Geospatial Zarr Datasets

Mix.install([
  {:ex_zarr, "~> 1.2"}
  # {:ex_zarr, path: Path.join(__DIR__, "../../..")},
])

Pattern

Climate and EO datasets often use (time, lat, lon). Stream along time with stream_slices/3, or extract windows with get_slice/2.

alias ExZarr.Array
alias ExZarr.Gallery.Pack

# Compact stand-in for (time, lat, lon)
{:ok, array} =
  Array.create(
    shape: {12, 18, 36},
    chunks: {1, 18, 36},
    dtype: :float32,
    compressor: :zlib,
    storage: :memory
  )

n = 12 * 18 * 36
vals = for i <- 0..(n - 1), do: i * 0.001
:ok = Array.set_slice(array, Pack.pack(vals, :float32), start: {0, 0, 0}, stop: {12, 18, 36})

timesteps =
  array
  |> Array.stream_slices(0, concurrency: 2)
  |> Enum.map(fn {start, data} ->
    {elem(start, 0), byte_size(data)}
  end)

%{timesteps: Enum.take(timesteps, 3), total_steps: length(timesteps)}

For spatial windows, use explicit start/stop (or named dimensions on v3 arrays).