Array Interfaces in Julia

An array interface is a syntactical contract for arrays in Julia that they must follow. This article describes the various methods which can be adopted to construct array interfaces in Julia. It also explains how to perform indexing in an array interface to access its elements.

An array interface can be constructed by taking a standard array and linking custom keys along with the indices of each dimension. This can be achieved by using the AxisIndices package.

Construction of an Array Using AxisArray Function:

An array can be created using the AxisArray() function present in the AxisIndices package of Julia. 

Example: 

Julia

# Julia program to declare an array interface

# Importing AxisIndices package
using AxisIndices

# Declare array
arr = [5 6; 7 8];

# Using AxisArray function to return an array
# with its dimension names and axis values
axis = AxisArray(arr, ["row1", "row2"],
                      [:col1, :col2])

Output:

An array interface can also be initialized using the following syntax:

Syntax:

Array{T}(undef, dims)

Example:

Julia

# Julia program to declare an array interface

# Importing AxisIndices package
using AxisIndices

# Declare array
arr = [5 6; 7 8];

# Using AxisArray function to return an array
# with its dimension names and axis values
axis = AxisArray{Int}(undef, ["row1", "row2"],
                             [:col1, :col2]);

axis[:, :] = arr;

axis

Output:

Construction of an Array Using NamedAxisArray() Function:

Each Dimension or axis of the array interface can be given names using NamedAxisArray present in the AxisIndices package of Julia.

Example:

Julia

# Julia program to declare an array interface

# Importing AxisIndices package
using AxisIndices

# Declare array
arr = [5 6; 7 8];

# Using AxisArray function to return an array
# with its dimension names and axis values
axis = AxisArray(arr, ["row1", "row2"],
                      [:col1, :col2])

# Using NamedAxisArray function
# to return an array with named axes
axis_named = NamedAxisArray{(:xdim, :ydim)}(axis)

Output:

Construction of an Array Using MetaAxisArray() Function:

Metadata can also be linked with the array interface.

Example:

Julia

# Julia program to declare an array interface

# Importing AxisIndices package
using AxisIndices

# Declare array
arr = [5 6; 7 8];

# Using MetaAxisArray function to return
# an array with its dimension names,
# axis values, metadata
axis_meta = MetaAxisArray(arr, (["row1", "row2"],
                                [:col1, :col2]),
                          metadata = "Array Interface")

Output:

Indexing of Arrays

Indexing in an Array Interface Using Unitful Package:

An array interface can be indexed just like standard arrays in Julia. This can be achieved by using the Unitful package.

Example:

Julia

# Julia program to perform indexing
# in an array interface

# Importing AxisIndices package
using AxisIndices

# Importing Unitful package
import Unitful: m

# Using reshape function
# to return an array with
# its specified dimensions
arr = reshape(1:6, 2, 3);

# Using AxisArray function to return
# an array with its dimension names
# and axis values
axis = AxisArray(arr, ((1:2)m,
                ["col1", "col2", "col3"]))

Output:

Example:

Julia

# Julia program to perform indexing
# in an array interface

# Importing AxisIndices package
using AxisIndices

# Importing Unitful package
import Unitful: m

# Using reshape function
# to return an array
# with its specified dimensions
arr = reshape(1:6, 2, 3);

# Using AxisArray function
# to return an array
# with its dimension names and axis values
axis = AxisArray(arr, ((1:2)m,
                ["col1", "col2", "col3"]))

# displaying only the first row
# and all the columns
axis[1,:]

Output:

Example:

Julia

# Julia program to perform indexing
# in an array interface

# Importing AxisIndices package
using AxisIndices

# Importing Unitful package
import Unitful: m

# Using reshape function to return
# an array with its specified dimensions
arr = reshape(1:6, 2, 3);

# Using AxisArray function to return
# an array with its dimension names
# and axis values
axis = AxisArray(arr, ((1:2)m,
                ["col1", "col2", "col3"]))

# displaying the first two rows
# and the first two columns
axis[1:2, 1:2]

Output:

Example:

Julia

# Julia program to perform indexing
# in an array interface

# Importing AxisIndices package
using AxisIndices

# Importing Unitful package
import Unitful: m

# Using reshape function to return
# an array with its specified dimensions
arr = reshape(1:6, 2, 3);

# Using AxisArray function to return
# an array with its dimension names and axis values
axis = AxisArray(arr, ((1:2)m,
                ["col1", "col2", "col3"]))

# displaying the first 3 elements of the array
axis[1:3]

Output:

Indexing in an Array Interface Using Keys:

An Array Interface can be indexed with the help of keys.

Example :

Julia

# Julia program to perform indexing
# in an array interface

# Importing AxisIndices package
using AxisIndices

# Importing Unitful package
import Unitful: m

# Using reshape function to return
# an array with its specified dimensions
arr = reshape(1:6, 2, 3);

# Using AxisArray function to return
# an array with its dimension names and axis values
axis = AxisArray(arr, ((1:2)m,
                ["col1", "col2", "col3"]))

# accessing an element using its specified keys
axis[1m, "col1"]

Output:

Indexing in an Array Interface Using Filtration Functions: 

An array interface can also be indexed using functions that filter the keys.

Example:

Julia

# Julia program to perform indexing
# in an array interface

# Importing AxisIndices package
using AxisIndices

# Importing Unitful package
import Unitful: m

# Using reshape function to return
# an array with its specified dimensions
arr = reshape(1:6, 2, 3);

# Using AxisArray function to return
# an array with its dimension names and axis values
axis = AxisArray(arr, ((1:2)m,
                ["col1", "col2", "col3"]))

# Applying filters to the keys
axis[!=(1m), in(["col1", "col2"])]

Output: