BlockArray is a package for working with arrays logically partitioned into blocks (or subarrays) in a nested format (also called nested arrays, nested matrices, block matrices, etc.). The main object is a BlockArray which can be created and indexed as shown:
from blockarray.blockarray import BlockArray
from blockarray.linalg import mult_mat_vec
# model the block vector
# [x0, x1]
# where,
x0 = np.array([1, 2, 3])
x1 = np.array([4, 5])
x = BlockArray([x0, x1])
# model the block matrix
# [[A00, A01],
# [A10, A11]]
# where
A00 = np.array(
[[1, 2, 3],
[4, 5, 6],
[7, 8, 9]])
A01 = np.array(
[[1, 2],
[3, 4],
[5, 6]])
A10 = np.array(
[[1, 2, 3],
[4, 5, 6]])
A11 = np.array(
[[1, 2],
[3, 4]])
A = BlockArray([[A00, A01], [A10, A11]])
## Indexing
# Select the (0,) subarray. This is equal to x0
x[0]
# Select the (0, 0) subarray. This is equal to A00
A[0, 0]
# Select the (0, 0) subarray as a BlockArray with 1 subarray
A[0:1, 0:1]
# Select the (0, 0) subarray as a BlockArray with 1 subarray
# Represents the block matrix
# [[A00]]
A[0:1, 0:1]
# Select the upper row of the BlockArray (2 subarrays)
# Represents the block matrix
# [[A00, A01]]
A[0:1, :]
# Select column zero of the BlockArray (2 subarrays)
# Represents the block matrix
# [[A00],
# [A10]]
A[:, 0:1] In addition, basic math operations can be applied on BlockArray arrays (addition, subtraction, multiplication, etc.) as well as most numpy.ufunc operations.
## Basic math operations
# Basic math operations (add, sub, scalar mul, etc) are defined
y = mult_mat_vec(2*A, x)
z = x+y
# There is limited support for numpy ufuncs
y = np.matmul(A, x)Documentation is available but incomplete. To build the documentation, go to the docs directory and use the command make html to build the html documentation in the build directory.
This project is a work in progress so there are likely bugs and missing features. If you would like to contribute a bug fix, a feature, refactor etc. thank you! All contributions are welcome.
There are similar projects that provide block matrix and block vector functionality such as the excellent FEniCS project and its associated projects (https://github.com/FEniCS, https://fenicsproject.org/, https://bitbucket.org/fenics-apps/cbc.block). The PETSc project also provides a nested matrix and vector format (https://petsc.org/release/#). This package provides a more generic block array to facilititate higher dimensional arrays.