simple_mesh.cxx

#include "simple_mesh.h"
#include "stl_reader.h"
#include "obj_loader.h"
#include <cgv/math/inv.h>
#include <cgv/utils/scan.h>
#include <cgv/utils/advanced_scan.h>
#include <cgv/media/mesh/obj_reader.h>
#include <cgv/math/bucket_sort.h>
#include <fstream>
 
namespace cgv {
    namespace media {
        namespace mesh {
 
std::string simple_mesh_base::get_attribute_name(attribute_type attr)
{
    const char* attribute_names[] = { "position", "texcoords", "normal", "tangent", "color" };
    return attribute_names[int(attr)];
}
simple_mesh_base::AttributeFlags simple_mesh_base::get_attribute_flag(attribute_type attr)
{
    AttributeFlags attribute_flags[] = { AF_position, AF_texcoords, AF_normal, AF_tangent, AF_color };
    return attribute_flags[int(attr)];
}
simple_mesh_base::simple_mesh_base() 
{
}
simple_mesh_base::simple_mesh_base(const simple_mesh_base& smb) :
    colored_model(smb),
    position_indices(smb.position_indices),
    normal_indices(smb.normal_indices),
    tex_coord_indices(smb.tex_coord_indices),
    faces(smb.faces),
    group_indices(smb.group_indices),
    group_names(smb.group_names),
    material_indices(smb.material_indices),
    materials(smb.materials)
{
}
simple_mesh_base::simple_mesh_base(simple_mesh_base&& smb) :
    colored_model(std::move(smb)),
    position_indices(std::move(smb.position_indices)),
    normal_indices(std::move(smb.normal_indices)),
    tex_coord_indices(std::move(smb.tex_coord_indices)),
    faces(std::move(smb.faces)),
    group_indices(std::move(smb.group_indices)),
    group_names(std::move(smb.group_names)),
    material_indices(std::move(smb.material_indices)),
    materials(std::move(smb.materials))
{
}
simple_mesh_base& simple_mesh_base::operator=(const simple_mesh_base& smb)
{
    colored_model::operator=(smb);
    position_indices=smb.position_indices;
    normal_indices=smb.normal_indices;
    tex_coord_indices=smb.tex_coord_indices;
    faces=smb.faces;
    group_indices=smb.group_indices;
    group_names=smb.group_names;
    material_indices=smb.material_indices;
    materials = smb.materials;
    return *this;
}
simple_mesh_base& simple_mesh_base::operator=(simple_mesh_base&& smb)
{
    colored_model::operator=(std::move(smb));
    position_indices=std::move(smb.position_indices);
    normal_indices=std::move(smb.normal_indices);
    tex_coord_indices=std::move(smb.tex_coord_indices);
    faces=std::move(smb.faces);
    group_indices=std::move(smb.group_indices);
    group_names=std::move(smb.group_names);
    material_indices=std::move(smb.material_indices);
    materials = std::move(smb.materials);
    return *this;
}
simple_mesh_base::idx_type simple_mesh_base::start_face()
{
    faces.push_back((cgv::type::uint32_type)position_indices.size());
    if (!materials.empty())
        material_indices.push_back(idx_type(materials.size()) - 1);
    if (!group_names.empty())
        group_indices.push_back(idx_type(group_names.size()) - 1);
    return idx_type(faces.size() - 1);
}
simple_mesh_base::idx_type simple_mesh_base::new_corner(idx_type position_index, idx_type normal_index,
                                                        idx_type tex_coord_index)
{
    position_indices.push_back(position_index);
    if (normal_index != -1) //FIXME: -1 underflows unsigned int!
        normal_indices.push_back(normal_index);
    if (tex_coord_index != -1) //FIXME: -1 underflows unsigned int!
        tex_coord_indices.push_back(tex_coord_index);
    return idx_type(position_indices.size());
}
void simple_mesh_base::revert_face_orientation()
{
    bool nmls = position_indices.size() == normal_indices.size();
    bool tcs  = position_indices.size() == tex_coord_indices.size();
    for (idx_type fi = 0; fi < get_nr_faces(); ++fi) {
        idx_type ci = begin_corner(fi);
        idx_type cj = end_corner(fi);
        while (ci + 1 < cj) {
            --cj;
            std::swap(position_indices[ci], position_indices[cj]);
            if (nmls)
                std::swap(normal_indices[ci], normal_indices[cj]);
            if (tcs)
                std::swap(tex_coord_indices[ci], tex_coord_indices[cj]);
            ++ci;
        }
    }
}
void simple_mesh_base::sort_faces(std::vector<idx_type>& perm, bool by_group, bool by_material) const
{
    if (by_group && by_material) {
        std::vector<idx_type> perm0;
        cgv::math::bucket_sort(group_indices, get_nr_groups(), perm0);
        cgv::math::bucket_sort(material_indices, get_nr_materials(), perm, &perm0);
    }
    else if (by_group)
        cgv::math::bucket_sort(group_indices, get_nr_groups(), perm);
    else
        cgv::math::bucket_sort(material_indices, get_nr_materials(), perm);
}
void simple_mesh_base::merge_indices(std::vector<idx_type>& indices, std::vector<idx4_type>& unique_quadruples, bool* include_tex_coords_ptr, bool* include_normals_ptr, bool* include_tangents_ptr) const
{
    bool include_tex_coords = false;
    if (include_tex_coords_ptr)
        *include_tex_coords_ptr = include_tex_coords = (tex_coord_indices.size() > 0) && *include_tex_coords_ptr;
 
    bool include_normals    = false;
    if (include_normals_ptr)
        *include_normals_ptr = include_normals = (normal_indices.size() > 0) && *include_normals_ptr;
 
    bool include_tangents   = false;
    if(include_tangents_ptr)
        *include_tangents_ptr = include_tangents = (tangent_indices.size() > 0) && *include_tangents_ptr;
 
    std::map<std::tuple<idx_type, idx_type, idx_type, idx_type>, idx_type> corner_to_index;
    for (idx_type ci = 0; ci < position_indices.size(); ++ci) {
        // construct corner
        idx4_type c(position_indices[ci], 
                (include_tex_coords && ci < tex_coord_indices.size()) ? tex_coord_indices[ci] : 0, 
                (include_normals && ci < normal_indices.size()) ? normal_indices[ci] : 0,
                (include_tangents && ci < tangent_indices.size()) ? tangent_indices[ci] : 0);
        std::tuple<idx_type, idx_type, idx_type, idx_type> quadruple(c(0),c(1),c(2),c(3));
        // look corner up in map
        auto iter = corner_to_index.find(quadruple);
        // determine vertex index
        idx_type vi;
        if (iter == corner_to_index.end()) {
            vi = idx_type(unique_quadruples.size());
            corner_to_index[quadruple] = vi;
            unique_quadruples.push_back(c);
        }
        else
            vi = iter->second;
 
