An Elixir downsampling library that retains the visual characteristics of your data.
ExLTTB is based on Largest-Triangle-Three-Buckets (LTTB) and it contains two implementation modules: ExLTTB implements the original algorithm using Enum and eager evaluation, ExLTTB.Stream implements a slightly modified version to cope with lazy evaluation and using Stream.
The data is assumed to be ordered by its x coordinate in both implementations.
Add ex_lttb to your list of dependencies in mix.exs:
def deps do
[
{:ex_lttb, "~> 0.3.0"}
]
endYou can read the documentation on HexDocs.
# Downsample a list of data in the default shape of %{x: x, y: y}
input_samples = for x <- 1..100, do: %{x: x, y: :random.uniform() * 100}
# 30 output samples
output_samples = ExLTTB.downsample_to(input_samples, 30)
# Downsample a list of data of arbitrary shape
input_samples =
for x <- 1..100 do
%{nested: %{timebase: x},
data: :random.uniform() * 100,
untouched_other_key: :random.uniform() * 3
}
end
sample_to_x_fun = fn sample -> sample[:nested][:timebase] end
sample_to_y_fun = fn sample -> sample[:data] end
xy_to_sample_fun = fn x, y -> %{nested: %{timebase: x}, data: y} end
output_samples =
ExLTTB.downsample_to(
input_samples,
30,
sample_to_x_fun: sample_to_x_fun,
sample_to_y_fun: sample_to_y_fun,
xy_to_sample_fun: xy_to_sample_fun
)
# Downsample a stream of data
input_stream =
Stream.iterate(%{x: 0, y: :random.uniform() * 100}, fn %{x: x} -> %{x: x + 1, y: :random.uniform() * 100} end)
# Downsample rate of 2.3
output_samples = ExLTTB.Stream.downsample(input_stream, 2.3) |> Enum.take(20)
# The options for arbitrary shaped data are the same for the streaming version