BitVectors of length 1 are saved as scalars not arrays

[bjarthur]

julia> using MAT

julia> matwrite("foo.mat", Dict("x"=>trues(1)))

julia> matread("foo.mat")
Dict{String, Any} with 1 entry:
  "x" => true

julia> matwrite("foo.mat", Dict("x"=>trues(2)))

julia> matread("foo.mat")
Dict{String, Any} with 1 entry:
  "x" => Bool[1, 1]

the workaround:

julia> matwrite("foo.mat", Dict("x"=>convert(Array{Any}, trues(1))))

julia> matread("foo.mat")
Dict{String, Any} with 1 entry:
  "x" => Any[true]

[inkydragon]

The behavior of the existing API is correct.

Generating test mat file in julia:

matwrite("ju.mat", 
    Dict(
        "x1"=>trues(1),
        "x2"=>trues(2),
        "x3"=>convert(Array{Any}, trues(1))     
))

Read the test mat file in MATLAB:

>> clear
>> load('ju.mat')
>> x1
x1 =

  logical

   1

>> x2
x2 =

  2×1 logical array

   1
   1

>> x3
x3 =

  1×1 cell array

    {[1]}

---

Write mat file in MATLAB:

>> clear
>> m1 = logical([1])
m1 =

  logical

   1

>> m2 = logical([1; 1])
m2 =

  2×1 logical array

   1
   1

>> m3 = logical([1 1])
m3 =

  1×2 logical array

   1   1

>> m4 = {1}
m4 =

  1×1 cell array

    {[1]}

>> m5 = {logical(1)}
m5 =

  1×1 cell array

    {[1]}

>> save('m.mat')

Read mat file in julia:

julia> matread("m.mat")
Dict{String, Any} with 5 entries:
  "m3" => Bool[1 1]
  "m2" => Bool[1; 1]
  "m4" => Any[1.0]
  "m5" => Any[true]
  "m1" => true

[matthijscox]

Unfortunately this is a MATLAB definition problem, MATLAB treats all 1x1 arrays as scalars. There is no actual numerical scalar type in MATLAB. Therefore MAT.jl has to read/writes scalars as 1x1 arrays, for all bitstypes (Bool, Float64, Int64, etc.). There's no way around this. If you think we need to clearly define this somewhere in the docs, please open another issue. Closing this one.