You are viewing the documentation for Deformable v0.4.1, an older version. Read the latest, v0.9.0

Python API Reference

Render surfaces

Embed a render surface in a tetrahedral mesh, and deform it with the simulation every frame.

On this page

A render surface is the mesh a host draws. embed_surface prepares it once against the rest mesh of an object, and the SurfaceEmbedding it returns deforms it with the simulated nodes every frame. A breakable surface gets inner faces, which lie on tetrahedron faces inside the object and stay hidden until a crack exposes them. With a splinter cage it takes the splinter path instead: it is cut into bricks, one per cell of the cage, and caps close the cuts.

ClipMode

Enum · deformable.ClipMode

python
class ClipMode(enum.IntEnum):
    ASSIGN_ONLY = 0
    CLIP_WITH_INNER_FACES = 2
    TETRAHEDRAL_MESH_BOUNDARY = 3

How embed_surface prepares a render surface against the rest mesh: whether it cuts the surface along the tetrahedron faces, and which inner faces it adds.

An inner face lies on a tetrahedron face inside the object, hidden until a crack exposes it. With a splinter cage (PreparationParameters.splinter_cage), the two modes other than ASSIGN_ONLY take the splinter path instead. The values are the library’s codes.

Values

  • ASSIGN_ONLY (0)Each vertex follows the tetrahedron it lies in, or the nearest one by its weights when it lies outside the mesh; nothing is cut and no inner face is made. For objects that bend but never break.
  • CLIP_WITH_INNER_FACES (2)Breakable: the surface is cut along the tetrahedron faces, and the part of each interior tetrahedron face inside the surface becomes an inner face. A polygon wholly outside the mesh is kept whole on the nearest tetrahedron by its weights.
  • TETRAHEDRAL_MESH_BOUNDARY (3)The tetrahedral mesh is the surface: each polygon follows the tetrahedron its centre lies in, uncut, and every interior tetrahedron face becomes an inner face. For showing the tetrahedrons themselves.

See also Guide: Choose how to prepare it · C DeformableClipMode · Rust ClipMode

PreparationParameters

Data class · deformable.PreparationParameters

python
@dataclasses.dataclass(frozen=True)
class PreparationParameters:
    clip_mode: ClipMode | int
    uv_scale: float
    splinter_cage: PolygonMesh | None = None

The settings of embed_surface: how it cuts the surface, the scale of the planar UVs it makes, and an optional splinter cage.

Fields

  • clip_mode (ClipMode or int)How the surface is cut, and which inner faces it gets. An int that is no member’s value makes embed_surface raise DeformableError with the status INVALID_ARGUMENT, and one that does not fit in 32 bits ValueError.
  • uv_scale (float)Scale of the planar UVs of inner and cap faces; the random offsets of inner faces are not scaled. 1 is a good start.
  • splinter_cage (PolygonMesh or None)The splinter cage in the surface’s space, already moved by the host’s splinter transform, or None. With vertices and a clip mode other than ASSIGN_ONLY, the surface takes the splinter path: it is cut into one brick per cage cell, and the cuts are capped. When no cell holds a part of the surface, the surface is prepared as without a cage. ASSIGN_ONLY ignores the cage, but its polygons are still checked. Default None.

Example

python
# A surface that bends with the object but never breaks.
parameters = deformable.PreparationParameters(
    deformable.ClipMode.ASSIGN_ONLY, uv_scale=1.0
)
surface = deformable.SurfaceMesh(polygons)
with deformable.embed_surface(mesh, surface, parameters) as embedding:
    len(embedding.arrays().surface.mesh.positions)
# 8

See also Guide: Choose how to prepare it · C DeformablePreparationParameters · Rust PreparationParameters

SkinningOptions

Data class · deformable.SkinningOptions

python
@dataclasses.dataclass(frozen=True)
class SkinningOptions:
    show_inner_faces: bool = False
    start_splintered: bool = False

The per-frame skinning options: which hidden faces SurfaceEmbedding.deform shows. SkinningOptions() turns both off, as deform does when it gets None.