        indices.push_back(vi);
    }
}
void simple_mesh_base::extract_triangle_element_buffer(
    const std::vector<idx_type>& vertex_indices, std::vector<idx_type>& triangle_element_buffer, 
    const std::vector<idx_type>* face_permutation_ptr, std::vector<idx3_type>* material_group_start_ptr) const
{
    idx_type mi = idx_type(-1);
    idx_type gi = idx_type(-1);
    // construct triangle element buffer
    for (idx_type fi = 0; fi < faces.size(); ++fi) {
        idx_type fj = face_permutation_ptr ? face_permutation_ptr->at(fi) : fi;
        if (material_group_start_ptr) {
            if (mi != material_indices[fj] || gi != group_indices[fj]) {
                mi = material_indices[fj];
                gi = group_indices[fj];
                material_group_start_ptr->push_back(idx3_type(mi, gi, idx_type(triangle_element_buffer.size())));
            }
        }
        if (face_degree(fj) == 3) {
            for (idx_type ci = begin_corner(fj); ci < end_corner(fj); ++ci)
                triangle_element_buffer.push_back(vertex_indices.at(ci));
        }
        else {
            // in case of non triangular faces do simplest triangulation approach that assumes convexity of faces
            for (idx_type ci = begin_corner(fj) + 2; ci < end_corner(fj); ++ci) {
                triangle_element_buffer.push_back(vertex_indices.at(begin_corner(fj)));
                triangle_element_buffer.push_back(vertex_indices.at(ci - 1));
                triangle_element_buffer.push_back(vertex_indices.at(ci));
            }
        }
    }
}
void simple_mesh_base::extract_wireframe_element_buffer(const std::vector<idx_type>& vertex_indices, std::vector<idx_type>& edge_element_buffer) const
{
    // map stores for each halfedge the number of times it has been seen before
    std::map<std::tuple<idx_type, idx_type>, idx_type> halfedge_to_count;
    for (idx_type fi = 0; fi < faces.size(); ++fi) {
        idx_type last_vi = vertex_indices.at(end_corner(fi) - 1);
        for (idx_type ci = begin_corner(fi); ci < end_corner(fi); ++ci) {
            // construct halfedge with sorted vertex indices
            idx_type vi = vertex_indices.at(ci);
            std::tuple<idx_type, idx_type> halfedge(last_vi, vi);
            if (vi < last_vi)
                std::swap(std::get<0>(halfedge), std::get<1>(halfedge));
 
            // lookup corner in map
            auto iter = halfedge_to_count.find(halfedge);
 
