Get optimal chunking size for an array. — optimal

This function will determine the optimal chunking sizes of the array dimensions based on weights per dimension.

Usage

optimal_chunking(dim_sizes, weights, chunk_values = 4L * 1024L * 1024L)

Arguments

dim_sizes: Integer array of dimension lengths, corresponding to the shape of the array.
weights: Optional, numeric vector with weights per dimension, in the same order as dim_sizes. If omitted, each dimension will have a weight of 1L, i.e. no preferential chunking on any dimension.
chunk_values: Optional, integer value given the maximum number of array elements per chunk. Default is 4 million, meaning that the chunk size of float32 data is at most 16MB uncompressed.

Value

An integer vector with chunk length per dimension in the same order as argument dim_sizes.

Examples

shape <- c(x = 50000L, y = 350L, time = 8192)

# Default chunking, approaching the maximum chunk size
optimal_chunking(dim_sizes = shape)
#> [1] 161 175 161

# Prioritize extractions over the "time" dimension
optimal_chunking(dim_sizes = shape, weights = c(1, 1, 2))
#> [1]   46   44 2048