Skip to content

Add functionality to use arbitrary singular values in ConcatDiskArrays #256

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 5 commits into
base: fg/cat
Choose a base branch
from
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 14 additions & 13 deletions src/cat.jl
Original file line number Diff line number Diff line change
@@ -1,4 +1,3 @@

"""
ConcatDiskArray <: AbstractDiskArray

Expand All @@ -15,7 +14,7 @@ Returned from `cat` on disk arrays.

It is also useful on its own as it can easily concatenate an array of disk arrays.
"""
struct ConcatDiskArray{T,N,P,C,HC,ID} <: AbstractDiskArray{T,N}
struct ConcatDiskArray{T,N,P,C,HC, ID} <: AbstractDiskArray{T,N}
parents::P
startinds::NTuple{N,Vector{Int}}
size::NTuple{N,Int}
Expand All @@ -31,16 +30,18 @@ function ConcatDiskArray(arrays::AbstractArray{Union{<:AbstractArray,Missing}})
M = ndims(et)
_ConcatDiskArray(arrays, T, Val(N), Val(M))
end

function infer_eltypes(arrays)
foldl(arrays, init=(-1, Union{})) do (M, T), a
if ismissing(a)
(M, promote_type(Missing, T))
if !isa(a, AbstractArray)
(M, promote_type(typeof(a), T))
else
M == -1 || ndims(a) == M || throw(ArgumentError("All arrays to concatenate must have equal ndims"))
(ndims(a), promote_type(eltype(a), T))
end
end
end

function ConcatDiskArray(arrays::AbstractArray{<:AbstractArray})
N = ndims(arrays)
T = eltype(eltype(arrays))
Expand Down Expand Up @@ -90,7 +91,7 @@ function arraysize_and_startinds(arrays1)
sizes = map(i -> zeros(Int, i), size(arrays1))
for i in CartesianIndices(arrays1)
ai = arrays1[i]
ismissing(ai) && continue
ai isa AbstractArray || continue
sizecur = extenddims(size(ai), size(arrays1), 1)
foreach(sizecur, i.I, sizes) do si, ind, sizeall
if sizeall[ind] == 0
Expand Down Expand Up @@ -123,10 +124,10 @@ function readblock!(a::ConcatDiskArray, aout, inds::AbstractUnitRange...)
# Find affected blocks and indices in blocks
_concat_diskarray_block_io(a, inds...) do outer_range, array_range, I
vout = view(aout, outer_range...)
if ismissing(I)
vout .= missing
else
if I isa CartesianIndex
readblock!(a.parents[I], vout, array_range...)
else
vout .= I
end
end
end
Expand Down Expand Up @@ -170,10 +171,10 @@ function _concat_diskarray_block_io(f, a::ConcatDiskArray, inds...)
#Shorten array range to shape of actual array
array_range = ntuple(j -> array_range[j], ID)
outer_range = fix_outerrangeshape(outer_range, array_range)
if ismissing(myar)
f(outer_range, array_range, missing)
else
if myar isa AbstractArray
f(outer_range, array_range, cI)
else
f(outer_range, array_range, myar)
end
end
end
Expand All @@ -189,13 +190,13 @@ function concat_chunksize(parents)
newchunks = map(s -> Vector{Union{RegularChunks,IrregularChunks}}(undef, s), size(parents))
for i in CartesianIndices(parents)
array = parents[i]
ismissing(array) && continue
array isa AbstractArray || continue
chunks = eachchunk(array)
foreach(chunks.chunks, i.I, newchunks) do c, ind, newc
if !isassigned(newc, ind)
newc[ind] = c
elseif c != newc[ind]
throw(ArgumentError("Chunk sizes don't forma grid"))
throw(ArgumentError("Chunk sizes don't form a grid"))
end
end
end
Expand Down
29 changes: 29 additions & 0 deletions test/runtests.jl
Original file line number Diff line number Diff line change
Expand Up @@ -492,6 +492,34 @@ end
@test slic == Float64[1, 2, 3, 4, 1, 2, 3, 4]
end

@testset "Concat DiskArray with fill zero tiles" begin
a = zeros(Int, 3, 4)
b = ones(Int, 2, 4)
c = fill(2, 3, 5)
d = fill(0, 2, 5)
aconc = DiskArrays.ConcatDiskArray(reshape([a, b, c, 0], 2, 2))
abase = [a c; b d]
@test all(isequal.(aconc[:, :], abase))
@test all(isequal.(aconc[3:4, 4:6], abase[3:4, 4:6]))
ch = DiskArrays.eachchunk(aconc)
@test ch.chunks[1] == [1:3, 4:5]
@test ch.chunks[2] == [1:4, 5:9]

a = ones(100, 50)
b = [rem(i.I[3], 5) == 0 ? 0 : a for i in CartesianIndices((1, 1, 100))]
b[1] = 0
a_conc = DiskArrays.ConcatDiskArray(b)
ch = eachchunk(a_conc)
@test ch.chunks[1] == [1:100]
@test ch.chunks[2] == [1:50]
@test ch.chunks[3] === DiskArrays.RegularChunks(1, 0, 100)

@test all(isequal.(a_conc[2, 2, 1:5], [0, 1.0, 1.0, 1.0, 0]))
@test all(isequal.(a_conc[end, end, 95:100], [0, 1.0, 1.0, 1.0, 1.0, 0]))

end


@testset "Concat DiskArray with missing tiles" begin
a = zeros(Int, 3, 4)
b = ones(Int, 2, 4)
Expand All @@ -518,6 +546,7 @@ end
@test all(isequal.(a_conc[end, end, 95:100], [missing, 1.0, 1.0, 1.0, 1.0, missing]))

end

end

@testset "Broadcast with length 1 and 0 final dim" begin
Expand Down