#include #include "3DScene.hpp" #include "libslic3r/ExtrusionEntity.hpp" #include "libslic3r/ExtrusionEntityCollection.hpp" #include "libslic3r/Geometry.hpp" #include "libslic3r/GCode/PreviewData.hpp" #include "libslic3r/Print.hpp" #include "libslic3r/SLAPrint.hpp" #include "libslic3r/Slicing.hpp" #include "libslic3r/GCode/Analyzer.hpp" #include "slic3r/GUI/PresetBundle.hpp" #include "libslic3r/Format/STL.hpp" #include "libslic3r/Utils.hpp" #include #include #include #include #include #include #include #include #include #include #include #include #include "GUI.hpp" #ifdef HAS_GLSAFE void glAssertRecentCallImpl(const char *file_name, unsigned int line, const char *function_name) { GLenum err = glGetError(); if (err == GL_NO_ERROR) return; const char *sErr = 0; switch (err) { case GL_INVALID_ENUM: sErr = "Invalid Enum"; break; case GL_INVALID_VALUE: sErr = "Invalid Value"; break; // be aware that GL_INVALID_OPERATION is generated if glGetError is executed between the execution of glBegin and the corresponding execution of glEnd case GL_INVALID_OPERATION: sErr = "Invalid Operation"; break; case GL_STACK_OVERFLOW: sErr = "Stack Overflow"; break; case GL_STACK_UNDERFLOW: sErr = "Stack Underflow"; break; case GL_OUT_OF_MEMORY: sErr = "Out Of Memory"; break; default: sErr = "Unknown"; break; } BOOST_LOG_TRIVIAL(error) << "OpenGL error in " << file_name << ":" << line << ", function " << function_name << "() : " << (int)err << " - " << sErr; assert(false); } #endif namespace Slic3r { void GLIndexedVertexArray::load_mesh_full_shading(const TriangleMesh &mesh) { assert(triangle_indices.empty() && vertices_and_normals_interleaved_size == 0); assert(quad_indices.empty() && triangle_indices_size == 0); assert(vertices_and_normals_interleaved.size() % 6 == 0 && quad_indices_size == vertices_and_normals_interleaved.size()); this->vertices_and_normals_interleaved.reserve(this->vertices_and_normals_interleaved.size() + 3 * 3 * 2 * mesh.facets_count()); unsigned int vertices_count = 0; for (int i = 0; i < (int)mesh.stl.stats.number_of_facets; ++i) { const stl_facet &facet = mesh.stl.facet_start[i]; for (int j = 0; j < 3; ++j) this->push_geometry(facet.vertex[j](0), facet.vertex[j](1), facet.vertex[j](2), facet.normal(0), facet.normal(1), facet.normal(2)); this->push_triangle(vertices_count, vertices_count + 1, vertices_count + 2); vertices_count += 3; } } void GLIndexedVertexArray::finalize_geometry(bool opengl_initialized) { assert(this->vertices_and_normals_interleaved_VBO_id == 0); assert(this->triangle_indices_VBO_id == 0); assert(this->quad_indices_VBO_id == 0); if (! opengl_initialized) { // Shrink the data vectors to conserve memory in case the data cannot be transfered to the OpenGL driver yet. this->shrink_to_fit(); return; } if (! this->vertices_and_normals_interleaved.empty()) { glsafe(::glGenBuffers(1, &this->vertices_and_normals_interleaved_VBO_id)); glsafe(::glBindBuffer(GL_ARRAY_BUFFER, this->vertices_and_normals_interleaved_VBO_id)); glsafe(::glBufferData(GL_ARRAY_BUFFER, this->vertices_and_normals_interleaved.size() * 4, this->vertices_and_normals_interleaved.data(), GL_STATIC_DRAW)); glsafe(::glBindBuffer(GL_ARRAY_BUFFER, 0)); this->vertices_and_normals_interleaved.clear(); } if (! this->triangle_indices.empty()) { glsafe(::glGenBuffers(1, &this->triangle_indices_VBO_id)); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, this->triangle_indices_VBO_id)); glsafe(::glBufferData(GL_ELEMENT_ARRAY_BUFFER, this->triangle_indices.size() * 4, this->triangle_indices.data(), GL_STATIC_DRAW)); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0)); this->triangle_indices.clear(); } if (! this->quad_indices.empty()) { glsafe(::glGenBuffers(1, &this->quad_indices_VBO_id)); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, this->quad_indices_VBO_id)); glsafe(::glBufferData(GL_ELEMENT_ARRAY_BUFFER, this->quad_indices.size() * 4, this->quad_indices.data(), GL_STATIC_DRAW)); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0)); this->quad_indices.clear(); } } void GLIndexedVertexArray::release_geometry() { if (this->vertices_and_normals_interleaved_VBO_id) { glsafe(::glDeleteBuffers(1, &this->vertices_and_normals_interleaved_VBO_id)); this->vertices_and_normals_interleaved_VBO_id = 0; } if (this->triangle_indices_VBO_id) { glsafe(::glDeleteBuffers(1, &this->triangle_indices_VBO_id)); this->triangle_indices_VBO_id = 0; } if (this->quad_indices_VBO_id) { glsafe(::glDeleteBuffers(1, &this->quad_indices_VBO_id)); this->quad_indices_VBO_id = 0; } this->clear(); } void GLIndexedVertexArray::render() const { assert(this->vertices_and_normals_interleaved_VBO_id != 0); assert(this->triangle_indices_VBO_id != 0 || this->quad_indices_VBO_id != 0); glsafe(::glBindBuffer(GL_ARRAY_BUFFER, this->vertices_and_normals_interleaved_VBO_id)); glsafe(::glVertexPointer(3, GL_FLOAT, 6 * sizeof(float), (const void*)(3 * sizeof(float)))); glsafe(::glNormalPointer(GL_FLOAT, 6 * sizeof(float), nullptr)); glsafe(::glEnableClientState(GL_VERTEX_ARRAY)); glsafe(::glEnableClientState(GL_NORMAL_ARRAY)); // Render using the Vertex Buffer Objects. if (this->triangle_indices_size > 0) { glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, this->triangle_indices_VBO_id)); glsafe(::glDrawElements(GL_TRIANGLES, GLsizei(this->triangle_indices_size), GL_UNSIGNED_INT, nullptr)); glsafe(glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0)); } if (this->quad_indices_size > 0) { glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, this->quad_indices_VBO_id)); glsafe(::glDrawElements(GL_QUADS, GLsizei(this->quad_indices_size), GL_UNSIGNED_INT, nullptr)); glsafe(glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0)); } glsafe(::glDisableClientState(GL_VERTEX_ARRAY)); glsafe(::glDisableClientState(GL_NORMAL_ARRAY)); glsafe(::glBindBuffer(GL_ARRAY_BUFFER, 0)); } void GLIndexedVertexArray::render( const std::pair& tverts_range, const std::pair& qverts_range) const { assert(this->vertices_and_normals_interleaved_VBO_id != 0); assert(this->triangle_indices_VBO_id != 0 || this->quad_indices_VBO_id != 0); // Render using the Vertex Buffer Objects. glsafe(::glBindBuffer(GL_ARRAY_BUFFER, this->vertices_and_normals_interleaved_VBO_id)); glsafe(::glVertexPointer(3, GL_FLOAT, 6 * sizeof(float), (const void*)(3 * sizeof(float)))); glsafe(::glNormalPointer(GL_FLOAT, 6 * sizeof(float), nullptr)); glsafe(::glEnableClientState(GL_VERTEX_ARRAY)); glsafe(::glEnableClientState(GL_NORMAL_ARRAY)); if (this->triangle_indices_size > 0) { glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, this->triangle_indices_VBO_id)); glsafe(::glDrawElements(GL_TRIANGLES, GLsizei(std::min(this->triangle_indices_size, tverts_range.second - tverts_range.first)), GL_UNSIGNED_INT, (const void*)(tverts_range.first * 4))); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0)); } if (this->quad_indices_size > 0) { glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, this->quad_indices_VBO_id)); glsafe(::glDrawElements(GL_QUADS, GLsizei(std::min(this->quad_indices_size, qverts_range.second - qverts_range.first)), GL_UNSIGNED_INT, (const void*)(qverts_range.first * 4))); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0)); } glsafe(::glDisableClientState(GL_VERTEX_ARRAY)); glsafe(::glDisableClientState(GL_NORMAL_ARRAY)); glsafe(::glBindBuffer(GL_ARRAY_BUFFER, 0)); } const float GLVolume::SELECTED_COLOR[4] = { 0.0f, 1.0f, 0.0f, 1.0f }; const float GLVolume::HOVER_SELECT_COLOR[4] = { 0.4f, 0.9f, 0.1f, 1.0f }; const float GLVolume::HOVER_DESELECT_COLOR[4] = { 1.0f, 0.75f, 0.75f, 1.0f }; const float GLVolume::OUTSIDE_COLOR[4] = { 0.0f, 0.38f, 0.8f, 1.0f }; const float GLVolume::SELECTED_OUTSIDE_COLOR[4] = { 0.19f, 0.58f, 1.0f, 1.0f }; const float GLVolume::DISABLED_COLOR[4] = { 0.25f, 0.25f, 0.25f, 1.0f }; const float GLVolume::MODEL_COLOR[4][4] = { { 1.0f, 1.0f, 0.0f, 1.f }, { 1.0f, 0.5f, 0.5f, 1.f }, { 0.5f, 1.0f, 0.5f, 1.f }, { 0.5f, 0.5f, 1.0f, 1.f } }; const float GLVolume::SLA_SUPPORT_COLOR[4] = { 0.75f, 0.75f, 0.75f, 1.0f }; const float GLVolume::SLA_PAD_COLOR[4] = { 0.0f, 0.2f, 0.0f, 1.0f }; GLVolume::GLVolume(float r, float g, float b, float a) : m_transformed_bounding_box_dirty(true) , m_sla_shift_z(0.0) , m_transformed_convex_hull_bounding_box_dirty(true) // geometry_id == 0 -> invalid , geometry_id(std::pair(0, 0)) , extruder_id(0) , selected(false) , disabled(false) , printable(true) , is_active(true) , zoom_to_volumes(true) , shader_outside_printer_detection_enabled(false) , is_outside(false) , hover(HS_None) , is_modifier(false) , is_wipe_tower(false) , is_extrusion_path(false) , force_transparent(false) , force_native_color(false) , tverts_range(0, size_t(-1)) , qverts_range(0, size_t(-1)) { color[0] = r; color[1] = g; color[2] = b; color[3] = a; set_render_color(r, g, b, a); } void