mpoly,
+ .polynors = polynors,
+ .poly_strength = CustomData_get_layer_named(
+ &result->pdata, CD_PROP_INT,
+ MOD_WEIGHTEDNORMALS_FACEWEIGHT_CDLAYER_ID),
+
+ .dvert = dvert,
+ .defgrp_index = defgrp_index,
+ .use_invert_vgroup = (wnmd->flag & MOD_WEIGHTEDNORMAL_INVERT_VGROUP) != 0,
+
+ .weight = weight,
+ .mode = wnmd->mode,
+ };
+
+ switch (wnmd->mode) {
+ case MOD_WEIGHTEDNORMAL_MODE_FACE:
+ wn_face_area(wnmd, &wn_data);
+ break;
+ case MOD_WEIGHTEDNORMAL_MODE_ANGLE:
+ wn_corner_angle(wnmd, &wn_data);
+ break;
+ case MOD_WEIGHTEDNORMAL_MODE_FACE_ANGLE:
+ wn_face_with_angle(wnmd, &wn_data);
+ break;
+ }
+
+ MEM_SAFE_FREE(wn_data.loop_to_poly);
+ MEM_SAFE_FREE(wn_data.mode_pair);
+ MEM_SAFE_FREE(wn_data.items_data);
+
+ return result;
+}
+
+static void initData(ModifierData *md)
+{
+ WeightedNormalModifierData *wnmd = (WeightedNormalModifierData *)md;
+ wnmd->mode = MOD_WEIGHTEDNORMAL_MODE_FACE;
+ wnmd->weight = 50;
+ wnmd->thresh = 1e-2f;
+ wnmd->flag = 0;
+}
+
+static CustomDataMask requiredDataMask(Object *UNUSED(ob), ModifierData *md)
+{
+ WeightedNormalModifierData *wnmd = (WeightedNormalModifierData *)md;
+ CustomDataMask dataMask = CD_CUSTOMLOOPNORMAL;
+
+ if (wnmd->defgrp_name[0]) {
+ dataMask |= CD_MASK_MDEFORMVERT;
+ }
+
+ if (wnmd->flag & MOD_WEIGHTEDNORMAL_FACE_INFLUENCE) {
+ dataMask |= CD_MASK_PROP_INT;
+ }
+
+ return dataMask;
+}
+
+static bool dependsOnNormals(ModifierData *UNUSED(md))
+{
+ return true;
+}
+
+ModifierTypeInfo modifierType_WeightedNormal = {
+ /* name */ "Weighted Normal",
+ /* structName */ "WeightedNormalModifierData",
+ /* structSize */ sizeof(WeightedNormalModifierData),
+ /* type */ eModifierTypeType_Constructive,
+ /* flags */ eModifierTypeFlag_AcceptsMesh |
+ eModifierTypeFlag_SupportsMapping |
+ eModifierTypeFlag_SupportsEditmode |
+ eModifierTypeFlag_EnableInEditmode,
+
+ /* copyData */ modifier_copyData_generic,
+
+ /* deformVerts_DM */ NULL,
+ /* deformMatrices_DM */ NULL,
+ /* deformVertsEM_DM */ NULL,
+ /* deformMatricesEM_DM*/NULL,
+ /* applyModifier_DM */ NULL,
+ /* applyModifierEM_DM */NULL,
+
+ /* deformVerts */ NULL,
+ /* deformMatrices */ NULL,
+ /* deformVertsEM */ NULL,
+ /* deformMatricesEM */ NULL,
+ /* applyModifier */ applyModifier,
+ /* applyModifierEM */ NULL,
+
+ /* initData */ initData,
+ /* requiredDataMask */ requiredDataMask,
+ /* freeData */ NULL,
+ /* isDisabled */ NULL,
+ /* updateDepsgraph */ NULL,
+ /* dependsOnTime */ NULL,
+ /* dependsOnNormals */ dependsOnNormals,
+ /* foreachObjectLink */ NULL,
+ /* foreachIDLink */ NULL,
+ /* foreachTexLink */ NULL,
+};