            // determine vertex index
            if (iter == halfedge_to_count.end()) {
                halfedge_to_count[halfedge] = 1;
                edge_element_buffer.push_back(last_vi);
                edge_element_buffer.push_back(vi);
            }
            else
                ++halfedge_to_count[halfedge];
            last_vi = vi;
        }
    }
}
simple_mesh_base::idx_type simple_mesh_base::extract_vertex_attribute_buffer_base(const std::vector<idx4_type>& unique_quadruples, AttributeFlags& flags, std::vector<uint8_t>& attrib_buffer) const
{
    // update flags of to be used attributes
    if (position_indices.empty() && (flags & AF_position))
        flags = AttributeFlags(flags & ~AF_position);
    if (tex_coord_indices.empty() && (flags & AF_texcoords))
        flags = AttributeFlags(flags & ~AF_texcoords);
    if (normal_indices.empty() && (flags & AF_normal))
        flags = AttributeFlags(flags & ~AF_normal);
    if (tangent_indices.empty() && (flags & AF_tangent))
        flags = AttributeFlags(flags & ~AF_tangent);
    if (!has_colors() && (flags & AF_color))
        flags = AttributeFlags(flags & ~AF_color);
    bool include_attribute[5] = { bool(flags&AF_position),bool(flags&AF_texcoords), 
        bool(flags&AF_normal),bool(flags&AF_tangent),bool(flags&AF_color) };
    // determine vertex size in bytes and allocate attribute buffer
    uint32_t cs = get_coord_size();
    uint32_t attribute_size[5] = { 3 * cs,2 * cs,3 * cs,3 * cs,uint32_t(get_color_size()) };
    uint32_t attribute_offset[5] = { 3 * cs,2 * cs,3 * cs,3 * cs,uint32_t(get_color_size() == 3 ? 4 : get_color_size()) };
    uint32_t vs = 0;
    for (int ai = 0; ai < 5; ++ai)
        if (include_attribute[ai])
            vs += attribute_offset[ai];
    attrib_buffer.resize(vs * unique_quadruples.size());
    // fill attribute buffer
    const uint8_t* attrib_ptrs[5] = {
        include_attribute[0] ? get_attribute_ptr(attribute_type::position) : nullptr,
        include_attribute[1] ? get_attribute_ptr(attribute_type::texcoords) : nullptr,
        include_attribute[2] ? get_attribute_ptr(attribute_type::normal) : nullptr,
        include_attribute[3] ? get_attribute_ptr(attribute_type::tangent) : nullptr,
        include_attribute[4] ? get_attribute_ptr(attribute_type::color) : nullptr
    };
    size_t loc = 0;
    for (auto t : unique_quadruples) {
        for (int ai = 0; ai < 5; ++ai)
            if (include_attribute[ai]) {
                const uint8_t* src_ptr = attrib_ptrs[ai] + attribute_size[ai] * t[ai & 3];
                std::copy(src_ptr, src_ptr + attribute_size[ai], &attrib_buffer[loc]);
                loc += attribute_offset[ai];
            }
    }
    return vs;
}
simple_mesh_base::idx_type simple_mesh_base::compute_inv(
    std::vector<idx_type>& inv,
    bool link_non_manifold_edges,
    std::vector<idx_type>* p2c_ptr,
    std::vector<idx_type>* next_ptr,
    std::vector<idx_type>* prev_ptr,
    std::vector<idx_type>* unmatched,
    std::vector<idx_type>* non_manifold,
    std::vector<idx_type>* unmatched_elements,
    std::vector<idx_type>* non_manifold_elements) const
{
    uint32_t fi, e = 0;
    if (p2c_ptr)
        p2c_ptr->resize(get_nr_positions());
    if (next_ptr)
        next_ptr->resize(get_nr_corners());
    if (prev_ptr)
        prev_ptr->resize(get_nr_corners());
    inv.resize(get_nr_corners(), uint32_t(-1));
    std::vector<idx_type> pis(get_nr_corners()), perm0, perm;
    // extract min position indices per corner
    //idx_type a = 0;
    //std::cout << "before sort" << std::endl;
    for (fi = 0; fi < get_nr_faces(); ++fi) {
        idx_type cb = begin_corner(fi), ce = end_corner(fi), cp = ce - 1, pp = c2p(cp);
        for (idx_type ci = cb; ci < ce; ++ci) {
            idx_type pi = c2p(ci);
            if (p2c_ptr)
                p2c_ptr->at(pi) = ci;
            if (next_ptr)
                next_ptr->at(cp) = ci;
            if (prev_ptr)
                prev_ptr->at(ci) = cp;
            pis[cp] = std::min(pp, pi);
            //if (++a < 20)
            //  std::cout << "   " << a - 1 << " : " << pp << "," << pi << std::endl;
            cp = ci;
            pp = pi;
        }
    }
    // sort corners by min position indices
    cgv::math::bucket_sort(pis, get_nr_positions(), perm0);
    //std::cout << "after min sort" << std::endl;
    //for (a=0; a < 20; ++a)
    //  std::cout << "   " << a << " : " << c2p(perm0[a]) << "," << c2p(next_ptr->at(perm0[a])) << " (" << pis[perm0[a]] << ")" << std::endl;
    // extract max position indices per corner and fill p2c, next and prev vectors
    for (fi = 0; fi < get_nr_faces(); ++fi) {
        idx_type cb = begin_corner(fi), ce = end_corner(fi), cp = ce - 1;
        for (idx_type ci = cb; ci < ce; ++ci) {
            idx_type pi = c2p(ci);
            pis[cp] = std::max(c2p(cp), pi);
            cp = ci;
        }
    }
    // sort corners by max position indices
    cgv::math::bucket_sort(pis, get_nr_positions(), perm, &perm0);
    //std::cout << "after max sort" << std::endl;
    //for (a = 0; a < 20; ++a)
    //  std::cout << "   " << a << " : " << c2p(perm[a]) << "," << c2p(next_ptr->at(perm[a])) << " (" << pis[perm[a]] << ")" << std::endl;
    // store target in pis
    for (fi = 0; fi < get_nr_faces(); ++fi) {
        idx_type cb = begin_corner(fi), ce = end_corner(fi), cp = ce - 1;
        for (idx_type ci = cb; ci < ce; ++ci) {
            pis[cp] = c2p(ci);
            cp = ci;
        }
    }
    perm0.clear();
    // finally perform matching
    idx_type i = 0;
    while (i < perm.size()) {
        idx_type ci = perm[i], pi0 = c2p(ci), pi1 = pis[ci];
        idx_type cnt = 1;
        while (i + cnt < perm.size()) {
            idx_type cj = perm[i + cnt], pj0 = c2p(cj), pj1 = pis[cj];
            if (std::min(pi0, pi1) == std::min(pj0, pj1) && std::max(pi0, pi1) == std::max(pj0, pj1)) {
                ++cnt;
            }
            else
                break;
        }
        if (cnt == 1) {
            if (unmatched)
                unmatched->push_back(ci);
            if (unmatched_elements) {
                unmatched_elements->push_back(pi0);
                unmatched_elements->push_back(pi1);
            }
        }
        else if (cnt == 2) {
            inv[perm[i]] = perm[i + 1];
            inv[perm[i + 1]] = perm[i];
            ++e;
        }
        else {
            if (link_non_manifold_edges) {
                idx_type cl = perm[i + cnt - 1];
                for (idx_type k = 0; k < cnt; ++k) {
                    idx_type ck = perm[i + k];
                    inv[cl] = ck;
                    cl = ck;
                }
            }
            if (non_manifold)
                non_manifold->push_back(ci);
            if (non_manifold_elements) {
                non_manifold_elements->push_back(pi0);
                non_manifold_elements->push_back(pi1);
            }
        }
        i += cnt;
    }
    return e;
    /*
    std::map<std::pair<uint32_t, uint32_t>, uint32_t> pipj2ci;
    for (fi = 0; fi < get_nr_faces(); ++fi) {
        uint32_t prev_ci = end_corner(fi) - 1;
        for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci) {
            uint32_t pi = c2p(ci);
            if (p2c_ptr)
                p2c_ptr->at(pi) = ci;
            uint32_t next_ci = ci + 1 == end_corner(fi) ? begin_corner(fi) : ci + 1;
            if (next_ptr)
                next_ptr->at(ci) = next_ci;
            if (prev_ptr)
                prev_ptr->at(ci) = prev_ci;
            prev_ci = ci;
            uint32_t pj = c2p(next_ci);
            std::pair<uint32_t, uint32_t> pipj(std::min(pi, pj), std::max(pi, pj));
            if (pipj2ci.find(pipj) == pipj2ci.end())
                pipj2ci[pipj] = ci;
            else {
                uint32_t inv_ci = pipj2ci[pipj];
                inv[ci] = inv_ci;
                inv[inv_ci] = ci;
                pipj2ci.erase(pipj2ci.find(pipj));
            }
        }
    }
    */
}
simple_mesh_base::idx_type simple_mesh_base::compute_c2e(const std::vector<uint32_t>& inv, std::vector<uint32_t>& c2e, std::vector<uint32_t>* e2c_ptr) const
{
    uint32_t e = 0;
    c2e.resize(get_nr_corners(), -1);
    if (e2c_ptr)
        e2c_ptr->resize(get_nr_corners() / 2);
    for (uint32_t fi = 0; fi < get_nr_faces(); ++fi) {
        for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci) {
            if (c2e[ci] == -1) {
                c2e[ci] = c2e[inv[ci]] = e;
                if (e2c_ptr)
                    e2c_ptr->at(e) = ci;
                ++e;
            }
        }
    }
    return e;
}
void simple_mesh_base::compute_c2f(std::vector<uint32_t>& c2f) const
{
    c2f.resize(get_nr_corners(), -1);
    for (uint32_t fi = 0; fi < get_nr_faces(); ++fi) {
        for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci)
            c2f[ci] = fi;
    }
}
template <typename T>
void simple_mesh<T>::construct(const obj_loader_generic<T>& loader, bool copy_grp_info, bool copy_material_info)
{
    for (unsigned vi = 0; vi < loader.vertices.size(); ++vi)
        new_position(loader.vertices[vi]);
    for (unsigned ni = 0; ni < loader.normals.size(); ++ni)
        new_normal(loader.normals[ni]);
    for (unsigned ti = 0; ti < loader.texcoords.size(); ++ti)
        new_tex_coord(loader.texcoords[ti]);
    if (copy_grp_info)
        for (unsigned gi = 0; gi < loader.groups.size(); ++gi)
            new_group(loader.groups[gi].name);
    if (copy_material_info)
        for (unsigned mi = 0; mi < loader.materials.size(); ++mi)
            ref_material(new_material()) = loader.materials[mi];    
    for (unsigned fi = 0; fi < loader.faces.size(); ++fi) {
        start_face();
        const auto& F = loader.faces[fi];
        if (copy_grp_info && !loader.groups.empty())
            group_index(fi) = F.group_index;
        if (copy_material_info && !loader.materials.empty())
            material_index(fi) = F.material_index;
        for (unsigned i = 0; i < F.degree; ++i) {
            new_corner(
                loader.vertex_indices[F.first_vertex_index + i], 
                F.first_normal_index != -1 ? loader.normal_indices[F.first_normal_index + i] : -1, 
                F.first_texcoord_index != -1 ? loader.texcoord_indices[F.first_texcoord_index + i] : -1
            );
        }
    }
}
 
