Chunker

class PlasmaCalcs.dimensions.maindims_chunking.Chunker(chunks, maindims_sizes, init_slices=None, *, lenient=False)

Bases: object

manages choosing slices based on chunks.

— CAUTION —

derivative results different with/without chunks, unless deriv_before_slice=True.

All other results should be the same with/without chunking.

chunks: dict

keys should be maindim possibly with ‘_size’ or ‘_slices’ appended, e.g. ‘x’ or ‘y_size’.

x –> break x into this many chunks total.

x_size –> break x into chunks of this size.

x_slices –> list of slices to use for x.

maindims_sizes: dict

{dim: full size} for dims in maindims.

init_slices: None or dict

dict of slices which already exist.

un-chunked dims’ slices will be maintained when creating chunk slices.

invalid chunking if any chunk dims appear in slices.

None –> internally store as empty dict.

lenient: bool

whether to allow any ChunkSlicer inputs, even if invalid.

default False, i.e. self.check_valid() (which will raise ChunkError if not valid).

Methods

__len__()	the number of chunk slices which will be generated by self.iter_slicers().
check_valid()	checks that self is a valid ChunkSlicer.
dim_chunk_info(x)	return chunking info for dim x: (mode, value).
iter_slicers()	iterates through all chunk slices, yielding slices dict including any non-chunked slices.
slicers()	list of all slices to use for chunking.
_iter_slicers_x(x)	iterates through all chunk slices for dim x.

Attributes

chunk_dims	the tuple of dims which are being chunked.
maindims	the maindims.

__len__()

the number of chunk slices which will be generated by self.iter_slicers().

[EFF] perhaps should make this more efficient instead of generating all slices…

_iter_slicers_x(x)

iterates through all chunk slices for dim x.

check_valid()

checks that self is a valid ChunkSlicer. crash with ChunkError if not.

Valid must satisfy:

- all chunk dims are maindims

- no overlap between chunks and slices (e.g. ‘x’ in chunks and slices)

property chunk_dims

the tuple of dims which are being chunked.

E.g., if chunks={‘x’:10, ‘y_size’:50}, then chunking_dims=’x’, ‘y’.

Computed as set of self.chunks keys, with ‘_size’ or ‘_slices’ removed if present.

dim_chunk_info(x)

return chunking info for dim x: (mode, value).

mode will be ‘n’, ‘size’, or ‘slices’,

correspond to x, x_size, or x_slices appearing in self.chunks.

value will be int, int, or iterable of slices, respectively.

iter_slicers()

iterates through all chunk slices, yielding slices dict including any non-chunked slices.

E.g. chunks=dict(x_size=50), slices=dict(z=7) –> yield:

dict(x=slice(0,50), z=7),

dict(x=slice(50,100), z=7),

…

property maindims

the maindims. Just the tuple of self.maindims_sizes keys.

slicers()

list of all slices to use for chunking. Equivalent: list(self.iter_slicers()).