ArrayStream.jl

This is an attempt to make a more convenient way to read/write Arrays in Julia. My initial focus is on images but if this is found to be useful to others it could proabably be adapted to any sort of array structure in Julia.

ArrayStreams.jl is not a registered package. If there is enough interest in this project I'd be happy to register it. If there's another infrastructure out there that's just better but could benefit from this sort of functionality I'd also be willing to assist in porting this sort of functionality elsewhere.

Usage

The main workhorse of this package is ArrayStream. It is a subtype of AbstractArrayStream which is an AbstractArray. An ArrayStream gains most of it's utility by containing an IO field that has some fairly basic indexing behavior being wrapped by other types that require an AbstractArray (especially AxisArrays). This results in several key contributions beyond just convenient syntax (almost identical to base).

• Reading to a sink. This is similar to Base.read!, but allows AxisArray and ImageMeta to be overwritten.
• Axis based interpolation of how to read struct elements. If you're familiar with Images.jl you may be aware of the colorview function and Axis traits that can detect which dimension is responsible for color. Leveraging AxisArrays allows reading in arrays where the desired sink eltype is a struct but this isn't containing in the stream elementwise but represented as one of the dimensions (e.g., the last dimension of a Float32 array represents each channel of an RGB struct). More on this below.
• Data may be read and written in chunks.

Basics

If an ArrayStream is constructed from an Array, an IOBuffer will be created that contains the contents of the arrays.

A = rand(3,4)
s = ArrayStream(A)

Once an ArrayStream is constructed it contains all the necessary information to construct an array from the underlying IO.

A2 = read(s)

An ArrayStream may also be treated similarly to any IO object.

position(s)
eof(s)
isopen(s)
seek(s, 0)

An ArrayStream also has the ability to read to a sink.

B = rand(3,4)
read!(s, B)
B == A

Sinks

TODO

io = IOBuffer()
s = ArrayStream{Tuple{2,5},Int,2}(io,0:80,true,false)
write(s, rand(2,5))

ChunkedStream

TODO