template <typename T>
class simple_mesh_obj_reader : public obj_reader_generic<T>
{
public:
    typedef T coord_type;
    typedef cgv::math::fvec<T,2> vec2_type;
    typedef cgv::math::fvec<T,3> vec3_type;
    typedef illum::obj_material::color_type color_type;
 
    typedef typename simple_mesh<T>::idx_type idx_type;
protected:
    simple_mesh<T> &mesh;
public:
    simple_mesh_obj_reader(simple_mesh<T>& _mesh) : mesh(_mesh) {}
    void process_vertex(const vec3_type& p) { mesh.positions.push_back(p); }
    void process_texcoord(const vec2_type& t) { mesh.tex_coords.push_back(vec2_type(t(0),t(1))); }
    void process_color(const color_type& c) { mesh.resize_colors(mesh.get_nr_colors() + 1); mesh.set_color(mesh.get_nr_colors()-1, c); }
    void process_normal(const vec3_type& n) { mesh.normals.push_back(n); }
    void process_face(unsigned vcount, int *vertices, int *texcoords, int *normals)
    {
        obj_reader_base::convert_to_positive(vcount, vertices, texcoords, normals, unsigned(mesh.positions.size()), unsigned(mesh.normals.size()), unsigned(mesh.tex_coords.size()));
        mesh.faces.push_back(idx_type(mesh.position_indices.size()));
        if (obj_reader_base::get_current_group() != -1)
            mesh.group_indices.push_back(obj_reader_base::get_current_group());
        if (obj_reader_base::get_current_material() != -1)
            mesh.material_indices.push_back(obj_reader_base::get_current_material());
        if (texcoords) {
            if (mesh.tex_coord_indices.size() < mesh.position_indices.size())
                mesh.tex_coord_indices.resize(mesh.position_indices.size(), 0);
        }
        if (normals) {
            if (mesh.normal_indices.size() < mesh.position_indices.size())
                mesh.normal_indices.resize(mesh.position_indices.size(), 0);
        }
        for (idx_type i = 0; i < vcount; ++i) {
            mesh.position_indices.push_back(idx_type(vertices[i]));
            if (texcoords)
                mesh.tex_coord_indices.push_back(idx_type(texcoords[i]));
            if (normals)
                mesh.normal_indices.push_back(idx_type(normals[i]));
        }
    }
    void process_group(const std::string& name, const std::string& parameters)
    {
        mesh.group_names.push_back(name);
    }
    void process_material(const cgv::media::illum::obj_material& mtl, unsigned idx)
    {
        if (idx >= mesh.materials.size())
            mesh.materials.resize(idx+1);
        mesh.materials[idx] = mtl;
    }
};
 
 
template <typename T>
simple_mesh<T>::simple_mesh(const simple_mesh<T>& sm)
    : simple_mesh_base(sm), positions(sm.positions), normals(sm.normals), tex_coords(sm.tex_coords)
{
}
 
template <typename T>
simple_mesh<T>::simple_mesh(simple_mesh<T>&& sm)
    : simple_mesh_base(std::move(sm)), positions(std::move(sm.positions)), normals(std::move(sm.normals)),
      tex_coords(std::move(sm.tex_coords))
{
}
 
template <typename T>
simple_mesh<T>& simple_mesh<T>::operator = (const simple_mesh<T>& sm)
{
    simple_mesh_base::operator = (sm);
    positions = sm.positions;
    normals = sm.normals;
    tangents = sm.tangents;
    tex_coords = sm.tex_coords;
    return *this;
}
 
template <typename T>
simple_mesh<T>& simple_mesh<T>::operator = (simple_mesh<T>&& sm)
{
    simple_mesh_base::operator = (std::move(sm));
    positions = std::move(sm.positions);
    normals = std::move(sm.normals);
    tangents = std::move(sm.tangents);
    tex_coords = std::move(sm.tex_coords);
    return *this;
}
 
template <typename T>
void simple_mesh<T>::clear() 
{
    positions.clear(); 
    normals.clear(); 
    tangents.clear();
    tex_coords.clear(); 
    position_indices.clear(); 
    tex_coord_indices.clear(); 
    normal_indices.clear(); 
    faces.clear();
    group_indices.clear();
    group_names.clear();
    material_indices.clear();
    materials.clear();
    destruct_colors();
}
 