GLVolume::set_render_color(float r, float g, float b, float a) { render_color[0] = r; render_color[1] = g; render_color[2] = b; render_color[3] = a; } void GLVolume::set_render_color(const float* rgba, unsigned int size) { ::memcpy((void*)render_color, (const void*)rgba, (size_t)(std::min((unsigned int)4, size) * sizeof(float))); } void GLVolume::set_render_color() { if (force_native_color) { if (is_outside && shader_outside_printer_detection_enabled) set_render_color(OUTSIDE_COLOR, 4); else set_render_color(color, 4); } else { if (hover == HS_Select) set_render_color(HOVER_SELECT_COLOR, 4); else if (hover == HS_Deselect) set_render_color(HOVER_DESELECT_COLOR, 4); else if (selected) set_render_color(is_outside ? SELECTED_OUTSIDE_COLOR : SELECTED_COLOR, 4); else if (disabled) set_render_color(DISABLED_COLOR, 4); else if (is_outside && shader_outside_printer_detection_enabled) set_render_color(OUTSIDE_COLOR, 4); else set_render_color(color, 4); } if (!printable) { render_color[0] /= 4; render_color[1] /= 4; render_color[2] /= 4; } if (force_transparent) render_color[3] = color[3]; } void GLVolume::set_color_from_model_volume(const ModelVolume *model_volume) { if (model_volume->is_modifier()) { color[0] = 0.2f; color[1] = 1.0f; color[2] = 0.2f; } else if (model_volume->is_support_blocker()) { color[0] = 1.0f; color[1] = 0.2f; color[2] = 0.2f; } else if (model_volume->is_support_enforcer()) { color[0] = 0.2f; color[1] = 0.2f; color[2] = 1.0f; } color[3] = model_volume->is_model_part() ? 1.f : 0.5f; } Transform3d GLVolume::world_matrix() const { Transform3d m = m_instance_transformation.get_matrix() * m_volume_transformation.get_matrix(); m.translation()(2) += m_sla_shift_z; return m; } bool GLVolume::is_left_handed() const { const Vec3d &m1 = m_instance_transformation.get_mirror(); const Vec3d &m2 = m_volume_transformation.get_mirror(); return m1.x() * m1.y() * m1.z() * m2.x() * m2.y() * m2.z() < 0.; } const BoundingBoxf3& GLVolume::transformed_bounding_box() const { const BoundingBoxf3& box = bounding_box(); assert(box.defined || box.min(0) >= box.max(0) || box.min(1) >= box.max(1) || box.min(2) >= box.max(2)); if (m_transformed_bounding_box_dirty) { m_transformed_bounding_box = box.transformed(world_matrix()); m_transformed_bounding_box_dirty = false; } return m_transformed_bounding_box; } const BoundingBoxf3& GLVolume::transformed_convex_hull_bounding_box() const { if (m_transformed_convex_hull_bounding_box_dirty) m_transformed_convex_hull_bounding_box = this->transformed_convex_hull_bounding_box(world_matrix()); return m_transformed_convex_hull_bounding_box; } BoundingBoxf3 GLVolume::transformed_convex_hull_bounding_box(const Transform3d &trafo) const { return (m_convex_hull && m_convex_hull->stl.stats.number_of_facets > 0) ? m_convex_hull->transformed_bounding_box(trafo) : bounding_box().transformed(trafo); } void GLVolume::set_range(double min_z, double max_z) { this->qverts_range.first = 0; this->qverts_range.second = this->indexed_vertex_array.quad_indices_size; this->tverts_range.first = 0; this->tverts_range.second = this->indexed_vertex_array.triangle_indices_size; if (! this->print_zs.empty()) { // The Z layer range is specified. // First test whether the Z span of this object is not out of (min_z, max_z) completely. if (this->print_zs.front() > max_z || this->print_zs.back() < min_z) { this->qverts_range.second = 0; this->tverts_range.second = 0; } else { // Then find the lowest layer to be displayed. size_t i = 0; for (; i < this->print_zs.size() && this->print_zs[i] < min_z; ++ i); if (i == this->print_zs.size()) { // This shall not happen. this->qverts_range.second = 0; this->tverts_range.second = 0; } else { // Remember start of the layer. this->qverts_range.first = this->offsets[i * 2]; this->tverts_range.first = this->offsets[i * 2 + 1]; // Some layers are above $min_z. Which? for (; i < this->print_zs.size() && this->print_zs[i] <= max_z; ++ i); if (i < this->print_zs.size()) { this->qverts_range.second = this->offsets[i * 2]; this->tverts_range.second = this->offsets[i * 2 + 1]; } } } } } void GLVolume::render() const { if (!is_active) return; if (this->is_left_handed()) glFrontFace(GL_CW); glsafe(::glCullFace(GL_BACK)); glsafe(::glPushMatrix()); glsafe(::glMultMatrixd(world_matrix().data())); this->indexed_vertex_array.render(this->tverts_range, this->qverts_range); glsafe(::glPopMatrix()); if (this->is_left_handed()) glFrontFace(GL_CCW); } void GLVolume::render(int color_id, int detection_id, int worldmatrix_id) const { if (color_id >= 0) glsafe(::glUniform4fv(color_id, 1, (const GLfloat*)render_color)); else glsafe(::glColor4fv(render_color)); if (detection_id != -1) glsafe(::glUniform1i(detection_id, shader_outside_printer_detection_enabled ? 1 : 0)); if (worldmatrix_id != -1) glsafe(::glUniformMatrix4fv(worldmatrix_id, 1, GL_FALSE, (const GLfloat*)world_matrix().cast().data())); render(); } bool GLVolume::is_sla_support() const { return this->composite_id.volume_id == -int(slaposSupportTree); } bool GLVolume::is_sla_pad() const { return this->composite_id.volume_id == -int(slaposPad); } std::vector GLVolumeCollection::load_object( const ModelObject *model_object, int obj_idx, const std::vector &instance_idxs, const std::string &color_by, bool opengl_initialized) { std::vector volumes_idx; for (int volume_idx = 0; volume_idx < int(model_object->volumes.size()); ++volume_idx) for (int instance_idx : instance_idxs) volumes_idx.emplace_back(this->GLVolumeCollection::load_object_volume(model_object, obj_idx, volume_idx, instance_idx, color_by, opengl_initialized)); return volumes_idx; } int GLVolumeCollection::load_object_volume( const ModelObject *model_object, int obj_idx, int volume_idx, int instance_idx, const std::string &color_by, bool opengl_initialized) { const ModelVolume *model_volume = model_object->volumes[volume_idx]; const int extruder_id = model_volume->extruder_id(); const ModelInstance *instance = model_object->instances[instance_idx]; const TriangleMesh &mesh = model_volume->mesh(); float color[4]; memcpy(color, GLVolume::MODEL_COLOR[((color_by == "volume") ? volume_idx : obj_idx) % 4], sizeof(float) * 3); /* if (model_volume->is_support_blocker()) { color[0] = 1.0f; color[1] = 0.2f; color[2] = 0.2f; } else if (model_volume->is_support_enforcer()) { color[0] = 0.2f; color[1] = 0.2f; color[2] = 1.0f; } color[3] = model_volume->is_model_part() ? 1.f : 0.5f; */ color[3] = model_volume->is_model_part() ? 1.f : 0.5f; this->volumes.emplace_back(new GLVolume(color)); GLVolume& v = *this->volumes.back(); v.set_color_from_model_volume(model_volume); v.indexed_vertex_array.load_mesh(mesh); v.indexed_vertex_array.finalize_geometry(opengl_initialized); v.composite_id = GLVolume::CompositeID(obj_idx, volume_idx, instance_idx); if (model_volume->is_model_part()) { // GLVolume will reference a convex hull from model_volume! v.set_convex_hull(model_volume->get_convex_hull_shared_ptr()); if (extruder_id != -1) v.extruder_id = extruder_id; } v.is_modifier = !model_volume->is_model_part(); v.shader_outside_printer_detection_enabled = model_volume->is_model_part(); v.set_instance_transformation(instance->get_transformation()); v.set_volume_transformation(model_volume->get_transformation()); return int(this->volumes.size() - 1); } // Load SLA auxiliary GLVolumes (for support trees or pad). // This function produces volumes for multiple instances in a single shot, // as some object specific mesh conversions may be expensive. void GLVolumeCollection::load_object_auxiliary( const SLAPrintObject *print_object, int obj_idx, // pairs of const std::vector>& instances, SLAPrintObjectStep milestone, // Timestamp of the last change of the milestone size_t timestamp, bool opengl_initialized) { assert(print_object->is_step_done(milestone)); Transform3d mesh_trafo_inv = print_object->trafo().inverse(); // Get the support mesh. TriangleMesh mesh = print_object->get_mesh(milestone); mesh.transform(mesh_trafo_inv); // Convex hull is required for out of print bed detection. TriangleMesh convex_hull = mesh.convex_hull_3d(); for (const std::pair& instance_idx : instances) { const ModelInstance& model_instance = *print_object->model_object()->instances[instance_idx.first]; this->volumes.emplace_back(new GLVolume((milestone == slaposPad) ? GLVolume::SLA_PAD_COLOR : GLVolume::SLA_SUPPORT_COLOR)); GLVolume& v = *this->volumes.back(); v.indexed_vertex_array.load_mesh(mesh); v.indexed_vertex_array.finalize_geometry(opengl_initialized); v.composite_id = GLVolume::CompositeID(obj_idx, -int(milestone), (int)instance_idx.first); v.geometry_id = std::pair(timestamp, model_instance.id().id); // Create a copy of the convex hull mesh for each instance. Use a move operator on the last instance. if (&instance_idx == &instances.back()) v.set_convex_hull(std::move(convex_hull)); else v.set_convex_hull(convex_hull); v.is_modifier = false; v.shader_outside_printer_detection_enabled = (milestone == slaposSupportTree); v.set_instance_transformation(model_instance.get_transformation()); // Leave the volume transformation at identity. // v.set_volume_transformation(model_volume->get_transformation()); } } int GLVolumeCollection::load_wipe_tower_preview( int obj_idx, float pos_x, float pos_y, float width, float depth, float height, float rotation_angle, bool size_unknown, float brim_width, bool opengl_initialized) { if (depth < 0.01f) return int(this->volumes.size() - 1); if (height == 0.0f) height = 0.1f; Point origin_of_rotation(0.f, 0.f); TriangleMesh mesh; float color[4] = { 0.5f, 0.5f, 0.0f, 1.f }; // In case we don't know precise dimensions of the wipe tower yet, we'll draw the box with different color with one side jagged: if (size_unknown) { color[0] = 0.9f; color[1] = 0.6f; depth = std::max(depth, 10.f); // Too narrow tower would interfere with the teeth. The estimate is not precise anyway. float min_width = 30.f; // We'll now create the box with jagged edge. y-coordinates of the pre-generated model are shifted so that the front // edge has y=0 and centerline of the back edge has y=depth: Pointf3s points; std::vector facets; float out_points_idx[][3] = { { 0, -depth, 0 }, { 0, 0, 0 }, { 38.453f, 0, 0 }, { 61.547f, 0, 0 }, { 100.0f, 0, 0 }, { 100.0f, -depth, 0 }, { 55.7735f, -10.0f, 0 }, { 44.2265f, 10.0f, 0 }, { 38.453f, 0, 1 }, { 0, 0, 1 }, { 0, -depth, 1 }, { 100.0f, -depth, 1 }, { 100.0f, 0, 1 }, { 61.547f, 0, 1 }, { 55.7735f, -10.0f, 1 }, { 44.2265f, 10.0f, 1 } }; int out_facets_idx[][3] = { { 0, 1, 2 }, { 3, 4, 5 }, { 6, 5, 0 }, { 3, 5, 6 }, { 6, 2, 7 }, { 6, 0, 2 }, { 8, 9, 10 }, { 11, 12, 13 }, { 10, 11, 14 }, { 14, 11, 13 }, { 15, 8, 14 }, {8, 10, 14}, {3, 12, 4}, {3, 13, 12}, {6, 13, 3}, {6, 14, 13}, {7, 14, 6}, {7, 15, 14}, {2, 15, 7}, {2, 8, 15}, {1, 8, 2}, {1, 9, 8}, {0, 9, 1}, {0, 10, 9}, {5, 10, 0}, {5, 11, 10}, {4, 11, 5}, {4, 12, 11} }; for (int i = 0; i < 16; ++i) points.push_back(Vec3d(out_points_idx[i][0] / (100.f / min_width), out_points_idx[i][1] + depth, out_points_idx[i][2])); for (int i = 0; i < 28; ++i) facets.push_back(Vec3crd(out_facets_idx[i][0], out_facets_idx[i][1], out_facets_idx[i][2])); TriangleMesh tooth_mesh(points, facets); // We have the mesh ready. It has one tooth and width of min_width. We will now append several of these together until we are close to // the required width of the block. Than we can scale it precisely. size_t n = std::max(1, int(width / min_width)); // How many shall be merged? for (size_t i = 0; i < n; ++i) { mesh.merge(tooth_mesh); tooth_mesh.translate(min_width, 0.f, 0.f); } mesh.scale(Vec3d(width / (n * min_width), 1.f, height)); // Scaling to proper width } else mesh = make_cube(width, depth, height); // We'll make another mesh to show the brim (fixed layer height): TriangleMesh brim_mesh = make_cube(width + 2.f * brim_width, depth + 2.f * brim_width, 0.2f); brim_mesh.translate(-brim_width, -brim_width, 0.f); mesh.merge(brim_mesh); this->volumes.emplace_back(new GLVolume(color)); GLVolume& v = *this->volumes.back(); v.indexed_vertex_array.load_mesh(mesh); v.indexed_vertex_array.finalize_geometry(opengl_initialized); v.set_volume_offset(Vec3d(pos_x, pos_y, 0.0)); v.set_volume_rotation(Vec3d(0., 0., (M_PI / 180.) * rotation_angle)); v.composite_id = GLVolume::CompositeID(obj_idx, 0, 0); v.geometry_id.first = 0; v.geometry_id.second = wipe_tower_instance_id().id; v.is_wipe_tower = true; v.shader_outside_printer_detection_enabled = !size_unknown; return int(this->volumes.size() - 1); } GLVolume* GLVolumeCollection::new_toolpath_volume(const float *rgba, size_t reserve_vbo_floats) { GLVolume *out = new_nontoolpath_volume(rgba, reserve_vbo_floats); out->is_extrusion_path = true; return out; } GLVolume* GLVolumeCollection::new_nontoolpath_volume(const float *rgba, size_t reserve_vbo_floats) { GLVolume *out = new GLVolume(rgba); out->is_extrusion_path = false; // Reserving number of vertices (3x position + 3x color) out->indexed_vertex_array.reserve(reserve_vbo_floats / 6); this->volumes.emplace_back(out); return out; } GLVolumeWithIdAndZList volumes_to_render(const GLVolumePtrs& volumes, GLVolumeCollection::ERenderType type, const Transform3d& view_matrix, std::function filter_func) { GLVolumeWithIdAndZList list; list.reserve(volumes.size()); for (unsigned int i = 0; i < (unsigned int)volumes.size(); ++i) { GLVolume* volume = volumes[i]; bool is_transparent = (volume->render_color[3] < 1.0f); if ((((type == GLVolumeCollection::Opaque) && !is_transparent) || ((type == GLVolumeCollection::Transparent) && is_transparent) || (type == GLVolumeCollection::All)) && (! filter_func || filter_func(*volume))) list.emplace_back(std::make_pair(volume, std::make_pair(i, 0.0))); } if ((type == GLVolumeCollection::Transparent) && (list.size() > 1)) { for (GLVolumeWithIdAndZ& volume : list) { volume.second.second = volume.first->bounding_box().transformed(view_matrix * volume.first->world_matrix()).max(2); } std::sort(list.begin(), list.end(), [](const GLVolumeWithIdAndZ& v1, const GLVolumeWithIdAndZ& v2) -> bool { return v1.second.second < v2.second.second; } ); } else if ((type == GLVolumeCollection::Opaque) && (list.size() > 1)) { std::sort(list.begin(), list.end(), [](const GLVolumeWithIdAndZ& v1, const GLVolumeWithIdAndZ& v2) -> bool { return v1.first->selected && !v2.first->selected; } ); } return list; } void GLVolumeCollection::render(GLVolumeCollection::ERenderType type, bool disable_cullface, const Transform3d& view_matrix, std::function filter_func) const { glsafe(::glEnable(GL_BLEND)); glsafe(::glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA)); glsafe(::glCullFace(GL_BACK)); if (disable_cullface) glsafe(::glDisable(GL_CULL_FACE)); glsafe(::glEnableClientState(GL_VERTEX_ARRAY)); glsafe(::glEnableClientState(GL_NORMAL_ARRAY)); GLint current_program_id; glsafe(::glGetIntegerv(GL_CURRENT_PROGRAM, ¤t_program_id)); GLint color_id = (current_program_id > 0) ? ::glGetUniformLocation(current_program_id, "uniform_color") : -1; GLint z_range_id = (current_program_id > 0) ? ::glGetUniformLocation(current_program_id, "z_range") : -1; GLint clipping_plane_id = (current_program_id > 0) ? ::glGetUniformLocation(current_program_id, "clipping_plane") : -1; GLint print_box_min_id = (current_program_id > 0) ? ::glGetUniformLocation(current_program_id, "print_box.min") : -1; GLint print_box_max_id = (current_program_id > 0) ? ::glGetUniformLocation(current_program_id, "print_box.max") : -1; GLint print_box_detection_id = (current_program_id > 0) ? ::glGetUniformLocation(current_program_id, "print_box.volume_detection") : -1; GLint print_box_worldmatrix_id = (current_program_id > 0) ? ::glGetUniformLocation(current_program_id, "print_box.volume_world_matrix") : -1; glcheck(); if (print_box_min_id != -1) glsafe(::glUniform3fv(print_box_min_id, 1, (const GLfloat*)print_box_min)); if (print_box_max_id != -1) glsafe(::glUniform3fv(print_box_max_id, 1, (const GLfloat*)print_box_max)); if (z_range_id != -1) glsafe(::glUniform2fv(z_range_id, 1, (const GLfloat*)z_range)); if (clipping_plane_id != -1) glsafe(::glUniform4fv(clipping_plane_id, 1, (const GLfloat*)clipping_plane)); GLVolumeWithIdAndZList to_render = volumes_to_render(this->volumes, type, view_matrix, filter_func); for (GLVolumeWithIdAndZ& volume : to_render) { volume.first->set_render_color(); volume.first->render(color_id, print_box_detection_id, print_box_worldmatrix_id); } glsafe(::glBindBuffer(GL_ARRAY_BUFFER, 0)); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0)); glsafe(::glDisableClientState(GL_VERTEX_ARRAY)); glsafe(::glDisableClientState(GL_NORMAL_ARRAY)); if (disable_cullface) glsafe(::glEnable(GL_CULL_FACE)); glsafe(::glDisable(GL_BLEND)); } bool GLVolumeCollection::check_outside_state(const DynamicPrintConfig* config, ModelInstance::EPrintVolumeState* out_state) { if (config == nullptr) return false; const ConfigOptionPoints* opt = dynamic_cast(config->option("bed_shape")); if (opt == nullptr) return false; BoundingBox bed_box_2D = get_extents(Polygon::new_scale(opt->values)); BoundingBoxf3 print_volume(Vec3d(unscale(bed_box_2D.min(0)), unscale(bed_box_2D.min(1)), 0.0), Vec3d(unscale(bed_box_2D.max(0)), unscale(bed_box_2D.max(1)), config->opt_float("max_print_height"))); // Allow the objects to protrude below the print bed print_volume.min(2) = -1e10; ModelInstance::EPrintVolumeState state = ModelInstance::PVS_Inside; bool all_contained = true; bool contained_min_one = false; for (GLVolume* volume : this->volumes) { if ((volume == nullptr) || !volume->printable || volume->is_modifier || (volume->is_wipe_tower && !volume->shader_outside_printer_detection_enabled) || ((volume->composite_id.volume_id < 0) && !volume->shader_outside_printer_detection_enabled)) continue; const BoundingBoxf3& bb = volume->transformed_convex_hull_bounding_box(); bool contained = print_volume.contains(bb); all_contained &= contained; if (contained) contained_min_one = true; volume->is_outside = !contained; if ((state == ModelInstance::PVS_Inside) && volume->is_outside) state = ModelInstance::PVS_Fully_Outside; if ((state == ModelInstance::PVS_Fully_Outside) && volume->is_outside && print_volume.intersects(bb)) state = ModelInstance::PVS_Partly_Outside; } if (out_state != nullptr) *out_state = state; return /*all_contained*/ contained_min_one; // #ys_FIXME_delete_after_testing } void GLVolumeCollection::reset_outside_state() { for (GLVolume* volume : this->volumes) { if (volume != nullptr) volume->is_outside = false; } } void GLVolumeCollection::update_colors_by_extruder(const DynamicPrintConfig* config) { static const float inv_255 = 1.0f / 255.0f; struct Color { std::string text; unsigned char rgb[3]; Color() : text("") { rgb[0] = 255; rgb[1] = 255; rgb[2] = 255; } void set(const std::string& text, unsigned char* rgb) { this->text = text; ::memcpy((void*)this->rgb, (const void*)rgb, 3 * sizeof(unsigned char)); } }; if (config == nullptr) return; const ConfigOptionStrings* extruders_opt = dynamic_cast(config->option("extruder_colour")); if (extruders_opt == nullptr) return; const ConfigOptionStrings* filamemts_opt = dynamic_cast(config->option("filament_colour")); if (filamemts_opt == nullptr) return; unsigned int colors_count = std::max((unsigned int)extruders_opt->values.size(), (unsigned int)filamemts_opt->values.size()); if (colors_count == 0) return; std::vector colors(colors_count); unsigned char rgb[3]; for (unsigned int i = 0; i < colors_count; ++i) { const std::string& txt_color = config->opt_string("extruder_colour", i); if (PresetBundle::parse_color(txt_color, rgb)) { colors[i].set(txt_color, rgb); } else { const std::string& txt_color = config->opt_string("filament_colour", i); if (PresetBundle::parse_color(txt_color, rgb)) colors[i].set(txt_color, rgb); } } for (GLVolume* volume : volumes) { if ((volume == nullptr) || volume->is_modifier || volume->is_wipe_tower || (volume->volume_idx() < 0)) continue; int extruder_id = volume->extruder_id - 1; if ((extruder_id < 0) || ((int)colors.size() <= extruder_id)) extruder_id = 0; const Color& color = colors[extruder_id]; if (!color.text.empty()) { for (int i = 0; i < 3; ++i) { volume->color[i] = (float)color.rgb[i] * inv_255; } } } } std::vector GLVolumeCollection::get_current_print_zs(bool active_only) const { // Collect layer top positions of all volumes. std::vector print_zs; for (GLVolume *vol : this->volumes) { if (!active_only || vol->is_active) append(print_zs, vol->print_zs); } std::sort(print_zs.begin(), print_zs.end()); // Replace intervals of layers with similar top positions with their average value. int n = int(print_zs.size()); int k = 0; for (int i = 0; i < n;) { int j = i + 1; coordf_t zmax = print_zs[i] + EPSILON; for (; j < n && print_zs[j] <= zmax; ++ j) ; print_zs[k ++] = (j > i + 1) ? (0.5 * (print_zs[i] + print_zs[j - 1])) : print_zs[i]; i = j; } if (k < n) print_zs.erase(print_zs.begin() + k, print_zs.end()); return print_zs; } size_t GLVolumeCollection::cpu_memory_used() const { size_t memsize = sizeof(*this) + this->volumes.capacity() * sizeof(GLVolume); for (const GLVolume *volume : this->volumes) memsize += volume->cpu_memory_used(); return memsize; } size_t GLVolumeCollection::gpu_memory_used() const { size_t memsize = 0; for (const GLVolume *volume : this->volumes) memsize += volume->gpu_memory_used(); return memsize; } std::string GLVolumeCollection::log_memory_info() const { return " (GLVolumeCollection RAM: " + format_memsize_MB(this->cpu_memory_used()) + " GPU: " + format_memsize_MB(this->gpu_memory_used()) + " Both: " + format_memsize_MB(this->gpu_memory_used()) + ")"; } bool can_export_to_obj(const GLVolume& volume) { if (!volume.is_active || !volume.is_extrusion_path) return false; if (volume.indexed_vertex_array.triangle_indices.empty() && (std::min(volume.indexed_vertex_array.triangle_indices_size, volume.tverts_range.second - volume.tverts_range.first) == 0)) return false; if (volume.indexed_vertex_array.quad_indices.empty() && (std::min(volume.indexed_vertex_array.quad_indices_size, volume.qverts_range.second - volume.qverts_range.first) == 0)) return false; return true; } bool GLVolumeCollection::has_toolpaths_to_export() const { for (const GLVolume* volume : this->volumes) { if (can_export_to_obj(*volume)) return true; } return false; } void GLVolumeCollection::export_toolpaths_to_obj(const char* filename) const { if (filename == nullptr) return; if (!has_toolpaths_to_export()) return; // collect color information to generate materials typedef std::array Color; std::set colors; for (const GLVolume* volume : this->volumes) { if (!can_export_to_obj(*volume)) continue; Color color; ::memcpy((void*)color.data(), (const void*)volume->color, 4 * sizeof(float)); colors.insert(color); } // save materials file boost::filesystem::path mat_filename(filename); mat_filename.replace_extension("mtl"); FILE* fp = boost::nowide::fopen(mat_filename.string().c_str(), "w"); if (fp == nullptr) { BOOST_LOG_TRIVIAL(error) << "GLVolumeCollection::export_toolpaths_to_obj: Couldn't open " << mat_filename.string().c_str() << " for writing"; return; } fprintf(fp, "# G-Code Toolpaths Materials\n"); fprintf(fp, "# Generated by %s based on Slic3r\n", SLIC3R_BUILD_ID); unsigned int colors_count = 1; for (const Color& color : colors) { fprintf(fp, "\nnewmtl material_%d\n", colors_count++); fprintf(fp, "Ka 1 1 1\n"); fprintf(fp, "Kd %f %f %f\n", color[0], color[1], color[2]); fprintf(fp, "Ks 0 0 0\n"); } fclose(fp); // save geometry file fp = boost::nowide::fopen(filename, "w"); if (fp == nullptr) { BOOST_LOG_TRIVIAL(error) << "GLVolumeCollection::export_toolpaths_to_obj: Couldn't open " << filename << " for writing"; return; } fprintf(fp, "# G-Code Toolpaths\n"); fprintf(fp, "# Generated by %s based on Slic3r\n", SLIC3R_BUILD_ID); fprintf(fp, "\nmtllib ./%s\n", mat_filename.filename().string().c_str()); unsigned int vertices_count = 0; unsigned int normals_count = 0; unsigned int volumes_count = 0; for (const GLVolume* volume : this->volumes) { if (!can_export_to_obj(*volume)) continue; std::vector src_vertices_and_normals_interleaved; std::vector src_triangle_indices; std::vector src_quad_indices; if (!volume->indexed_vertex_array.vertices_and_normals_interleaved.empty()) // data are in CPU memory src_vertices_and_normals_interleaved = volume->indexed_vertex_array.vertices_and_normals_interleaved; else if ((volume->indexed_vertex_array.vertices_and_normals_interleaved_VBO_id != 0) && (volume->indexed_vertex_array.vertices_and_normals_interleaved_size != 0)) { // data are in GPU memory src_vertices_and_normals_interleaved = std::vector(volume->indexed_vertex_array.vertices_and_normals_interleaved_size, 0.0f); glsafe(::glBindBuffer(GL_ARRAY_BUFFER, volume->indexed_vertex_array.vertices_and_normals_interleaved_VBO_id)); glsafe(::glGetBufferSubData(GL_ARRAY_BUFFER, 0, src_vertices_and_normals_interleaved.size() * sizeof(float), src_vertices_and_normals_interleaved.data())); glsafe(::glBindBuffer(GL_ARRAY_BUFFER, 0)); } else continue; if (!volume->indexed_vertex_array.triangle_indices.empty()) { // data are in CPU memory size_t size = std::min(volume->indexed_vertex_array.triangle_indices.size(), volume->tverts_range.second - volume->tverts_range.first); if (size != 0) { std::vector::const_iterator it_begin = volume->indexed_vertex_array.triangle_indices.begin() + volume->tverts_range.first; std::vector::const_iterator it_end = volume->indexed_vertex_array.triangle_indices.begin() + volume->tverts_range.first + size; std::copy(it_begin, it_end, std::back_inserter(src_triangle_indices)); } } else if ((volume->indexed_vertex_array.triangle_indices_VBO_id != 0) && (volume->indexed_vertex_array.triangle_indices_size != 0)) { // data are in GPU memory size_t size = std::min(volume->indexed_vertex_array.triangle_indices_size, volume->tverts_range.second - volume->tverts_range.first); if (size != 0) { src_triangle_indices = std::vector(size, 0); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, volume->indexed_vertex_array.triangle_indices_VBO_id)); glsafe(::glGetBufferSubData(GL_ELEMENT_ARRAY_BUFFER, volume->tverts_range.first * sizeof(int), size * sizeof(int), src_triangle_indices.data())); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0)); } } if (!volume->indexed_vertex_array.quad_indices.empty()) { // data are in CPU memory size_t size = std::min(volume->indexed_vertex_array.quad_indices.size(), volume->qverts_range.second - volume->qverts_range.first); if (size != 0) { std::vector::const_iterator it_begin = volume->indexed_vertex_array.quad_indices.begin() + volume->qverts_range.first; std::vector::const_iterator it_end = volume->indexed_vertex_array.quad_indices.begin() + volume->qverts_range.first + size; std::copy(it_begin, it_end, std::back_inserter(src_quad_indices)); } } else if ((volume->indexed_vertex_array.quad_indices_VBO_id != 0) && (volume->indexed_vertex_array.quad_indices_size != 0)) { // data are in GPU memory size_t size = std::min(volume->indexed_vertex_array.quad_indices_size, volume->qverts_range.second - volume->qverts_range.first); if (size != 0) { src_quad_indices = std::vector(size, 0); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, volume->indexed_vertex_array.quad_indices_VBO_id)); glsafe(::glGetBufferSubData(GL_ELEMENT_ARRAY_BUFFER, volume->qverts_range.first * sizeof(int), size * sizeof(int), src_quad_indices.data())); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0)); } } if (src_triangle_indices.empty() && src_quad_indices.empty()) continue; ++volumes_count; // reduce output size by keeping only used vertices and normals struct Vector { std::array vector; explicit Vector(float* ptr) { vector[0] = scale_(*(ptr + 0)); vector[1] = scale_(*(ptr + 1)); vector[2] = scale_(*(ptr + 2)); } }; typedef std::vector Vectors; auto vector_less = [](const Vector& v1, const Vector& v2)->bool { return v1.vector < v2.vector; }; auto vector_equal = [](const Vector& v1, const Vector& v2)->bool { return (v1.vector[0] == v2.vector[0]) && (v1.vector[1] == v2.vector[1]) && (v1.vector[2] == v2.vector[2]); }; // copy used vertices and normals data Vectors dst_normals; Vectors dst_vertices; unsigned int src_triangle_indices_size = (unsigned int)src_triangle_indices.size(); for (unsigned int i = 0; i < src_triangle_indices_size; ++i) { float* src_ptr = src_vertices_and_normals_interleaved.data() + src_triangle_indices[i] * 6; dst_normals.emplace_back(src_ptr + 0); dst_vertices.emplace_back(src_ptr + 3); } unsigned int src_quad_indices_size = (unsigned int)src_quad_indices.size(); for (unsigned int i = 0; i < src_quad_indices_size; ++i) { float* src_ptr = src_vertices_and_normals_interleaved.data() + src_quad_indices[i] * 6; dst_normals.emplace_back(src_ptr + 0); dst_vertices.emplace_back(src_ptr + 3); } // sort vertices and normals std::sort(dst_normals.begin(), dst_normals.end(), vector_less); std::sort(dst_vertices.begin(), dst_vertices.end(), vector_less); // remove duplicated vertices and normals dst_normals.erase(std::unique(dst_normals.begin(), dst_normals.end(), vector_equal), dst_normals.end()); dst_vertices.erase(std::unique(dst_vertices.begin(), dst_vertices.end(), vector_equal), dst_vertices.end()); // reindex triangles and quads struct IndicesPair { int vertex; int normal; IndicesPair(int vertex, int normal) : vertex(vertex), normal(normal) {} }; typedef std::vector Indices; unsigned int src_vertices_count = (unsigned int)src_vertices_and_normals_interleaved.size() / 6; std::vector src_dst_vertex_indices_map(src_vertices_count, -1); std::vector src_dst_normal_indices_map(src_vertices_count, -1); for (unsigned int i = 0; i < src_vertices_count; ++i) { float* src_ptr = src_vertices_and_normals_interleaved.data() + i * 6; src_dst_normal_indices_map[i] = std::distance(dst_normals.begin(), std::lower_bound(dst_normals.begin(), dst_normals.end(), Vector(src_ptr + 0), vector_less)); src_dst_vertex_indices_map[i] = std::distance(dst_vertices.begin(), std::lower_bound(dst_vertices.begin(), dst_vertices.end(), Vector(src_ptr + 3), vector_less)); } Indices dst_triangle_indices; if (src_triangle_indices_size > 0) dst_triangle_indices.reserve(src_triangle_indices_size); for (unsigned int i = 0; i < src_triangle_indices_size; ++i) { int id = src_triangle_indices[i]; dst_triangle_indices.emplace_back(src_dst_vertex_indices_map[id], src_dst_normal_indices_map[id]); } Indices dst_quad_indices; if (src_quad_indices_size > 0) dst_quad_indices.reserve(src_quad_indices_size); for (unsigned int i = 0; i < src_quad_indices_size; ++i) { int id = src_quad_indices[i]; dst_quad_indices.emplace_back(src_dst_vertex_indices_map[id], src_dst_normal_indices_map[id]); } // save to file fprintf(fp, "\n# vertices volume %d\n", volumes_count); for (const Vector& v : dst_vertices) { fprintf(fp, "v %g %g %g\n", unscale(v.vector[0]), unscale(v.vector[1]), unscale(v.vector[2])); } fprintf(fp, "\n# normals volume %d\n", volumes_count); for (const Vector& n : dst_normals) { fprintf(fp, "vn %g %g %g\n", unscale(n.vector[0]), unscale(n.vector[1]), unscale(n.vector[2])); } Color color; ::memcpy((void*)color.data(), (const void*)volume->color, 4 * sizeof(float)); fprintf(fp, "\n# material volume %d\n", volumes_count); fprintf(fp, "usemtl material_%lld\n", (long long)(1 + std::distance(colors.begin(), colors.find(color)))); int base_vertex_id = vertices_count + 1; int base_normal_id = normals_count + 1; if (!dst_triangle_indices.empty()) { fprintf(fp, "\n# triangular facets volume %d\n", volumes_count); for (unsigned int i = 0; i < (unsigned int)dst_triangle_indices.size(); i += 3) { fprintf(fp, "f %d//%d %d//%d %d//%d\n", base_vertex_id + dst_triangle_indices[i + 0].vertex, base_normal_id + dst_triangle_indices[i + 0].normal, base_vertex_id + dst_triangle_indices[i + 1].vertex, base_normal_id + dst_triangle_indices[i + 1].normal, base_vertex_id + dst_triangle_indices[i + 2].vertex, base_normal_id + dst_triangle_indices[i + 2].normal); } } if (!dst_quad_indices.empty()) { fprintf(fp, "\n# quadrangular facets volume %d\n", volumes_count); for (unsigned int i = 0; i < (unsigned int)src_quad_indices.size(); i += 4) { fprintf(fp, "f %d//%d %d//%d %d//%d %d//%d\n", base_vertex_id + dst_quad_indices[i + 0].vertex, base_normal_id + dst_quad_indices[i + 0].normal, base_vertex_id + dst_quad_indices[i + 1].vertex, base_normal_id + dst_quad_indices[i + 1].normal, base_vertex_id + dst_quad_indices[i + 2].vertex, base_normal_id + dst_quad_indices[i + 2].normal, base_vertex_id + dst_quad_indices[i + 3].vertex, base_normal_id + dst_quad_indices[i + 3].normal); } } vertices_count += (unsigned int)dst_vertices.size(); normals_count += (unsigned int)dst_normals.size(); } fclose(fp); } // caller is responsible for supplying NO lines with zero length static void thick_lines_to_indexed_vertex_array( const Lines &lines, const std::vector &widths, const std::vector &heights, bool closed, double top_z, GLIndexedVertexArray &volume) { assert(! lines.empty()); if (lines.empty()) return; #define LEFT 0 #define RIGHT 1 #define TOP 2 #define BOTTOM 3 // right, left, top, bottom int idx_prev[4] = { -1, -1, -1, -1 }; double bottom_z_prev = 0.; Vec2d b1_prev(Vec2d::Zero()); Vec2d v_prev(Vec2d::Zero()); int idx_initial[4] = { -1, -1, -1, -1 }; double width_initial = 0.; double bottom_z_initial = 0.0; double len_prev = 0.0; // loop once more in case of closed loops size_t lines_end = closed ? (lines.size() + 1) : lines.size(); for (size_t ii = 0; ii < lines_end; ++ ii) { size_t i = (ii == lines.size()) ? 0 : ii; const Line &line = lines[i]; double bottom_z = top_z - heights[i]; double middle_z = 0.5 * (top_z + bottom_z); double width = widths[i]; bool is_first = (ii == 0); bool is_last = (ii == lines_end - 1); bool is_closing = closed && is_last; Vec2d v = unscale(line.vector()).normalized(); double len = unscale(line.length()); Vec2d a = unscale(line.a); Vec2d b = unscale(line.b); Vec2d a1 = a; Vec2d a2 = a; Vec2d b1 = b; Vec2d b2 = b; { double dist = 0.5 * width; // scaled double dx = dist * v(0); double dy = dist * v(1); a1 += Vec2d(+dy, -dx); a2 += Vec2d(-dy, +dx); b1 += Vec2d(+dy, -dx); b2 += Vec2d(-dy, +dx); } // calculate new XY normals Vec2d xy_right_normal = unscale(line.normal()).normalized(); int idx_a[4] = { 0, 0, 0, 0 }; // initialized to avoid warnings int idx_b[4] = { 0, 0, 0, 0 }; // initialized to avoid warnings int idx_last = int(volume.vertices_and_normals_interleaved.size() / 6); bool bottom_z_different = bottom_z_prev != bottom_z; bottom_z_prev = bottom_z; if (!is_first && bottom_z_different) { // Found a change of the layer thickness -> Add a cap at the end of the previous segment. volume.push_quad(idx_b[BOTTOM], idx_b[LEFT], idx_b[TOP], idx_b[RIGHT]); } // Share top / bottom vertices if possible. if (is_first) { idx_a[TOP] = idx_last++; volume.push_geometry(a(0), a(1), top_z , 0., 0., 1.); } else { idx_a[TOP] = idx_prev[TOP]; } if (is_first || bottom_z_different) { // Start of the 1st line segment or a change of the layer thickness while maintaining the print_z. idx_a[BOTTOM] = idx_last ++; volume.push_geometry(a(0), a(1), bottom_z, 0., 0., -1.); idx_a[LEFT ] = idx_last ++; volume.push_geometry(a2(0), a2(1), middle_z, -xy_right_normal(0), -xy_right_normal(1), 0.0); idx_a[RIGHT] = idx_last ++; volume.push_geometry(a1(0), a1(1), middle_z, xy_right_normal(0), xy_right_normal(1), 0.0); } else { idx_a[BOTTOM] = idx_prev[BOTTOM]; } if (is_first) { // Start of the 1st line segment. width_initial = width; bottom_z_initial = bottom_z; memcpy(idx_initial, idx_a, sizeof(int) * 4); } else { // Continuing a previous segment. // Share left / right vertices if possible. double v_dot = v_prev.dot(v); // To reduce gpu memory usage, we try to reuse vertices // To reduce the visual artifacts, due to averaged normals, we allow to reuse vertices only when any of two adjacent edges // is longer than a fixed threshold. // The following value is arbitrary, it comes from tests made on a bunch of models showing the visual artifacts double len_threshold = 2.5; // Generate new vertices if the angle between adjacent edges is greater than 45 degrees or thresholds conditions are met bool sharp = (v_dot < 0.707) || (len_prev > len_threshold) || (len > len_threshold); if (sharp) { if (!bottom_z_different) { // Allocate new left / right points for the start of this segment as these points will receive their own normals to indicate a sharp turn. idx_a[RIGHT] = idx_last++; volume.push_geometry(a1(0), a1(1), middle_z, xy_right_normal(0), xy_right_normal(1), 0.0); idx_a[LEFT] = idx_last++; volume.push_geometry(a2(0), a2(1), middle_z, -xy_right_normal(0), -xy_right_normal(1), 0.0); if (cross2(v_prev, v) > 0.) { // Right turn. Fill in the right turn wedge. volume.push_triangle(idx_prev[RIGHT], idx_a[RIGHT], idx_prev[TOP]); volume.push_triangle(idx_prev[RIGHT], idx_prev[BOTTOM], idx_a[RIGHT]); } else { // Left turn. Fill in the left turn wedge. volume.push_triangle(idx_prev[LEFT], idx_prev[TOP], idx_a[LEFT]); volume.push_triangle(idx_prev[LEFT], idx_a[LEFT], idx_prev[BOTTOM]); } } } else { if (!bottom_z_different) { // The two successive segments are nearly collinear. idx_a[LEFT ] = idx_prev[LEFT]; idx_a[RIGHT] = idx_prev[RIGHT]; } } if (is_closing) { if (!sharp) { if (!bottom_z_different) { // Closing a loop with smooth transition. Unify the closing left / right vertices. memcpy(volume.vertices_and_normals_interleaved.data() + idx_initial[LEFT ] * 6, volume.vertices_and_normals_interleaved.data() + idx_prev[LEFT ] * 6, sizeof(float) * 6); memcpy(volume.vertices_and_normals_interleaved.data() + idx_initial[RIGHT] * 6, volume.vertices_and_normals_interleaved.data() + idx_prev[RIGHT] * 6, sizeof(float) * 6); volume.vertices_and_normals_interleaved.erase(volume.vertices_and_normals_interleaved.end() - 12, volume.vertices_and_normals_interleaved.end()); // Replace the left / right vertex indices to point to the start of the loop. for (size_t u = volume.quad_indices.size() - 16; u < volume.quad_indices.size(); ++ u) { if (volume.quad_indices[u] == idx_prev[LEFT]) volume.quad_indices[u] = idx_initial[LEFT]; else if (volume.quad_indices[u] == idx_prev[RIGHT]) volume.quad_indices[u] = idx_initial[RIGHT]; } } } // This is the last iteration, only required to solve the transition. break; } } // Only new allocate top / bottom vertices, if not closing a loop. if (is_closing) { idx_b[TOP] = idx_initial[TOP]; } else { idx_b[TOP] = idx_last ++; volume.push_geometry(b(0), b(1), top_z , 0., 0., 1.); } if (is_closing && (width == width_initial) && (bottom_z == bottom_z_initial)) { idx_b[BOTTOM] = idx_initial[BOTTOM]; } else { idx_b[BOTTOM] = idx_last ++; volume.push_geometry(b(0), b(1), bottom_z, 0., 0., -1.); } // Generate new vertices for the end of this line segment. idx_b[LEFT ] = idx_last ++; volume.push_geometry(b2(0), b2(1), middle_z, -xy_right_normal(0), -xy_right_normal(1), 0.0); idx_b[RIGHT ] = idx_last ++; volume.push_geometry(b1(0), b1(1), middle_z, xy_right_normal(0), xy_right_normal(1), 0.0); memcpy(idx_prev, idx_b, 4 * sizeof(int)); bottom_z_prev = bottom_z; b1_prev = b1; v_prev = v; len_prev = len; if (bottom_z_different && (closed || (!is_first && !is_last))) { // Found a change of the layer thickness -> Add a cap at the beginning of this segment. volume.push_quad(idx_a[BOTTOM], idx_a[RIGHT], idx_a[TOP], idx_a[LEFT]); } if (! closed) { // Terminate open paths with caps. if (is_first) volume.push_quad(idx_a[BOTTOM], idx_a[RIGHT], idx_a[TOP], idx_a[LEFT]); // We don't use 'else' because both cases are true if we have only one line. if (is_last) volume.push_quad(idx_b[BOTTOM], idx_b[LEFT], idx_b[TOP], idx_b[RIGHT]); } // Add quads for a straight hollow tube-like segment. // bottom-right face volume.push_quad(idx_a[BOTTOM], idx_b[BOTTOM], idx_b[RIGHT], idx_a[RIGHT]); // top-right face volume.push_quad(idx_a[RIGHT], idx_b[RIGHT], idx_b[TOP], idx_a[TOP]); // top-left face volume.push_quad(idx_a[TOP], idx_b[TOP], idx_b[LEFT], idx_a[LEFT]); // bottom-left face volume.push_quad(idx_a[LEFT], idx_b[LEFT], idx_b[BOTTOM], idx_a[BOTTOM]); } #undef LEFT #undef RIGHT #undef TOP #undef BOTTOM } // caller is responsible for supplying NO lines with zero length static void thick_lines_to_indexed_vertex_array(const Lines3& lines, const std::vector& widths, const std::vector& heights, bool closed, GLIndexedVertexArray& volume) { assert(!lines.empty()); if (lines.empty()) return; #define LEFT 0 #define RIGHT 1 #define TOP 2 #define BOTTOM 3 // left, right, top, bottom int idx_initial[4] = { -1, -1, -1, -1 }; int idx_prev[4] = { -1, -1, -1, -1 }; double z_prev = 0.0; double len_prev = 0.0; Vec3d n_right_prev = Vec3d::Zero(); Vec3d n_top_prev = Vec3d::Zero(); Vec3d unit_v_prev = Vec3d::Zero(); double width_initial = 0.0; // new vertices around the line endpoints // left, right, top, bottom Vec3d a[4] = { Vec3d::Zero(), Vec3d::Zero(), Vec3d::Zero(), Vec3d::Zero() }; Vec3d b[4] = { Vec3d::Zero(), Vec3d::Zero(), Vec3d::Zero(), Vec3d::Zero() }; // loop once more in case of closed loops size_t lines_end = closed ? (lines.size() + 1) : lines.size(); for (size_t ii = 0; ii < lines_end; ++ii) { size_t i = (ii == lines.size()) ? 0 : ii; const Line3& line = lines[i]; double height = heights[i]; double width = widths[i]; Vec3d unit_v = unscale(line.vector()).normalized(); double len = unscale(line.length()); Vec3d n_top = Vec3d::Zero(); Vec3d n_right = Vec3d::Zero(); if ((line.a(0) == line.b(0)) && (line.a(1) == line.b(1))) { // vertical segment n_top = Vec3d::UnitY(); n_right = Vec3d::UnitX(); if (line.a(2) < line.b(2)) n_right = -n_right; } else { // horizontal segment n_right = unit_v.cross(Vec3d::UnitZ()).normalized(); n_top = n_right.cross(unit_v).normalized(); } Vec3d rl_displacement = 0.5 * width * n_right; Vec3d tb_displacement = 0.5 * height * n_top; Vec3d l_a = unscale(line.a); Vec3d l_b = unscale(line.b); a[RIGHT] = l_a + rl_displacement; a[LEFT] = l_a - rl_displacement; a[TOP] = l_a + tb_displacement; a[BOTTOM] = l_a - tb_displacement; b[RIGHT] = l_b + rl_displacement; b[LEFT] = l_b - rl_displacement; b[TOP] = l_b + tb_displacement; b[BOTTOM] = l_b - tb_displacement; Vec3d n_bottom = -n_top; Vec3d n_left = -n_right; int idx_a[4]; int idx_b[4]; int idx_last = int(volume.vertices_and_normals_interleaved.size() / 6); bool z_different = (z_prev != l_a(2)); z_prev = l_b(2); // Share top / bottom vertices if possible. if (ii == 0) { idx_a[TOP] = idx_last++; volume.push_geometry(a[TOP], n_top); } else idx_a[TOP] = idx_prev[TOP]; if ((ii == 0) || z_different) { // Start of the 1st line segment or a change of the layer thickness while maintaining the print_z. idx_a[BOTTOM] = idx_last++; volume.push_geometry(a[BOTTOM], n_bottom); idx_a[LEFT] = idx_last++; volume.push_geometry(a[LEFT], n_left); idx_a[RIGHT] = idx_last++; volume.push_geometry(a[RIGHT], n_right); } else idx_a[BOTTOM] = idx_prev[BOTTOM]; if (ii == 0) { // Start of the 1st line segment. width_initial = width; ::memcpy(idx_initial, idx_a, sizeof(int) * 4); } else { // Continuing a previous segment. // Share left / right vertices if possible. double v_dot = unit_v_prev.dot(unit_v); bool is_right_turn = n_top_prev.dot(unit_v_prev.cross(unit_v)) > 0.0; // To reduce gpu memory usage, we try to reuse vertices // To reduce the visual artifacts, due to averaged normals, we allow to reuse vertices only when any of two adjacent edges // is longer than a fixed threshold. // The following value is arbitrary, it comes from tests made on a bunch of models showing the visual artifacts double len_threshold = 2.5; // Generate new vertices if the angle between adjacent edges is greater than 45 degrees or thresholds conditions are met bool is_sharp = (v_dot < 0.707) || (len_prev > len_threshold) || (len > len_threshold); if (is_sharp) { // Allocate new left / right points for the start of this segment as these points will receive their own normals to indicate a sharp turn. idx_a[RIGHT] = idx_last++; volume.push_geometry(a[RIGHT], n_right); idx_a[LEFT] = idx_last++; volume.push_geometry(a[LEFT], n_left); if (is_right_turn) { // Right turn. Fill in the right turn wedge. volume.push_triangle(idx_prev[RIGHT], idx_a[RIGHT], idx_prev[TOP]); volume.push_triangle(idx_prev[RIGHT], idx_prev[BOTTOM], idx_a[RIGHT]); } else { // Left turn. Fill in the left turn wedge. volume.push_triangle(idx_prev[LEFT], idx_prev[TOP], idx_a[LEFT]); volume.push_triangle(idx_prev[LEFT], idx_a[LEFT], idx_prev[BOTTOM]); } } else { // The two successive segments are nearly collinear. idx_a[LEFT] = idx_prev[LEFT]; idx_a[RIGHT] = idx_prev[RIGHT]; } if (ii == lines.size()) { if (!is_sharp) { // Closing a loop with smooth transition. Unify the closing left / right vertices. ::memcpy(volume.vertices_and_normals_interleaved.data() + idx_initial[LEFT] * 6, volume.vertices_and_normals_interleaved.data() + idx_prev[LEFT] * 6, sizeof(float) * 6); ::memcpy(volume.vertices_and_normals_interleaved.data() + idx_initial[RIGHT] * 6, volume.vertices_and_normals_interleaved.data() + idx_prev[RIGHT] * 6, sizeof(float) * 6); volume.vertices_and_normals_interleaved.erase(volume.vertices_and_normals_interleaved.end() - 12, volume.vertices_and_normals_interleaved.end()); // Replace the left / right vertex indices to point to the start of the loop. for (size_t u = volume.quad_indices.size() - 16; u < volume.quad_indices.size(); ++u) { if (volume.quad_indices[u] == idx_prev[LEFT]) volume.quad_indices[u] = idx_initial[LEFT]; else if (volume.quad_indices[u] == idx_prev[RIGHT]) volume.quad_indices[u] = idx_initial[RIGHT]; } } // This is the last iteration, only required to solve the transition. break; } } // Only new allocate top / bottom vertices, if not closing a loop. if (closed && (ii + 1 == lines.size())) idx_b[TOP] = idx_initial[TOP]; else { idx_b[TOP] = idx_last++; volume.push_geometry(b[TOP], n_top); } if (closed && (ii + 1 == lines.size()) && (width == width_initial)) idx_b[BOTTOM] = idx_initial[BOTTOM]; else { idx_b[BOTTOM] = idx_last++; volume.push_geometry(b[BOTTOM], n_bottom); } // Generate new vertices for the end of this line segment. idx_b[LEFT] = idx_last++; volume.push_geometry(b[LEFT], n_left); idx_b[RIGHT] = idx_last++; volume.push_geometry(b[RIGHT], n_right); ::memcpy(idx_prev, idx_b, 4 * sizeof(int)); n_right_prev = n_right; n_top_prev = n_top; unit_v_prev = unit_v; len_prev = len; if (!closed) { // Terminate open paths with caps. if (i == 0) volume.push_quad(idx_a[BOTTOM], idx_a[RIGHT], idx_a[TOP], idx_a[LEFT]); // We don't use 'else' because both cases are true if we have only one line. if (i + 1 == lines.size()) volume.push_quad(idx_b[BOTTOM], idx_b[LEFT], idx_b[TOP], idx_b[RIGHT]); } // Add quads for a straight hollow tube-like segment. // bottom-right face volume.push_quad(idx_a[BOTTOM], idx_b[BOTTOM], idx_b[RIGHT], idx_a[RIGHT]); // top-right face volume.push_quad(idx_a[RIGHT], idx_b[RIGHT], idx_b[TOP], idx_a[TOP]); // top-left face volume.push_quad(idx_a[TOP], idx_b[TOP], idx_b[LEFT], idx_a[LEFT]); // bottom-left face volume.push_quad(idx_a[LEFT], idx_b[LEFT], idx_b[BOTTOM], idx_a[BOTTOM]); } #undef LEFT #undef RIGHT #undef TOP #undef BOTTOM } static void point_to_indexed_vertex_array(const Vec3crd& point, double width, double height, GLIndexedVertexArray& volume) { // builds a double piramid, with vertices on the local axes, around the point Vec3d center = unscale(point); double scale_factor = 1.0; double w = scale_factor * width; double h = scale_factor * height; // new vertices ids int idx_last = int(volume.vertices_and_normals_interleaved.size() / 6); int idxs[6]; for (int i = 0; i < 6; ++i) { idxs[i] = idx_last + i; } Vec3d displacement_x(w, 0.0, 0.0); Vec3d displacement_y(0.0, w, 0.0); Vec3d displacement_z(0.0, 0.0, h); Vec3d unit_x(1.0, 0.0, 0.0); Vec3d unit_y(0.0, 1.0, 0.0); Vec3d unit_z(0.0, 0.0, 1.0); // vertices volume.push_geometry(center - displacement_x, -unit_x); // idxs[0] volume.push_geometry(center + displacement_x, unit_x); // idxs[1] volume.push_geometry(center - displacement_y, -unit_y); // idxs[2] volume.push_geometry(center + displacement_y, unit_y); // idxs[3] volume.push_geometry(center - displacement_z, -unit_z); // idxs[4] volume.push_geometry(center + displacement_z, unit_z); // idxs[5] // top piramid faces volume.push_triangle(idxs[0], idxs[2], idxs[5]); volume.push_triangle(idxs[2], idxs[1], idxs[5]); volume.push_triangle(idxs[1], idxs[3], idxs[5]); volume.push_triangle(idxs[3], idxs[0], idxs[5]); // bottom piramid faces volume.push_triangle(idxs[2], idxs[0], idxs[4]); volume.push_triangle(idxs[1], idxs[2], idxs[4]); volume.push_triangle(idxs[3], idxs[1], idxs[4]); volume.push_triangle(idxs[0], idxs[3], idxs[4]); } void _3DScene::thick_lines_to_verts( const Lines &lines, const std::vector &widths, const std::vector &heights, bool closed, double top_z, GLVolume &volume) { thick_lines_to_indexed_vertex_array(lines, widths, heights, closed, top_z, volume.indexed_vertex_array); } void _3DScene::thick_lines_to_verts(const Lines3& lines, const std::vector& widths, const std::vector& heights, bool closed, GLVolume& volume) { thick_lines_to_indexed_vertex_array(lines, widths, heights, closed, volume.indexed_vertex_array); } static void thick_point_to_verts(const Vec3crd& point, double width, double height, GLVolume& volume) { point_to_indexed_vertex_array(point, width, height, volume.indexed_vertex_array); } void _3DScene::extrusionentity_to_verts(const Polyline &polyline, float width, float height, float print_z, GLVolume& volume) { if (polyline.size() >= 2) { size_t num_segments = polyline.size() - 1; thick_lines_to_verts(polyline.lines(), std::vector(num_segments, width), std::vector(num_segments, height), false, print_z, volume); } } // Fill in the qverts and tverts with quads and triangles for the extrusion_path. void _3DScene::extrusionentity_to_verts(const ExtrusionPath &extrusion_path, float print_z, GLVolume &volume) { extrusionentity_to_verts(extrusion_path.polyline, extrusion_path.width, extrusion_path.height, print_z, volume); } // Fill in the qverts and tverts with quads and triangles for the extrusion_path. void _3DScene::extrusionentity_to_verts(const ExtrusionPath &extrusion_path, float print_z, const Point ©, GLVolume &volume) { Polyline polyline = extrusion_path.polyline; polyline.remove_duplicate_points(); polyline.translate(copy); Lines lines = polyline.lines(); std::vector widths(lines.size(), extrusion_path.width); std::vector heights(lines.size(), extrusion_path.height); thick_lines_to_verts(lines, widths, heights, false, print_z, volume); } // Fill in the qverts and tverts with quads and triangles for the extrusion_loop. void _3DScene::extrusionentity_to_verts(const ExtrusionLoop &extrusion_loop, float print_z, const Point ©, GLVolume &volume) { Lines lines; std::vector widths; std::vector heights; for (const ExtrusionPath &extrusion_path : extrusion_loop.paths) { Polyline polyline = extrusion_path.polyline; polyline.remove_duplicate_points(); polyline.translate(copy); Lines lines_this = polyline.lines(); append(lines, lines_this); widths.insert(widths.end(), lines_this.size(), extrusion_path.width); heights.insert(heights.end(), lines_this.size(), extrusion_path.height); } thick_lines_to_verts(lines, widths, heights, true, print_z, volume); } // Fill in the qverts and tverts with quads and triangles for the extrusion_multi_path. void _3DScene::extrusionentity_to_verts(const ExtrusionMultiPath &extrusion_multi_path, float print_z, const Point ©, GLVolume &volume) { Lines lines; std::vector widths; std::vector heights; for (const ExtrusionPath &extrusion_path : extrusion_multi_path.paths) { Polyline polyline = extrusion_path.polyline; polyline.remove_duplicate_points(); polyline.translate(copy); Lines lines_this = polyline.lines(); append(lines, lines_this); widths.insert(widths.end(), lines_this.size(), extrusion_path.width); heights.insert(heights.end(), lines_this.size(), extrusion_path.height); } thick_lines_to_verts(lines, widths, heights, false, print_z, volume); } void _3DScene::extrusionentity_to_verts(const ExtrusionEntityCollection &extrusion_entity_collection, float print_z, const Point ©, GLVolume &volume) { for (const ExtrusionEntity *extrusion_entity : extrusion_entity_collection.entities) extrusionentity_to_verts(extrusion_entity, print_z, copy, volume); } void _3DScene::extrusionentity_to_verts(const ExtrusionEntity *extrusion_entity, float print_z, const Point ©, GLVolume &volume) { if (extrusion_entity != nullptr) { auto *extrusion_path = dynamic_cast(extrusion_entity); if (extrusion_path != nullptr) extrusionentity_to_verts(*extrusion_path, print_z, copy, volume); else { auto *extrusion_loop = dynamic_cast(extrusion_entity); if (extrusion_loop != nullptr) extrusionentity_to_verts(*extrusion_loop, print_z, copy, volume); else { auto *extrusion_multi_path = dynamic_cast(extrusion_entity); if (extrusion_multi_path != nullptr) extrusionentity_to_verts(*extrusion_multi_path, print_z, copy, volume); else { auto *extrusion_entity_collection = dynamic_cast(extrusion_entity); if (extrusion_entity_collection != nullptr) extrusionentity_to_verts(*extrusion_entity_collection, print_z, copy, volume); else { throw std::runtime_error("Unexpected extrusion_entity type in to_verts()"); } } } } } } void _3DScene::polyline3_to_verts(const Polyline3& polyline, double width, double height, GLVolume& volume) { Lines3 lines = polyline.lines(); std::vector widths(lines.size(), width); std::vector heights(lines.size(), height); thick_lines_to_verts(lines, widths, heights, false, volume); } void _3DScene::point3_to_verts(const Vec3crd& point, double width, double height, GLVolume& volume) { thick_point_to_verts(point, width, height, volume); } GUI::GLCanvas3DManager _3DScene::s_canvas_mgr; GLModel::GLModel() : m_filename("") { m_volume.shader_outside_printer_detection_enabled = false; } GLModel::~GLModel() { reset(); } void GLModel::set_color(const float* color, unsigned int size) { ::memcpy((void*)m_volume.color, (const void*)color, (size_t)(std::min((unsigned int)4, size) * sizeof(float))); m_volume.set_render_color(color, size); } const Vec3d& GLModel::get_offset() const { return m_volume.get_volume_offset(); } void GLModel::set_offset(const Vec3d& offset) { m_volume.set_volume_offset(offset); } const Vec3d& GLModel::get_rotation() const { return m_volume.get_volume_rotation(); } void GLModel::set_rotation(const Vec3d& rotation) { m_volume.set_volume_rotation(rotation); } const Vec3d& GLModel::get_scale() const { return m_volume.get_volume_scaling_factor(); } void GLModel::set_scale(const Vec3d& scale) { m_volume.set_volume_scaling_factor(scale); } void GLModel::reset() { m_volume.indexed_vertex_array.release_geometry(); m_filename = ""; } void GLModel::render() const { glsafe(::glEnable(GL_BLEND)); glsafe(::glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA)); glsafe(::glCullFace(GL_BACK)); glsafe(::glEnableClientState(GL_VERTEX_ARRAY)); glsafe(::glEnableClientState(GL_NORMAL_ARRAY)); GLint current_program_id; glsafe(::glGetIntegerv(GL_CURRENT_PROGRAM, ¤t_program_id)); GLint color_id = (current_program_id > 0) ? ::glGetUniformLocation(current_program_id, "uniform_color") : -1; glcheck(); m_volume.render(color_id, -1, -1); glsafe(::glBindBuffer(GL_ARRAY_BUFFER, 0)); glsafe(::glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0)); glsafe(::glDisableClientState(GL_VERTEX_ARRAY)); glsafe(::glDisableClientState(GL_NORMAL_ARRAY)); glsafe(::glDisable(GL_BLEND)); } bool GLArrow::on_init() { Pointf3s vertices; std::vector triangles; // bottom face vertices.emplace_back(0.5, 0.0, -0.1); vertices.emplace_back(0.5, 2.0, -0.1); vertices.emplace_back(1.0, 2.0, -0.1); vertices.emplace_back(0.0, 3.0, -0.1); vertices.emplace_back(-1.0, 2.0, -0.1); vertices.emplace_back(-0.5, 2.0, -0.1); vertices.emplace_back(-0.5, 0.0, -0.1); // top face vertices.emplace_back(0.5, 0.0, 0.1); vertices.emplace_back(0.5, 2.0, 0.1); vertices.emplace_back(1.0, 2.0, 0.1); vertices.emplace_back(0.0, 3.0, 0.1); vertices.emplace_back(-1.0, 2.0, 0.1); vertices.emplace_back(-0.5, 2.0, 0.1); vertices.emplace_back(-0.5, 0.0, 0.1); // bottom face triangles.emplace_back(0, 6, 1); triangles.emplace_back(6, 5, 1); triangles.emplace_back(5, 4, 3); triangles.emplace_back(5, 3, 1); triangles.emplace_back(1, 3, 2); // top face triangles.emplace_back(7, 8, 13); triangles.emplace_back(13, 8, 12); triangles.emplace_back(12, 10, 11); triangles.emplace_back(8, 10, 12); triangles.emplace_back(8, 9, 10); // side face triangles.emplace_back(0, 1, 8); triangles.emplace_back(8, 7, 0); triangles.emplace_back(1, 2, 9); triangles.emplace_back(9, 8, 1); triangles.emplace_back(2, 3, 10); triangles.emplace_back(10, 9, 2); triangles.emplace_back(3, 4, 11); triangles.emplace_back(11, 10, 3); triangles.emplace_back(4, 5, 12); triangles.emplace_back(12, 11, 4); triangles.emplace_back(5, 6, 13); triangles.emplace_back(13, 12, 5); triangles.emplace_back(6, 0, 7); triangles.emplace_back(7, 13, 6); m_volume.indexed_vertex_array.load_mesh(TriangleMesh(vertices, triangles)); m_volume.indexed_vertex_array.finalize_geometry(true); return true; } GLCurvedArrow::GLCurvedArrow(unsigned int resolution) : GLModel() , m_resolution(resolution) { if (m_resolution == 0) m_resolution = 1; } bool GLCurvedArrow::on_init() { Pointf3s vertices; std::vector triangles; double ext_radius = 2.5; double int_radius = 1.5; double step = 0.5 * (double)PI / (double)m_resolution; unsigned int vertices_per_level = 4 + 2 * m_resolution; // bottom face vertices.emplace_back(0.0, 1.5, -0.1); vertices.emplace_back(0.0, 1.0, -0.1); vertices.emplace_back(-1.0, 2.0, -0.1); vertices.emplace_back(0.0, 3.0, -0.1); vertices.emplace_back(0.0, 2.5, -0.1); for (unsigned int i = 1; i <= m_resolution; ++i) { double angle = (double)i * step; double x = ext_radius * ::sin(angle); double y = ext_radius * ::cos(angle); vertices.emplace_back(x, y, -0.1); } for (unsigned int i = 0; i < m_resolution; ++i) { double angle = (double)i * step; double x = int_radius * ::cos(angle); double y = int_radius * ::sin(angle); vertices.emplace_back(x, y, -0.1); } // top face vertices.emplace_back(0.0, 1.5, 0.1); vertices.emplace_back(0.0, 1.0, 0.1); vertices.emplace_back(-1.0, 2.0, 0.1); vertices.emplace_back(0.0, 3.0, 0.1); vertices.emplace_back(0.0, 2.5, 0.1); for (unsigned int i = 1; i <= m_resolution; ++i) { double angle = (double)i * step; double x = ext_radius * ::sin(angle); double y = ext_radius * ::cos(angle); vertices.emplace_back(x, y, 0.1); } for (unsigned int i = 0; i < m_resolution; ++i) { double angle = (double)i * step; double x = int_radius * ::cos(angle); double y = int_radius * ::sin(angle); vertices.emplace_back(x, y, 0.1); } // bottom face triangles.emplace_back(0, 1, 2); triangles.emplace_back(0, 2, 4); triangles.emplace_back(4, 2, 3); int first_id = 4; int last_id = (int)vertices_per_level; triangles.emplace_back(last_id, 0, first_id); triangles.emplace_back(last_id, first_id, first_id + 1); for (unsigned int i = 1; i < m_resolution; ++i) { triangles.emplace_back(last_id - i, last_id - i + 1, first_id + i); triangles.emplace_back(last_id - i, first_id + i, first_id + i + 1); } // top face last_id += 1; triangles.emplace_back(last_id + 0, last_id + 2, last_id + 1); triangles.emplace_back(last_id + 0, last_id + 4, last_id + 2); triangles.emplace_back(last_id + 4, last_id + 3, last_id + 2); first_id = last_id + 4; last_id = last_id + 4 + 2 * (int)m_resolution; triangles.emplace_back(last_id, first_id, (int)vertices_per_level + 1); triangles.emplace_back(last_id, first_id + 1, first_id); for (unsigned int i = 1; i < m_resolution; ++i) { triangles.emplace_back(last_id - i, first_id + i, last_id - i + 1); triangles.emplace_back(last_id - i, first_id + i + 1, first_id + i); } // side face for (unsigned int i = 0; i < 4 + 2 * (unsigned int)m_resolution; ++i) { triangles.emplace_back(i, vertices_per_level + 2 + i, i + 1); triangles.emplace_back(i, vertices_per_level + 1 + i, vertices_per_level + 2 + i); } triangles.emplace_back(vertices_per_level, vertices_per_level + 1, 0); triangles.emplace_back(vertices_per_level, 2 * vertices_per_level + 1, vertices_per_level + 1); m_volume.indexed_vertex_array.load_mesh(TriangleMesh(vertices, triangles)); m_volume.indexed_vertex_array.finalize_geometry(true); return true; } bool GLBed::on_init_from_file(const std::string& filename) { reset(); if (!boost::filesystem::exists(filename)) return false; if (!boost::algorithm::iends_with(filename, ".stl")) return false; Model model; try { model = Model::read_from_file(filename); } catch (std::exception & /* ex */) { return false; } m_filename = filename; m_volume.indexed_vertex_array.load_mesh(model.mesh()); m_volume.indexed_vertex_array.finalize_geometry(true); float color[4] = { 0.235f, 0.235f, 0.235f, 1.0f }; set_color(color, 4); return true; } std::string _3DScene::get_gl_info(bool format_as_html, bool extensions) { return Slic3r::GUI::GLCanvas3DManager::get_gl_info().to_string(format_as_html, extensions); } bool _3DScene::add_canvas(wxGLCanvas* canvas, GUI::Bed3D& bed, GUI::Camera& camera, GUI::GLToolbar& view_toolbar) { return s_canvas_mgr.add(canvas, bed, camera, view_toolbar); } bool _3DScene::remove_canvas(wxGLCanvas* canvas) { return s_canvas_mgr.remove(canvas); } void _3DScene::remove_all_canvases() { s_canvas_mgr.remove_all(); } bool _3DScene::init(wxGLCanvas* canvas) { return s_canvas_mgr.init(canvas); } void _3DScene::destroy() { s_canvas_mgr.destroy(); } GUI::GLCanvas3D* _3DScene::get_canvas(wxGLCanvas* canvas) { return s_canvas_mgr.get_canvas(canvas); } } // namespace Slic3r