SmallVectors

Introduction

A module that defines small (mutable and immutable) vectors with a maximum length. Externally they have a dynamic/runtime length, but internally they are backed by a statically sized vector. This makes it so that operations can be performed faster because they can remain on the stack, but it provides some more convenience compared to StaticArrays.jl where the length is encoded in the type.

Examples

For example:

using NDTensors.SmallVectors

mv = MSmallVector{10}([1, 2, 3]) # Mutable vector with length 3, maximum length 10
push!(mv, 4)
mv[2] = 12
sort!(mv; rev=true)

v = SmallVector{10}([1, 2, 3]) # Immutable vector with length 3, maximum length 10
v = SmallVectors.push(v, 4)
v = SmallVectors.setindex(v, 12, 2)
v = SmallVectors.sort(v; rev=true)

This also has the advantage that you can efficiently store collections of SmallVector/MSmallVector that have different runtime lengths, as long as they have the same maximum length.

List of functionality

SmallVector and MSmallVector are subtypes of AbstractVector and therefore can be used in Base AbstractVector functions, though SmallVector will fail for mutating functions like setindex! because it is immutable.

MSmallVector has specialized implementations of Base functions that involve resizing such as:

• resize!
• push!
• pushfirst!
• pop!
• popfirst!
• append!
• prepend!
• insert!
• deleteat! which are guaranteed to not realocate memory, and instead just use the memory buffer that already exists, unlike Base's Vector which may have to reallocate memory depending on the operation. However, they will error if they involve operations that resize beyond the maximum length of the MSmallVector, which you can access with SmallVectors.maxlength(v).

In addition, SmallVector and MSmallVector implement basic non-mutating operations such as:

• SmallVectors.setindex , non-mutating resizing operations:
• SmallVector.resize
• SmallVector.push
• SmallVector.pushfirst
• SmallVector.pop
• SmallVector.popfirst
• SmallVector.append
• SmallVector.prepend
• SmallVector.insert
• SmallVector.deleteat which output a new vector. In addition, it implements:
• SmallVectors.circshift
• sort (overloaded from Base).

Finally, it provides some new helpful functions that are not in Base:

• SmallVectors.insertsorted[!]
• SmallVectors.insertsortedunique[!]
• SmallVectors.mergesorted[!]
• SmallVectors.mergesortedunique[!]

TODO

Add specialized overloads for:

• splice[!]
• union[!] (∪)
• intersect[!] (∩)
• setdiff[!]
• symdiff[!]
• unique[!]

Please let us know if there are other operations that would warrant specialized implmentations for AbstractSmallVector.