using HDF5
using Test

@testset "extend" begin
    fn = tempname()

    fid = h5open(fn, "w")
    g = create_group(fid, "shoe")
    d = create_dataset(g, "foo", datatype(Float64), ((10, 20), (100, 200)); chunk=(1, 1))
    #println("d is size current $(map(int,HDF5.get_extent_dims(d)[1])) max $(map(int,HDF5.get_extent_dims(d)[2]))")
    dims, max_dims = HDF5.get_extent_dims(d)
    @test dims == (UInt64(10), UInt64(20))
    @test max_dims == (UInt64(100), UInt64(200))
    HDF5.set_extent_dims(d, (100, 150))
    dims, max_dims = HDF5.get_extent_dims(d)
    @test dims == (UInt64(100), UInt64(150))
    @test max_dims == (UInt64(100), UInt64(200))
    d[1, 1:5] = [1.1231, 1.313, 5.123, 2.231, 4.1231]
    HDF5.set_extent_dims(d, (1, 5))
    @test size(d) == (1, 5)

    # Indexing returns correct array dimensions
    @test d[1, end] ≈ 4.1231
    @test d[:, end] ≈ [4.1231]
    @test d[end, :] == [1.1231, 1.313, 5.123, 2.231, 4.1231]
    @test d[:, :] == [1.1231 1.313 5.123 2.231 4.1231]

    if VERSION >= v"1.4"
        include("extend_test_begin.jl")
        extend_test_begin(d)
    end

    # Test all integer types work
    @test d[UInt8(1), UInt16(1)] == 1.1231
    @test d[UInt32(1), UInt128(1)] == 1.1231
    @test d[Int8(1), Int16(1)] == 1.1231
    @test d[Int32(1), Int128(1)] == 1.1231

    # Test ranges work with steps
    @test d[1, 1:2:5] == [1.1231, 5.123, 4.1231]
    @test d[1:1, 2:2:4] == [1.313 2.231]

    # Test Array constructor
    Array(d) == [1.1231 1.313 5.123 2.231 4.1231]

    #println("d is size current $(map(int,HDF5.get_extent_dims(d)[1])) max $(map(int,HDF5.get_extent_dims(d)[2]))")
    b = create_dataset(fid, "b", Int, ((1000,), (-1,)); chunk=(100,)) #-1 is equivalent to typemax(hsize_t) as far as I can tell
    #println("b is size current $(map(int,HDF5.get_extent_dims(b)[1])) max $(map(int,HDF5.get_extent_dims(b)[2]))")
    b[1:200] = ones(200)
    dims, max_dims = HDF5.get_extent_dims(b)
    @test dims == (UInt64(1000),)
    @test max_dims == (HDF5.API.H5S_UNLIMITED % Int,)
    HDF5.set_extent_dims(b, (10000,))
    dims, max_dims = HDF5.get_extent_dims(b)
    @test dims == (UInt64(10000),)
    @test max_dims == (HDF5.API.H5S_UNLIMITED % Int,)
    #println("b is size current $(map(int,HDF5.get_extent_dims(b)[1])) max $(map(int,HDF5.get_extent_dims(b)[2]))")
    # b[:] = [1:10000] # gave error no method lastindex(HDF5.Dataset{PlainHDF5File},),
    # so I defined lastindex(dset::HDF5.Dataset) = length(dset), and exported lastindex
    # but that didn't fix the error, despite the lastindex function working
    # d[1] produces error ERROR: Wrong number of indices supplied, should datasets support linear indexing?
    b[1:10000] = [1:10000;]
    #println(b[1:100])

    close(fid)

    fid = h5open(fn, "r")
    d_again = fid["shoe/foo"]
    dims, max_dims = HDF5.get_extent_dims(d_again)
    @test dims == (UInt64(1), UInt64(5))
    @test max_dims == (UInt64(100), UInt64(200))
    @test (sum(d_again[1, 1:5]) - sum([1.1231, 1.313, 5.123, 2.231, 4.1231])) == 0
    #println("d is size current $(map(int,HDF5.get_extent_dims(re_d)[1])) max $(map(int,HDF5.get_extent_dims(re_d)[2]))")
    @test fid["b"][1:10000] == [1:10000;]
    b_again = fid["b"]
    dims, max_dims = HDF5.get_extent_dims(b_again)
    @test dims == (UInt64(10000),)
    @test max_dims == (HDF5.API.H5S_UNLIMITED % Int,)
    #println("b is size current $(map(int,HDF5.get_extent_dims(b)[1])) max $(map(int,HDF5.get_extent_dims(b)[2]))")

    close(fid)
    rm(fn)
end # testset extend_test