template <typename T>
bool read_off(const std::string& file_name, 
    std::vector<cgv::math::fvec<T,3>>& positions, std::vector<cgv::rgba>& vertex_colors, 
    std::vector<std::vector<uint32_t>>& faces, std::vector<cgv::rgba>& face_colors)
{
    std::string content;
    if (!cgv::utils::file::read(file_name, content, true))
        return false;
    std::vector<cgv::utils::line> lines;
    cgv::utils::split_to_lines(content, lines);
    if (!(lines[0] == "OFF")) {
        std::cerr << "WARNING: first line in OFF file " << file_name << " does not contain 'OFF'" << std::endl;
        return false;
    }
    unsigned real_li = 1;
    int v, f, e;
    for (unsigned li = 1; li < lines.size(); ++li) {
        if (lines[li].empty())
            continue;
        if (lines[li].begin[0] == '#')
            continue;
        ++real_li;
        std::vector<cgv::utils::token> toks;
        cgv::utils::split_to_tokens(lines[li], toks, "");
        if (real_li == 2) {
            if (toks.size() != 3) {
                std::cerr << "WARNING: second line in OFF file " << file_name << " does provide 3 tokens" << std::endl;
                return false;
            }
            int I[3];
            for (int i = 0; i < 3; ++i) {
                if (!cgv::utils::is_integer(toks[i].begin, toks[i].end, I[i])) {
                    std::cerr << "WARNING: token " << i << " on second line in OFF file " << file_name << " is not an integer value" << std::endl;
                    return false;
                }
            }
            v = I[0]; f = I[1]; e = I[2];
            std::cout << "OFF file " << file_name << " found " << v << " vertices, " << f << " faces, and " << e << " edges." << std::endl;
            continue;
        }
        if (int(real_li) < v+3) {
            if (!(toks.size() == 3 || toks.size() == 6 || toks.size() == 7)) {
                std::cerr << "WARNING: line of vertex " << real_li - 3 << " contains " << toks.size() << " tokens instead of 3 or 7." << std::endl;
                return false;
            }
            double x[3];
            for (unsigned i = 0; i < 3; ++i) {
                if (!cgv::utils::is_double(toks[i].begin, toks[i].end, x[i])) {
                    std::cerr << "WARNING: line of vertex " << real_li - 3 << " no double in XYZ component " << i << " but <" << toks[i] << ">." << std::endl;
                    return false;
                }
            }
            positions.push_back(cgv::math::fvec<T, 3>(T(x[0]), T(x[1]), T(x[2])));
            if (toks.size() >= 6) {
                double c[4] = { 0,0,0,1 };
                for (unsigned i = 0; i+3 < toks.size(); ++i) {
                    if (!cgv::utils::is_double(toks[i+3].begin, toks[i+3].end, c[i])) {
                        std::cerr << "WARNING: line of vertex " << real_li - 3 << " no double in RGB[A] component " << i << " but <" << toks[i+3] << ">." << std::endl;
                        return false;
                    }
                }
                while (vertex_colors.size() + 1 < positions.size())
                    vertex_colors.push_back(vertex_colors.empty() ? cgv::rgba(0.5, 0.5, 0.5, 1.0f) : vertex_colors.back());
                vertex_colors.push_back(cgv::rgba(float(c[0]), float(c[1]), float(c[2]), float(1.0-c[3])));
            }
        }
        else {
            int n;
            if (!cgv::utils::is_integer(toks[0].begin, toks[0].end, n)) {
                std::cerr << "WARNING: first token on face " << faces.size() << " is not of type integer " << std::endl;
                return false;
            }
            if (!(toks.size() == n + 1 || toks.size() == n + 4 || toks.size() == n + 5)) {
                std::cerr << "WARNING: line of face " << faces.size() << " contains " << toks.size() << " tokens instead of " << n+1 << " or " << n+5 << std::endl;
                return false;
            }
            faces.push_back({});
            auto& face = faces.back();
            for (int i = 0; i<n; ++i) {
                int pi;
                if (!cgv::utils::is_integer(toks[i+1].begin, toks[i + 1].end, pi)) {
                    std::cerr << "WARNING: token " << i+1 << " on face " << faces.size()-1 << " is not of type integer " << std::endl;
                    return false;
                }
                face.push_back(uint32_t(pi));
            }
            if (toks.size() >= n + 4) {
                double c[4] = { 0,0,0,1 };
                for (unsigned i = 0; i < toks.size()-n-1; ++i) {
                    if (!cgv::utils::is_double(toks[i + n + 1].begin, toks[i + n + 1].end, c[i])) {
                        std::cerr << "WARNING: line of vertex " << real_li - 3 << " no double in RGBA component " << i << " but <" << toks[i + n + 1] << ">." << std::endl;
                        return false;
                    }
                }
                while (face_colors.size()+1 < faces.size())
                    face_colors.push_back(face_colors.empty() ? cgv::rgba(0.5, 0.5, 0.5, 1.0f) : face_colors.back());
                face_colors.push_back(cgv::rgba(float(c[0]), float(c[1]), float(c[2]), float(1.0 - c[3])));
            }
        }
    }
    return true;
}
 
 
template <typename T>
bool simple_mesh<T>::read(const std::string& file_name)
{ 
    std::string ext = cgv::utils::to_lower(cgv::utils::file::get_extension(file_name));
    if (ext == "obj") {
        simple_mesh_obj_reader<T> reader(*this);
        return reader.read_obj(file_name);
    }
    if (ext == "stl") {
        try {
            stl_reader::StlMesh <T, unsigned> mesh(file_name);
 
            // copy vertices
            for (size_t vi = 0; vi < mesh.num_vrts(); ++vi)
                new_position(cgv::math::fvec<T, 3>(3, mesh.vrt_coords(vi)));
 
            // copy triangles and normals
            bool has_normals = mesh.raw_normals();
            for (size_t ti = 0; ti < mesh.num_tris(); ++ti) {
                if (has_normals)
                    new_normal(cgv::math::fvec<T, 3>(3, mesh.tri_normal(ti)));
                start_face();
                for (size_t ci = 0; ci < 3; ++ci)
                    new_corner(mesh.tri_corner_ind(ti, ci), has_normals ? (unsigned)ti : -1);
            }
            return true;
        }
        catch (std::exception& e) {
            std::cout << e.what() << std::endl;
            return false;
        }
    }
    if (ext == "off") {
        ensure_colors(cgv::media::ColorType::CT_RGBA);
        auto& vertex_colors = *reinterpret_cast<std::vector<cgv::rgba>*>(ref_color_data_vector_ptr());
        std::vector<cgv::rgba> face_colors;
        std::vector<std::vector<idx_type>> faces;
        if (!read_off(file_name, ref_positions(), vertex_colors, faces, face_colors))
            return false;
        for (const auto& f : faces) {
            start_face();
            for (auto pi : f)
                new_corner(pi);
        }
        return true;
    }
    std::cerr << "unknown mesh file extension '*." << ext << "'" << std::endl;
    return false;
}
 
template <typename T>
bool simple_mesh<T>::write(const std::string& file_name) const
{
    std::ofstream os(file_name);
    if (os.fail())
        return false;
    for (const auto& p : positions)
        os << "v " << p << std::endl;
    for (const auto& t : tex_coords)
        os << "vt " << t << std::endl;
    for (const auto& n : normals)
        os << "vn " << n << std::endl;
 
    bool nmls = position_indices.size() == normal_indices.size();
    bool tcs = position_indices.size() == tex_coord_indices.size();
 
    for (idx_type fi = 0; fi < faces.size(); ++fi) {
        os << "f";
        for (idx_type ci = begin_corner(fi); ci < end_corner(fi); ++ci) {
            os << " " << position_indices[ci] + 1;
            if (!nmls && !tcs)
                continue;
            os << "/";
            if (tcs)
                os << tex_coord_indices[ci] + 1;
            if (nmls)
                os << "/" << normal_indices[ci] + 1;
        }
        os << "\n";
    }
    return true;
}
 
template <typename T>
typename simple_mesh<T>::box_type simple_mesh<T>::compute_box() const
{
    box_type box;
    for (const auto& p : positions)
        box.add_point(p);
    return box;
}
template <typename T>
void simple_mesh<T>::compute_vertex_normals(bool use_parallel_implementation)
{
    // clear previous normal info
    if (has_normals())
        normals.clear();
    // copy position indices to normals
    normal_indices = position_indices;
    // initialize normals to null vectors
    normals.resize(positions.size(), vec3_type(T(0)));
    if (use_parallel_implementation) {
#pragma omp parallel for
        for (int fi = 0; fi < int(get_nr_faces()); ++fi) {
            vec3_type nml;
            if (compute_face_normal(fi, nml))
                for (idx_type ci = begin_corner(fi); ci < end_corner(fi); ++ci)
                    normal(normal_indices[ci]) += nml;
        }
#pragma omp parallel for
        for (int ni = 0; ni < int(normals.size()); ++ni)
            normals[ni].normalize();
    }
    else {
        vec3_type nml;
        for (idx_type fi = 0; fi < get_nr_faces(); ++fi)
            if (compute_face_normal(fi, nml))
                for (idx_type ci = begin_corner(fi); ci < end_corner(fi); ++ci)
                    normal(normal_indices[ci]) += nml;
        for (auto& n : normals)
            n.normalize();
    }
}
 
template <typename T>
unsigned simple_mesh<T>::extract_vertex_attribute_buffer(const std::vector<idx4_type>& unique_quadruples,
                                                         bool include_tex_coords, bool include_normals,
                                                         bool include_tangents, std::vector<T>& attrib_buffer,
                                                         bool* include_colors_ptr, int* num_floats_in_vertex) const
{
    // correct inquiry in case data is missing
    include_tex_coords = include_tex_coords && !tex_coord_indices.empty() && !tex_coords.empty();
    include_normals = include_normals && !normal_indices.empty() && !normals.empty();
    include_tangents = include_tangents && !tangent_indices.empty() && !tangents.empty();
    bool include_colors = false;
    if (include_colors_ptr)
        *include_colors_ptr = include_colors = has_colors() && get_nr_colors() > 0 && *include_colors_ptr;
 
    // determine number floats per vertex
    unsigned nr_floats = 3;
    nr_floats += include_tex_coords ? 2 : 0;
    nr_floats += include_normals ? 3 : 0;
    nr_floats += include_tangents ? 3 : 0;
    unsigned color_increment = 0;
    if (include_colors) {
        color_increment = (int)ceil((float)get_color_size() / sizeof(T));
        nr_floats += color_increment;
    }
 
    if (num_floats_in_vertex)
        *num_floats_in_vertex = nr_floats;
 
    attrib_buffer.resize(nr_floats * unique_quadruples.size());
    T* data_ptr = &attrib_buffer.front();
    for (auto t : unique_quadruples) {
        *reinterpret_cast<vec3_type*>(data_ptr) = positions[t[0]];
        data_ptr += 3;
        if (include_tex_coords) {
            *reinterpret_cast<vec2_type*>(data_ptr) = tex_coords[t[1]];
            data_ptr += 2;
        }
        if (include_normals) {
            *reinterpret_cast<vec3_type*>(data_ptr) = normals[t[2]];
            data_ptr += 3;
        }
        if (include_tangents) {
            *reinterpret_cast<vec3_type*>(data_ptr) = tangents[t[3]];
            data_ptr += 3;
        }
        if (include_colors) {
            put_color(t[0], data_ptr);
            data_ptr += color_increment;
        }
    }
    return color_increment;
}
 
template <typename T> void simple_mesh<T>::transform(const mat3_type& linear_transformation, const vec3_type& translation)
{
    mat3_type inverse_linear_transform = inv(linear_transformation);
    transform(linear_transformation, translation, inverse_linear_transform);
}
template <typename T> void simple_mesh<T>::transform(const mat3_type& linear_transform, const vec3_type& translation, const mat3_type& inverse_linear_transform)
{
    for (auto& p : positions)
        p = linear_transform * p + translation;
    for (auto& n : normals)
        n = n * inverse_linear_transform;
    for(auto& t : tangents)
        t = t * inverse_linear_transform;
}
 
template <typename T> typename simple_mesh<T>::vec3_type simple_mesh<T>::compute_face_center(idx_type fi) const
{
    vec3_type ctr = vec3_type(0.0f);
    uint32_t nr = 0;
    for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci) {
        ctr += position(c2p(ci));
        ++nr;
    }
    ctr /= float(nr);
    return ctr;
}
template <typename T> bool simple_mesh<T>::compute_face_normal(idx_type fi, vec3_type& nml_out, bool normalize) const
{
    idx_type c0 = begin_corner(fi);
    idx_type ce = end_corner(fi);
    vec3_type p0 = position(position_indices[c0]);
    vec3_type dj = position(position_indices[c0 + 1]) - p0;
    vec3_type nml(0.0f);
    for (idx_type ci = c0 + 2; ci < ce; ++ci) {
        vec3_type di = position(position_indices[ci]) - p0;
        nml += cross(dj, di);
        dj = di;
    }
    if (!normalize) {
        nml_out = nml;
        return true;
    }
    T nl = nml.length();
    if (nl > T(1e-8f)) {
        nml *= T(1) / nl;
        nml_out = nml;
        return true;
    }
    return false;
}
template <typename T> void simple_mesh<T>::compute_face_normals(bool construct_normal_indices)
{
    if (construct_normal_indices)
        normal_indices.clear();
    normals.clear();
    // compute per face normals
    for (uint32_t fi = 0; fi < get_nr_faces(); ++fi) {
        vec3_type nml = vec3_type(T(1),0,0);
        compute_face_normal(fi, nml);
        uint32_t ni = new_normal(nml);
        if (construct_normal_indices) {
            for (idx_type ci = begin_corner(fi); ci < end_corner(fi); ++ci) {
                normal_indices.push_back(ni);
            }
        }
    }
}
template <typename T> typename simple_mesh<T>::vec3_type simple_mesh<T>::compute_normal(const vec3_type& p0, const vec3_type& p1, const vec3_type& p2)
{
    return normalize(cross(p1 - p0, p2 - p0));
}
template <typename T> void simple_mesh<T>::compute_face_tangents(bool construct_tangent_indices) {
    // compute per face tangents
    if(!has_tex_coords())
        return;
 
    for(uint32_t fi = 0; fi < get_nr_faces(); ++fi) {
        std::vector<vec3_type> _P;
        std::vector<vec2_type> _T;
        vec3_type ctr(0.0f);
        uint32_t ci, nr = 0;
        for(ci = begin_corner(fi); ci < end_corner(fi); ++ci) {
            _P.push_back(position(c2p(ci)));
            _T.push_back(tex_coord(c2t(ci)));
            ctr += _P.back();
            ++nr;
        }
        vec3_type tng(1.0f, 0.0f, 0.0f);
        // calculate tangents for faces with at least three corners
        // for more than 3 corners only use the first two edges and assume the face to be planar
        if(_P.size() > 2) {
            vec3_type edge0 = _P[1] - _P[0];
            vec3_type edge1 = _P[2] - _P[0];
            vec2_type delta_uv0 = _T[1] - _T[0];
            vec2_type delta_uv1 = _T[2] - _T[0];
 
            float dir_correction = (delta_uv1.x() * delta_uv0.y() - delta_uv1.y() * delta_uv0.x()) < 0.0f ? -1.0f : 1.0f;
 
