PartitionerRegistry

Registry and manager for Femora mesh partitioners.

This registry maintains a collection of partitioner algorithms (both built-in and optional dependency-based, like METIS) that map a PyVista UnstructuredGrid's mesh cells to integer partition IDs in [0, num_partitions - 1].

Attributes:

Name	Type	Description
`_partitioners`	`Dict[str, _RegisteredPartitioner]`	Internal registry map storing registered partitioner functions and their availability checks.

Example

import pyvista as pv
from femora.components.partitioner.partitioner import PartitionerRegistry

# Construct a simple VTK grid and partition its cells
grid = pv.UnstructuredGrid(...)
labels = PartitionerRegistry.partition(
    grid,
    num_partitions=4,
    partitioner="hilbert",
)

Methods:

register `staticmethod`

register(name: str, func: Callable[..., ndarray], *, is_available: Optional[Callable[[], bool]] = None) -> None

Register a custom partitioner algorithm in Femora.

Parameters:

Name	Type	Description	Default
`name`	`str`	Unique string identifier for the partitioner (case-insensitive).	required
`func`	`Callable[..., ndarray]`	A callable of type PartitionerFunc mapping a PyVista UnstructuredGrid and partition count to a 1D NumPy array of cell partition IDs.	required
`is_available`	`Optional[Callable[[], bool]]`	Optional callback returning True if the partitioner's system dependencies are met. Defaults to None (always available).	`None`

Raises:

Type	Description
`ValueError`	If `name` is empty or whitespace-only.

get_available_types `staticmethod`

get_available_types(*, include_unavailable: bool = False) -> List[str]

Get the names of all registered partitioner algorithms.

Parameters:

Name	Type	Description	Default
`include_unavailable`	`bool`	If True, returns all registered algorithms even if their external library dependencies (like pymetis) are missing. Defaults to False.	`False`

Returns:

Type	Description
`List[str]`	List[str]: A sorted list of registered partitioner name keys.

validate `staticmethod`

validate(name: str) -> None

Validate if a partitioner name is registered and available.

Parameters:

Name	Type	Description	Default
`name`	`str`	The case-insensitive name of the partitioner to validate.	required

Raises:

Type	Description
`ValueError`	If the partitioner name is not registered.
`ImportError`	If the partitioner requires missing external dependencies.

partition `staticmethod`

partition(mesh: UnstructuredGrid, num_partitions: int, *, partitioner: str = 'kd-tree', **kwargs: Any) -> np.ndarray

Partition a mesh's cells into a specified number of balanced domains.

Parameters:

Name	Type	Description	Default
`mesh`	`UnstructuredGrid`	The PyVista UnstructuredGrid representing the global assembled mesh.	required
`num_partitions`	`int`	Target number of partitions/cores (must be >= 1).	required
`partitioner`	`str`	Registered case-insensitive name of the partitioning algorithm to use. Defaults to "kd-tree".	`'kd-tree'`
`**kwargs`	`Any`	Extra parameters passed to the underlying partitioner function.	`{}`

Returns:

Type	Description
`ndarray`	np.ndarray: A 1D integer NumPy array of shape `(mesh.n_cells,)` mapping each cell index to a partition ID in `[0, num_partitions - 1]`.

Raises:

Type	Description
`ValueError`	If `num_partitions` is less than 1.
`PartitionerError`	If the partitioning function fails, returns an incorrect array shape, or produces out-of-bound partition IDs.

PartitionerRegistry