https://github.com/juliaaplavin/unioncollections.jl

https://github.com/juliaaplavin/unioncollections.jl

arrays data-structures dictionaries

Last synced: 2 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/juliaaplavin/unioncollections.jl
• Owner: JuliaAPlavin
• License: mit
• Created: 2024-02-18T18:00:26.000Z (11 months ago)
• Default Branch: master
• Last Pushed: 2024-10-17T12:05:24.000Z (3 months ago)
• Last Synced: 2024-11-19T05:59:52.420Z (2 months ago)
• Topics: arrays, data-structures, dictionaries
• Language: Julia
• Homepage:
• Size: 15.6 KB
• Stars: 5
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# UnionCollections.jl

Efficient collections (arrays and dictionaries) with `Union` element types. Compared to `Vector{Union{...}}`, more operations are type-stable: most notably, `map()`.

Under the hood, elements of different types are stored in separate sub-collections. This allows for operations on the collection as a whole to be type-stable, while `X[i]` fundamentally remain type-unstable.

# Usage

```julia

# create a regular vector and a union vector:

julia> V = Union{Int, String}[1, 2, "x", 3, "yy"]

5-element Vector{Union{Int64, String}} <...>

julia> A = unioncollection(V)

5-element UnionVector{Union{Int64, String}} <...>

# these two work basically the same, but the union vector is more efficient

## Vector:

julia> map(x -> x^3, V)

5-element Vector{Any}:

  1

  8

   "xxx"

 27

   "yyyyyy"

julia> @code_warntype map(x -> x^3, V)

Body::Union{Vector{Any}, Vector{Int64}, Vector{String}}

julia> @btime map(x -> x^2, map(x -> x^3, $V))

  3.445 μs (19 allocations: 792 bytes)

## UnionVector:

julia> map(x -> x^3, A)

5-element UnionVector{Union{Int64, String}, Tuple{Vector{Int64}, Vector{String}}}:

  1

  8

   "xxx"

 27

   "yyyyyy"

julia> @code_warntype map(x -> x^3, A)

Body::UnionVector{Union{Int64, String}}

julia> @btime map(x -> x^2, map(x -> x^3, $A))

  197.597 ns (8 allocations: 392 bytes)

```