            // when t1, t2, t3 in same position in UV space, just use default UV direction.
            if(delta_uv0.x() * delta_uv1.y() == delta_uv0.y() * delta_uv1.x()) {
                delta_uv0.x() = 0.0f;
                delta_uv0.y() = 1.0f;
                delta_uv1.x() = 1.0f;
                delta_uv1.y() = 0.0f;
            }
 
            tng.x() = delta_uv0.y() * edge1.x() - delta_uv1.y() * edge0.x();
            tng.y() = delta_uv0.y() * edge1.y() - delta_uv1.y() * edge0.y();
            tng.z() = delta_uv0.y() * edge1.z() - delta_uv1.y() * edge0.z();
            tng *= dir_correction;
        } else {
            std::cout << "could not compute tangent for non-triangular face" << std::endl;
        }
 
        tng.normalize();
        uint32_t ti = new_tangent(tng);
        if(construct_tangent_indices) {
            for(ci = begin_corner(fi); ci < end_corner(fi); ++ci) {
                tangent_indices.push_back(ti);
            }
        }
    }
}
 
template <typename T> simple_mesh<T>::simple_mesh(const std::string& conway_notation)
{
    if (!conway_notation.empty())
        construct_conway_polyhedron(conway_notation);
}
template <typename T> void simple_mesh<T>::ambo()
{
    std::vector<uint32_t> c2e;
    std::vector<uint32_t> e2c;
    std::vector<uint32_t> inv;
    std::vector<uint32_t> prev;
    std::vector<uint32_t> p2c;
    compute_inv(inv, /*link_non_manifold_edges*/false, &p2c, 0, &prev);
    uint32_t e = compute_c2e(inv, c2e, &e2c);
    mesh_type new_M;
    // create one vertex per edge
    for (uint32_t ei = 0; ei < e; ++ei) {
        uint32_t pi = c2p(e2c[ei]);
        uint32_t pj = c2p(inv[e2c[ei]]);
        new_M.new_position(normalize(position(pi) + position(pj)));
    }
    // create one face for original faces
    for (uint32_t fi = 0; fi < get_nr_faces(); ++fi) {
        uint32_t new_fi = new_M.start_face();
        for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci)
            new_M.new_corner(c2e[ci], new_fi);
    }
    // create one face for original vertices
    for (uint32_t pi = 0; pi < get_nr_positions(); ++pi) {
        uint32_t c0 = p2c[pi];
        uint32_t ci = c0;
        uint32_t new_fi = new_M.start_face();
        do {
            new_M.new_corner(c2e[ci], new_fi);
            ci = prev[ci];
            ci = inv[ci];
        } while (ci != c0);
    }
    new_M.compute_face_normals();
    *this = new_M;
}
template <typename T> void simple_mesh<T>::truncate(T lambda)
{
    std::vector<uint32_t> inv;
    std::vector<uint32_t> prev;
    std::vector<uint32_t> p2c;
    compute_inv(inv, /*link_non_manifold_edges*/false, &p2c, 0, &prev);
    uint32_t c = get_nr_corners();
    mesh_type new_M;
    // create one vertex per corner
    for (uint32_t ci = 0; ci < c; ++ci) {
        uint32_t pi = c2p(ci);
        uint32_t pj = c2p(inv[ci]);
        new_M.new_position(normalize((1 - lambda) * position(pi) + lambda * position(pj)));
    }
    // create one face for original faces
    for (uint32_t fi = 0; fi < get_nr_faces(); ++fi) {
        uint32_t new_fi = new_M.start_face();
        for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci) {
            new_M.new_corner(ci, new_fi);
            new_M.new_corner(inv[ci], new_fi);
        }
    }
    // create one face for original vertices
    for (uint32_t pi = 0; pi < get_nr_positions(); ++pi) {
        uint32_t c0 = p2c[pi];
        uint32_t ci = c0;
        uint32_t new_fi = new_M.start_face();
        do {
            new_M.new_corner(ci, new_fi);
            ci = prev[ci];
            ci = inv[ci];
        } while (ci != c0);
    }
    new_M.compute_face_normals();
    *this = new_M;
}
template <typename T> void simple_mesh<T>::snub(T lambda)
{
    std::vector<uint32_t> inv;
    std::vector<uint32_t> prev;
    std::vector<uint32_t> p2c;
    compute_inv(inv, /*link_non_manifold_edges*/false, &p2c, 0, &prev);
    uint32_t c = get_nr_corners();
    mesh_type new_M;
    // create one vertex per corner
    for (uint32_t ci = 0; ci < c; ++ci) {
        uint32_t pi = c2p(ci);
        uint32_t pj = c2p(inv[ci]);
        new_M.new_position(normalize((1 - lambda) * position(pi) + lambda * position(pj)));
    }
    // create central face for original faces
    uint32_t fi;
    for (fi = 0; fi < get_nr_faces(); ++fi) {
        uint32_t new_fi = new_M.start_face();
        for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci)
            new_M.new_corner(ci, new_fi);
    }
    // create one per corner inside original faces
    for (fi = 0; fi < get_nr_faces(); ++fi) {
        for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci) {
            uint32_t new_fi = new_M.start_face();
            uint32_t prev_ci = prev[ci];
            new_M.new_corner(prev_ci, new_fi);
            new_M.new_corner(inv[prev_ci], new_fi);
            new_M.new_corner(ci, new_fi);
        }
    }
    // create one face for original vertices
    for (uint32_t pi = 0; pi < get_nr_positions(); ++pi) {
        uint32_t c0 = p2c[pi];
        uint32_t ci = c0;
        uint32_t new_fi = new_M.start_face();
        do {
            new_M.new_corner(ci, new_fi);
            ci = prev[ci];
            ci = inv[ci];
        } while (ci != c0);
    }
    new_M.compute_face_normals();
    *this = new_M;
}
template <typename T> void simple_mesh<T>::dual()
{
    std::vector<uint32_t> c2f;
    std::vector<uint32_t> p2c;
    std::vector<uint32_t> inv;
    std::vector<uint32_t> prev;
    compute_inv(inv, /*link_non_manifold_edges*/false, &p2c, 0, &prev);
    compute_c2f(c2f);
    uint32_t f = get_nr_faces();
    mesh_type new_M;
    // create one vertex per face
    for (uint32_t fi = 0; fi < f; ++fi) {
        vec3_type ctr = vec3_type(0.0f);
        for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci)
            ctr += position(c2p(ci));
        ctr.normalize();
        new_M.new_position(ctr);
    }
    // create one face for original vertices
    for (uint32_t pi = 0; pi < get_nr_positions(); ++pi) {
        uint32_t c0 = p2c[pi];
        uint32_t ci = c0;
        uint32_t new_fi = new_M.start_face();
        do {
            new_M.new_corner(c2f[ci], new_fi);
            ci = prev[ci];
            ci = inv[ci];
        } while (ci != c0);
    }
    new_M.compute_face_normals();
    *this = new_M;
}
template <typename T> void simple_mesh<T>::gyro(T lambda)
{
    std::vector<uint32_t> c2f;
    std::vector<uint32_t> p2c;
    std::vector<uint32_t> inv;
    std::vector<uint32_t> prev;
    compute_inv(inv, /*link_non_manifold_edges*/false, &p2c, 0, &prev);
    compute_c2f(c2f);
    uint32_t v = get_nr_positions();
    uint32_t f = get_nr_faces();
    uint32_t c = get_nr_corners();
    mesh_type new_M;
    // copy vertices
    for (uint32_t pi = 0; pi < v; ++pi)
        new_M.new_position(position(pi));
    // create one vertex per face
    for (uint32_t fi = 0; fi < f; ++fi) {
        vec3_type ctr = vec3_type(0.0f);
        for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci)
            ctr += position(c2p(ci));
        ctr.normalize();
        new_M.new_position(ctr);
    }
    // create one vertex per corner
    uint32_t ci;
    for (ci = 0; ci < c; ++ci) {
        uint32_t pi = c2p(ci);
        uint32_t pj = c2p(inv[ci]);
        new_M.new_position(normalize((1 - lambda) * position(pi) + lambda * position(pj)));
    }
    // create one face for per original corner
    for (ci = 0; ci < c; ++ci) {
        uint32_t inv_ci = inv[ci];
        uint32_t prev_inv_ci = inv[prev[ci]];
        uint32_t fi = c2f[ci];
        uint32_t pi = c2p(ci);
        uint32_t new_fi = new_M.start_face();
        new_M.new_corner(pi, new_fi);
        new_M.new_corner(ci + v + f, new_fi);
        new_M.new_corner(inv_ci + v + f, new_fi);
        new_M.new_corner(v + fi, new_fi);
        new_M.new_corner(prev_inv_ci + v + f, new_fi);
    }
    new_M.compute_face_normals();
    *this = new_M;
}
template <typename T> void simple_mesh<T>::join()
{
    std::vector<uint32_t> c2e;
    std::vector<uint32_t> e2c;
    std::vector<uint32_t> inv;
    std::vector<uint32_t> prev;
    std::vector<uint32_t> p2c;
    compute_inv(inv, /*link_non_manifold_edges*/false, &p2c, 0, &prev);
    uint32_t e = compute_c2e(inv, c2e, &e2c);
    std::vector<uint32_t> c2f;
    compute_c2f(c2f);
 
