https://github.com/perrutquist/triangularindices.jl

https://github.com/perrutquist/triangularindices.jl

Last synced: about 1 month ago
JSON representation

• Host: GitHub
• URL: https://github.com/perrutquist/triangularindices.jl
• Owner: perrutquist
• License: mit
• Created: 2022-06-23T11:29:46.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2022-07-01T06:11:42.000Z (over 2 years ago)
• Last Synced: 2024-11-10T06:28:14.355Z (about 2 months ago)
• Language: Julia
• Size: 20.5 KB
• Stars: 8
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
[![Build status](https://github.com/perrutquist/TriangularIndices.jl/workflows/CI/badge.svg)](https://github.com/perrutquist/TriangularIndices.jl/actions)

# TriangularIndices.jl

This small package is intended to help with iterating over the indices of triangular matrices. For example:

```julia

using TriangularIndices

z = zeros(Int, (3,3))

for (k, (i,j)) in enumerate(UpperTriangularIndices(z))

    z[i,j] = k

end

```

will result in 

```julia

z == [ 1  2  4

       0  3  5

       0  0  6 ]

```

Of course, the above code could just as easily have been written

```julia

z = zeros(Int, (3,3))

k = 1

for j in 1:size(z,1)

    for i in 1:j

        z[i,j] = k

        k += 1

    end

end

```

which would not require the use of TriangularIndices.jl, and which might even be faster for single-threaded code. However, `UpperTriangularIndices` objects support `length` as well as partitioning using `getindex`, which makes them

useful with macros that distribute the iterator over threads or processes, such as `@distributed` from [Distributed](https://docs.julialang.org/en/v1/manual/distributed-computing/), `@threads` from [Threads](https://docs.julialang.org/en/v1/manual/multi-threading/), or `@floop` from [FLoops](https://github.com/JuliaFolds/FLoops.jl).

For reverse indexing, there is a fast specialization of `findfirst`

```julia

findfirst(==((2,3)), UpperTriangularIndices(3)) # returns 5

```

Note: There's currently no corresponding implementation of `LowerTriangularIndices`. If somebody is interested in writing one, the linear-to-cartesian indexing operation could use [this code](https://discourse.julialang.org/t/iterating-over-elements-of-upper-triangular-matrix-but-cartesian-indices-are-needed/65498/3).