Fields

  • show_inner_faces (bool)Shows every inner face and cap, exposed or not, so no face is hidden: useful to inspect the preparation. Default False.
  • start_splintered (bool)With splinter data, shows the caps from the start instead of once a tetrahedron of theirs is flagged, as for an object that starts out split. Default False.

Example

python
clip = deformable.ClipMode.CLIP_WITH_INNER_FACES
parameters = deformable.PreparationParameters(clip, uv_scale=1.0)
surface = deformable.SurfaceMesh(polygons)
embedding = deformable.embed_surface(mesh, surface, parameters)
states = scene.tetrahedron_states(object)
# Every inner face shown, to inspect the preparation.
options = deformable.SkinningOptions(show_inner_faces=True)
deformed = embedding.deform(states, deformable.IDENTITY, options)
sum(deformed.hidden_faces)
# 0

See also Guide: Skin it every frame · C DeformableSkinningOptions · Rust SkinningOptions

DeformedSurface

Data class · deformable.DeformedSurface

python
@dataclasses.dataclass(frozen=True)
class DeformedSurface:
    positions: array.array
    normals: array.array
    hidden_faces: array.array

A skinned surface in the object space of the host’s shape: 32-bit float positions per output vertex, unit 32-bit float normals per corner (empty when the surface has none) and the faces to hide.

SurfaceEmbedding.deform returns a new one every call. Its fields are flat arrays a host hands to its mesh API as they are: Blender’s foreach_set takes them without a per-item conversion.

Fields

  • positions (array.array)x, y and z of the deformed position of each output vertex, mapped by the inverse world transform, flat 32-bit floats (array("f")).
  • normals (array.array)x, y and z of the deformed normal of each corner, embed(p + n) - embed(p) mapped by the linear part of the inverse world transform and made unit length, a zero vector left zero, flat 32-bit floats (array("f")); empty when the surface has no normals.
  • hidden_faces (array.array)1 for each face the host hides (an inner face or cap not shown yet), 0 for the others (array("b")).

See also Guide: Skin it every frame · C DeformableDeformedSurface · C DeformableDeformedSurfaceView · Rust DeformedSurface

SurfaceEmbedding

Class · deformable.SurfaceEmbedding

python
class SurfaceEmbedding

A prepared surface in rest space, made by embed_surface: the output surface, the tetrahedron and barycentric weights each output vertex follows, the flag word of each corner, the weights of the corner normals, and on the splinter path each vertex’s cell tetrahedron.

An input vertex becomes one output vertex for each flag word and tetrahedron among its corners, so the surface can open along the cracks. A vertex outside the rest mesh follows the nearest tetrahedron by its weights (the one whose smallest barycentric weight is largest), with weights that extrapolate. Reading and skinning never change the embedding, so several threads may use one embedding at the same time.

It works as a context manager: a with block calls close() when it ends.

See also Guide: Read the prepared surface · C DeformableSurfaceEmbedding · Rust SurfaceEmbedding

SurfaceEmbedding.arrays

Method · deformable.SurfaceEmbedding.arrays

python
def arrays(self) -> SurfaceEmbeddingArrays

A copy of the embedding’s arrays: the prepared surface in rest space, the flag word of each corner, the source face of each face, and the tetrahedron and weights each vertex follows. Later calls leave the copy as it is.

Raises

ValueError when the embedding is closed.

See also C deformable_surface_embedding_view

SurfaceEmbedding.deform

Method · deformable.SurfaceEmbedding.deform

python
def deform(
    self,
    states: Sequence[TetrahedronState],
    world_transform: AffineTransform,
    options: SkinningOptions | None = None,
) -> DeformedSurface

Skins the surface on the current tetrahedron states, in the object space of the host’s shape: positions and unit normals in 32-bit floats, mapped by the inverse of world_transform, and the faces to hide.

Each output vertex follows its tetrahedron by its weights. On the splinter path it switches to its cell tetrahedron once its cell has broken loose: once that tetrahedron or its own has a state flag word other than 0, or the two sit in different islands. Each corner normal is mapped by the linear part of the inverse, then made unit length; a zero one stays zero. A face is hidden while it is an inner face whose tetrahedron face no fracture has exposed, or a cap whose brick tetrahedron and the own tetrahedron of its first corner both still have a state flag word of 0, unless options say to show it. The states are not checked: a NaN or infinite corner gives NaN or infinite positions. Skinning never changes the embedding, so several threads may skin one embedding at the same time.

Parameters

Returns

DeformedSurface: The skinned surface, as new arrays.

Raises

DeformableError with one of these statuses:

  • LENGTH_MISMATCHstates does not hold one state per tetrahedron of the rest mesh.
  • NON_FINITE_VALUEAn entry of world_transform is NaN or infinite.
  • SINGULAR_TRANSFORMworld_transform has no inverse: the determinant of its linear part is 0, or the inverse overflows.

ValueError when the embedding is closed, the linear part of world_transform does not hold three rows of three numbers, or its translation or a corner of a state does not hold three numbers.

Example

python
clip = deformable.ClipMode.ASSIGN_ONLY
parameters = deformable.PreparationParameters(clip, uv_scale=1.0)
surface = deformable.SurfaceMesh(polygons)
embedding = deformable.embed_surface(mesh, surface, parameters)
# Every frame, after the step:
states = scene.tetrahedron_states(object)
deformed = embedding.deform(states, deformable.IDENTITY)
len(deformed.positions) // 3, len(deformed.hidden_faces)
# (8, 6)

See also Guide: Skin it every frame · C deformable_surface_embedding_deform · Rust SurfaceEmbedding::deform

SurfaceEmbedding.closed

Property · deformable.SurfaceEmbedding.closed

python
@property
def closed(self) -> bool

True once the handle has been destroyed.

SurfaceEmbedding.close

Method · deformable.SurfaceEmbedding.close

python
def close(self) -> None

Destroys the handle once the calls in flight on it end; later calls on the object raise ValueError. Closing twice does nothing.

See also C deformable_surface_embedding_destroy

SurfaceEmbeddingArrays

Data class · deformable.SurfaceEmbeddingArrays

python
@dataclasses.dataclass(frozen=True)
class SurfaceEmbeddingArrays:
    surface: SurfaceMesh
    corner_flags: list[int]
    source_faces: list[int]
    vertex_tetrahedrons: list[int]
    vertex_weights: list[tuple[float, float, float, float]]
    normal_weights: list[tuple[float, float, float, float]]
    brick_tetrahedrons: list[int]
    vertex_cell_tetrahedrons: list[int]

A copy of the arrays of a surface embedding, as SurfaceEmbedding.arrays gives it.

Fields

  • surface (SurfaceMesh)The prepared surface in rest space: output vertices, polygons and per-corner channels. It can have more vertices than the input, as a vertex is split where a crack can open; with ClipMode.ASSIGN_ONLY it keeps one vertex per input vertex a polygon uses. Build the render mesh from it once.
  • corner_flags (list[int])The flag word of each corner of surface, in corner order: bits 0 to 28 the controlling tetrahedron, bits 29 and 30 the tetrahedron face of an inner face, bit 31 set on the corners of inner faces. On the splinter path, the tetrahedron is that of the corner’s brick, the face bits are not set, and bit 31 marks the corners of caps.
  • source_faces (list[int])The input face each output face was cut from, in face order; -1 for inner faces and caps.
  • vertex_tetrahedrons (list[int])The controlling tetrahedron of each output vertex, the one it follows.
  • vertex_weights (list[tuple[float, float, float, float]])The barycentric weights of each output vertex in its tetrahedron; some are negative for a vertex outside it.
  • normal_weights (list[tuple[float, float, float, float]])The barycentric weights of p + n for each corner, the tip of its normal n at its position p, in its vertex’s tetrahedron; empty when the surface has no normals.
  • brick_tetrahedrons (list[int])The controlling tetrahedron of each brick of the splinter path, one brick per cage cell that holds a part of the surface; empty without splinter data.
  • vertex_cell_tetrahedrons (list[int])The cell tetrahedron of each output vertex on the splinter path, the one it follows once its cell has broken loose; empty without splinter data.

See also Guide: Read the prepared surface · C DeformableSurfaceEmbeddingView

embed_surface

Function · deformable.embed_surface

python
def embed_surface(
    rest: TetrahedralMesh,
    surface: SurfaceMesh,
    parameters: PreparationParameters,
) -> SurfaceEmbedding

Prepares a render surface against the rest mesh of an object: cuts it as the clip mode says, adds inner faces with random planar UVs times uv_scale, takes the splinter path when a splinter cage is given, then embeds each output vertex in a tetrahedron by barycentric weights.

The prepared surface keeps every channel the input has, with opaque white as the colour of inner faces. With ClipMode.CLIP_WITH_INNER_FACES and on the splinter path, a polygon of fewer than 3 corners is left out. The random stream that turns and offsets the UVs of inner faces has a fixed seed, so equal input gives equal UVs.

Parameters

  • rest (TetrahedralMesh)The object’s tetrahedral mesh at rest, in the space of surface: the mesh given to Scene.add_object, or that mesh before the host moved it to world space, with the same tetrahedrons in the same order.
  • surface (SurfaceMesh)The render surface: its polygons, and per-corner normals, UV sets and colours where it has them.
  • parameters (PreparationParameters)How the surface is cut, the scale of the UVs the call makes, and an optional splinter cage.

Returns

SurfaceEmbedding: The prepared surface with its embedding, to skin every frame with SurfaceEmbedding.deform.

Raises

DeformableError with one of these statuses:

  • EMPTY_MESHrest has no tetrahedrons.
  • INDEX_OUT_OF_RANGEA tetrahedron of rest names a node past the end of its positions, rest has more tetrahedrons than a corner flag word addresses (2 to the power 29), or a corner of surface or of the splinter cage names a vertex past the end of its positions.
  • LENGTH_MISMATCHThe polygon sizes of surface or of the splinter cage do not add up to its corner count, the normals or the colours of surface hold neither 0 entries nor one per corner, or surface has more than 2,147,483,647 polygons.
  • NON_FINITE_VALUEA position of rest, surface or the splinter cage, a normal, UV or colour of surface, or uv_scale is NaN or infinite.
  • NON_MANIFOLD_FACEThree or more tetrahedrons of rest share a face.
  • INVALID_SPLINTER_CAGEA polygon of the splinter cage has fewer than 3 vertices, or on the splinter path, a brick has 3 faces or fewer.

ValueError when the positions of rest, surface or the splinter cage, or the normals of surface, do not form whole groups of three numbers, the tetrahedrons of rest whole groups of four, a UV set does not hold one coordinate pair per corner, the colours do not form whole groups of four numbers, an index or polygon size is negative or does not fit in 32 bits, or clip_mode does not fit in 32 bits.

TypeError when clip_mode is not an int.

Example

python
clip = deformable.ClipMode.CLIP_WITH_INNER_FACES
parameters = deformable.PreparationParameters(clip, uv_scale=1.0)
surface = deformable.SurfaceMesh(polygons)
with deformable.embed_surface(mesh, surface, parameters) as embedding:
    arrays = embedding.arrays()
-1 in arrays.source_faces  # inner faces
# True

See also Guide: Choose how to prepare it · C deformable_embed_surface · Rust embed_surface