    uint32_t f = get_nr_faces();
    uint32_t v = get_nr_positions();
    mesh_type new_M;
    // copy vertices
    for (uint32_t pi = 0; pi < v; ++pi)
        new_M.new_position(position(pi));
 
    // append one vertex per face
    for (uint32_t fi = 0; fi < f; ++fi) {
        vec3_type ctr = vec3_type(0.0f);
        for (uint32_t ci = begin_corner(fi); ci < end_corner(fi); ++ci)
            ctr += position(c2p(ci));
        ctr.normalize();
        new_M.new_position(ctr);
    }
 
    // create one face per edge
    for (uint32_t ei = 0; ei < e; ++ei) {
        uint32_t ci = e2c[ei];
        uint32_t fi = c2f[ci];
        uint32_t pi = c2p(ci);
        uint32_t cj = inv[ci];
        uint32_t fj = c2f[cj];
        uint32_t pj = c2p(cj);
        new_M.start_face();
        new_M.new_corner(fi + v, ei);
        new_M.new_corner(pi, ei);
        new_M.new_corner(fj + v, ei);
        new_M.new_corner(pj, ei);
    }
    new_M.compute_face_normals();
    *this = new_M;
}
template <typename T> void simple_mesh<T>::ortho()
{
    join();
    join();
}
template <typename T> void simple_mesh<T>::construct_conway_polyhedron(const std::string& conway_notation)
{
    if (conway_notation.back() == 'C' || conway_notation.back() == 'O') {
        // cube
        static float V[8 * 3] = { -1,-1,+1, +1,-1,+1, -1,+1,+1, +1,+1,+1, -1,-1,-1, +1,-1,-1, -1,+1,-1, +1,+1,-1 };
        static float N[6 * 3] = { -1,0,0, +1,0,0, 0,-1,0, 0,+1,0, 0,0,-1, 0,0,+1 };
        static int F[6 * 4] = { 0,2,6,4, 1,5,7,3, 0,4,5,1, 2,3,7,6, 4,6,7,5, 0,1,3,2 };
        for (int vi = 0; vi < 8; ++vi)
            new_position(normalize(vec3_type(3, &V[3 * vi])));
        for (int ni = 0; ni < 6; ++ni)
            new_normal(vec3_type(3, &N[3 * ni]));
        for (int fi = 0; fi < 6; ++fi) {
            start_face();
            for (int ci = 0; ci < 4; ++ci)
                new_corner(F[4 * fi + ci], fi);
        }
        if (conway_notation.back() == 'O')
            dual();
    }
    if (conway_notation.back() == 'T' || conway_notation.back() == 'I' || conway_notation.back() == 'D') {
        static const float a = float(1.0 / (2 * sqrt(3.0)));
        static const float b = float(1.0 / (3 * sqrt(3.0 / 2)));
        static const float V[4 * 3] = { -0.5f, -a, -b,  0.5f, -a, -b,  0,2 * a, -b,  0,  0,2 * b };
        static const int F[4 * 3] = { 0,2,1,3,2,0,3,0,1,3,1,2 };
        for (int vi = 0; vi < 4; ++vi)
            new_position(normalize(vec3_type(3, &V[3 * vi])));
        for (int fi = 0; fi < 4; ++fi) {
            start_face();
            new_normal(compute_normal(position(F[3 * fi]), position(F[3 * fi + 1]), position(F[3 * fi + 2])));
            for (int ci = 0; ci < 3; ++ci)
                new_corner(F[3 * fi + ci], fi);
        }
        if (conway_notation.back() != 'T') {
            snub();
            dual();
            if (conway_notation.back() == 'I')
                dual();
        }
    }
    for (int ci = (int)conway_notation.size() - 1; ci >= 0; --ci) {
        switch (conway_notation[ci]) {
        case 't': truncate(); break;
        case 'a': ambo(); break;
        case 'd': dual(); break;
        case 'j': join(); break;
        case 'o': ortho(); break;
        case 's': snub(); break;
        case 'g': gyro(); break;
        }
    }
}
 
template class simple_mesh<float>;
template class simple_mesh<double>;
 
        }
    }
}