reshaped views and reinterpreted arrays #161
Conversation
Codecov Report
@@ Coverage Diff @@
## master JuliaArrays/ArrayInterface.jl#161 +/- ##
==========================================
+ Coverage 82.74% 82.80% +0.06%
==========================================
Files 11 11
Lines 1721 1733 +12
==========================================
+ Hits 1424 1435 +11
- Misses 297 298 +1
Continue to review full report at Codecov.
|
src/stridelayout.jl
Outdated
| @@ -310,10 +310,11 @@ _is_column_major(sr::R, cbs::StaticInt) where {R} = False() | |||
| _is_column_major(sr::R, cbs::Union{StaticInt{0},StaticInt{-1}}) where {R} = is_increasing(sr) | |||
|
|
|||
| """ | |||
| dense_dims(::Type{T}) -> NTuple{N,Bool} | |||
| dense_dims(::Type{<:AbstractArray{N}}) -> NTuple{N,Bool} | |||
There was a problem hiding this comment.
Thanks! Implemented in 7335b49
| # TODO: Should be generalized to reshaped arrays wrapping more general array types | ||
| function strides(A::ReshapedArray{T,N,P}) where {T, N, P<:AbstractVector} | ||
| if defines_strides(A) | ||
| return size_to_strides(size(A), first(strides(parent(A)))) |
There was a problem hiding this comment.
What about
julia> A = rand(5,12);
julia> B = reshape(A', (3,4,5));ArrayInterface.strides(B) should be (5,15,StaticInt{1}()).
There was a problem hiding this comment.
I would prefer to fix that later since I would need to think a bit more about it. I opened #162 to track this bug.
|
I pushed a fix for #163. I also updated the version in Project.toml since I would really like to have a new release to be able to use these fixes. |
I have fixed the issues reported in #157 and JuliaArrays/StaticArrayInterface.jl#2. Here, I concentrated on the exact applications using reshaped versions of views that are abstract vectors. In this case, perfect might be the enemy of good again: This solution isn't perfect but it solves the problems I have right now and can be extended later to reshaped arrays of more general array types. I also adapted some docstrings slightly to look more accurate to me.