contract

module of dascore.lazy

source

Chunk contracts: what one chunk of an operation’s output needs of its input.

An operation which can run on a lazy array says so by returning a contract from PatchProcessor.chunk_contract, for the input’s dtype and the plan get_metadata worked out. The contract is what lets the work be split into chunks later without changing the result: each output chunk is computed from the input it names, and nothing else. Returning None says the operation cannot run on a lazy array.

This module imports nothing from the rest of DASCore, because the processors in dascore.proc and dascore.transform import it.

Examples

import numpy as np
import dascore as dc
from dascore.lazy.contract import Pointwise, Unchanged

class Double(dc.PatchProcessor):
    '''Double every sample.'''

    name = None  # so this example claims no name of its own

    def chunk_contract(self, dtype, plan):
        return Pointwise()

    def kernel(self, data):
        return data * 2

assert isinstance(Double().chunk_contract(np.dtype("f8"), {}), Pointwise)
assert dc.PatchProcessor().chunk_contract(np.dtype("f8"), {}) is None

Classes

Name	Description
ChunkContract	What one chunk of an operation’s output needs of its input.
Pointwise	An output chunk needs only the input chunk at the same place.
Unchanged	The kernel hands its input back, for this dtype and plan.