Added subsurface levels option to normal projection.

Now when doing normal projeciton is possible to ask it to project along the normals
that the vertex would have if it was subsurfaced... this gives "better" projections on
low polys.

Point of this commit is to add the feature request from Eclectiel
http://blenderartists.org/forum/showpost.php?p=1181531&postcount=9

workflow as Eclectiel wanted is now possible:

to archieve a nice low-res retopology that aproximates a high-res mesh when subsurfaced:
1 - make base low-poly retopo
2 - apply a shrinkwrap with projection along normal.. and with SS level = N (where N>0)
3 - add a Subsurface with level N
4 - run bretch's script (ss_fit)

1 files changed, 63 insertions, 11 deletions

diff --git a/source/blender/blenkernel/intern/shrinkwrap.c b/source/blender/blenkernel/intern/shrinkwrap.c
index fda1fbeab56..5c020bf2339 100644
--- a/source/blender/blenkernel/intern/shrinkwrap.c
+++ b/source/blender/blenkernel/intern/shrinkwrap.c
@@ -32,10 +32,12 @@
 #include <memory.h>
 #include <stdio.h>
 #include <time.h>
+#include <assert.h>
 
 #include "DNA_object_types.h"
 #include "DNA_modifier_types.h"
 #include "DNA_meshdata_types.h"
+#include "DNA_mesh_types.h"
 
 #include "BKE_shrinkwrap.h"
 #include "BKE_DerivedMesh.h"
@@ -44,6 +46,7 @@
 #include "BKE_cdderivedmesh.h"
 #include "BKE_displist.h"
 #include "BKE_global.h"
+#include "BKE_subsurf.h"
 
 #include "BLI_arithb.h"
 #include "BLI_kdtree.h"
@@ -344,6 +347,7 @@ void shrinkwrap_calc_normal_projection(ShrinkwrapCalcData *calc)
 	const char use_normal    = calc->smd->shrinkOpts;
 	float proj_axis[3] = {0.0f, 0.0f, 0.0f};
 	MVert *vert  = NULL; //Needed in case of vertex normal
+	DerivedMesh* ss_mesh = NULL;
 
 	//Vertex group data
 	const int vgroup		   = get_named_vertexgroup_num(calc->ob, calc->smd->vgroup_name);
@@ -359,11 +363,46 @@ void shrinkwrap_calc_normal_projection(ShrinkwrapCalcData *calc)
 	BVHTreeFromMesh auxData= NULL_BVHTreeFromMesh;
 	SpaceTransform local2aux;
 
+do
+{
 
 	//Prepare data to retrieve the direction in which we should project each vertex
 	if(calc->smd->projAxis == MOD_SHRINKWRAP_PROJECT_OVER_NORMAL)
 	{
-		vert = calc->original ? calc->original->getVertDataArray(calc->original, CD_MVERT) : NULL;
+		//No Mvert information: jump to "free memory and return" part
+		if(calc->original == NULL) break;
+
+		if(calc->smd->subsurfLevels)
+		{
+			SubsurfModifierData smd;
+			memset(&smd, 0, sizeof(smd));
+			smd.subdivType = ME_CC_SUBSURF;			//catmull clark
+			smd.levels = calc->smd->subsurfLevels;	//levels
+
+			ss_mesh = subsurf_make_derived_from_derived(calc->original, &smd, FALSE, NULL, 0, 0);
+
+			if(ss_mesh)
+			{
+				vert = ss_mesh->getVertDataArray(ss_mesh, CD_MVERT);
+				if(vert)
+				{
+					//TRICKY: this code assumes subsurface will have the transformed original vertices
+					//in their original order at the end of the vert array.
+					vert = vert
+						 + ss_mesh->getNumVerts(ss_mesh)
+						 - calc->original->getNumVerts(calc->original);
+				}
+			}
+
+			//To make sure we are not letting any memory behind
+			assert(smd.emCache == NULL);
+			assert(smd.mCache == NULL);
+		}
+		else
+			vert = calc->original->getVertDataArray(calc->original, CD_MVERT);
+
+		//Not able to get vert information: jump to "free memory and return" part
+		if(vert == NULL) break;
 	}
 	else
 	{
@@ -373,22 +412,21 @@ void shrinkwrap_calc_normal_projection(ShrinkwrapCalcData *calc)
 		if(calc->smd->projAxis & MOD_SHRINKWRAP_PROJECT_OVER_Z_AXIS) proj_axis[2] = 1.0f;
 
 		Normalize(proj_axis);
-	}
 
-	if(vert == NULL && (INPR(proj_axis, proj_axis) < FLT_EPSILON))
-	{
-		printf("Shrinkwrap can't project witouth normal information");
-		return;
+		//Invalid projection direction: jump to "free memory and return" part
+		if(INPR(proj_axis, proj_axis) < FLT_EPSILON) break; 
 	}
 
 	//If the user doesn't allows to project in any direction of projection axis... then theres nothing todo.
 	if((use_normal & (MOD_SHRINKWRAP_PROJECT_ALLOW_POS_DIR | MOD_SHRINKWRAP_PROJECT_ALLOW_NEG_DIR)) == 0)
-		return;
+		break; //jump to "free memory and return" part
 
 
 	//Build target tree
 	BENCH(bvhtree_from_mesh_faces(&treeData, calc->target, calc->keepDist, 4, 6));
-	if(treeData.tree == NULL) return OUT_OF_MEMORY();
+	if(treeData.tree == NULL)
+		break; //jump to "free memory and return" part
+
 
 	//Build auxiliar target
 	if(calc->smd->auxTarget)
@@ -414,7 +452,15 @@ void shrinkwrap_calc_normal_projection(ShrinkwrapCalcData *calc)
 
 		if(weight == 0.0f) continue;
 
-		VECCOPY(tmp_co, co);
+		if(ss_mesh)
+		{
+			VECCOPY(tmp_co, vert[i].co);
+		}
+		else
+		{
+			VECCOPY(tmp_co, co);
+		}
+
 
 		if(vert)
 			NormalShortToFloat(tmp_no, vert[i].no);
@@ -456,6 +502,9 @@ void shrinkwrap_calc_normal_projection(ShrinkwrapCalcData *calc)
 	}
 
 
+//Simple do{} while(0) structure to allow to easily jump to the "free memory and return" part
+} while(0);
+
 	//free data structures
 
 	free_bvhtree_from_mesh(&treeData);
@@ -463,6 +512,9 @@ void shrinkwrap_calc_normal_projection(ShrinkwrapCalcData *calc)
 
 	if(aux_mesh)
 		aux_mesh->release(aux_mesh);
+
+	if(ss_mesh)
+		ss_mesh->release(ss_mesh);
 }
 
 /*
@@ -492,7 +544,7 @@ void shrinkwrap_calc_nearest_surface_point(ShrinkwrapCalcData *calc)
 	nearest.dist = FLT_MAX;
 
 
-	//Find the nearest vertex 
+	//Find the nearest vertex
 #pragma omp parallel for default(none) private(i) firstprivate(nearest) shared(calc,treeData) schedule(static)
 	for(i = 0; i<calc->numVerts; ++i)
 	{
@@ -518,7 +570,7 @@ void shrinkwrap_calc_nearest_surface_point(ShrinkwrapCalcData *calc)
 		BLI_bvhtree_find_nearest(treeData.tree, tmp_co, &nearest, treeData.nearest_callback, &treeData);
 
 		//Found the nearest vertex
-		if(nearest.index)
+		if(nearest.index != -1)
 		{
 			if(calc->smd->shrinkOpts & MOD_SHRINKWRAP_KEEP_ABOVE_SURFACE)
 			{