Good questions, thanks for asking! Here are a few examples:

>>> import graphblas as gb
>>> import numpy as np

>>> A = gb.Matrix.from_dense(np.arange(9).reshape(3, 3))
>>> A
"M_0"      nvals  nrows  ncols  dtype  format
gb.Matrix      9      3      3  INT64   fullr
---------------------------------------------
   0  1  2
0  0  1  2
1  3  4  5
2  6  7  8

>>> # Diagonal of a Matrix as a Vector
>>> A.diag()
"v_0"      nvals  size  dtype  format
gb.Vector      3     3  INT64    full
-------------------------------------
index  0  1  2
value  0  4  8

>>> # Off-diagonal of a Matrix as a Vector
>>> A.diag(1)
A"v_1"      nvals  size  dtype  format
gb.Vector      2     2  INT64    full
-------------------------------------
index  0  1
value  1  5

>>> # Select diagonal of a Matrix as a Matrix
>>> A.select("diag")
gb.MatrixExpression                         nrows  ncols  dtype
M_0.select(select.diag[BOOL], thunk=False)      3      3  INT64

"Result"   nvals  nrows  ncols  dtype
gb.Matrix      3      3      3  INT64
-------------------------------------
   0  1  2
0  0
1     4
2        8

Do expr.new() or other << expr to calculate the expression.

>>> # Select off-diagonal of a Matrix as a Matrix (using alternative API)
>>> gb.select.diag(A, 1)
gb.MatrixExpression                     nrows  ncols  dtype
M_0.select(select.diag[BOOL], thunk=1)      3      3  INT64

"Result"   nvals  nrows  ncols  dtype
gb.Matrix      2      3      3  INT64
-------------------------------------
  0  1  2
0    1
1       5
2

Do expr.new() or other << expr to calculate the expression.

>>> # Remove diagonal elements of a Matrix using select
>>> A << gb.select.offdiag(A)
>>> A
"M_0"      nvals  nrows  ncols  dtype   format
gb.Matrix      6      3      3  INT64  bitmapr
----------------------------------------------
   0  1  2
0     1  2
1  3     5
2  6  7

>>> # Assign a Vector to the diagonal of a Matrix
>>> v = gb.Vector.from_dense([1, 2, 3])
>>> v
A(Diag.S) << Diag"v_2"      nvals  size  dtype  format
gb.Vector      3     3  INT64    full
-------------------------------------
index  0  1  2
value  1  2  3

>>> A(gb.op.second) << v.diag()

>>> # or
>>> Diag = v.diag()
>>> A(Diag.S) << Diag
>>> A
"M_0"      nvals  nrows  ncols  dtype  format
gb.Matrix      9      3      3  INT64   fullr
---------------------------------------------
   0  1  2
0  1  1  2
1  3  2  5
2  6  7  3

Does this help? Is there any way we can make this better?

Question: Most efficient way to set or remove the diagonal in large matrices #487

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions