julia_port

julia_port copied to clipboard

julia_port

example project to invoke julia functions in elixir to do scientific computing using port and metaprogramming

news

• support julia v1.0.1
• support elixir v1.7.3
• support erlang 21.0
• the script_test is supported
• the real_test is removed because it does not support julia v1.0.1

remarks

• the author will experiment with tensorflow.jl for good reason

prerequisite

• julia installed and its access from shell

Installation

the package can be installed by adding julia_port to your list of dependencies in mix.exs:

def deps do
  [{:julia_port, "~> 0.2.0"}]
end

remark: if you install the package as a dependency, in order to run script_test, you have to move ./deps/julia_port/julia and ./deps/julia_port/data to ./julia and ./data respectively

usage

• simple_test: arithmetics

def simple_test(port) do
  port_send(port, "1+2")
  IO.puts port_receive(port, true)
end

• complex_test: linear algebra

use GenFunction, [rand: 2, sum: 1, *: 2]

def complex_test(port) do
  rand(port, :a, 3, 3)
  rand(port, :b, 3, 3)
  JuliaPort.*(port, :c, :a, :b)
  port_receive(port, false)
  sum(port, :d, :c)
  IO.puts port_receive(port, true)
end

• script_test: linear regression

use GenFunction, [load_data: 1, lr_train: 2, lr_test: 3]

def script_test(port) do
  include_script(port, "./julia/lr.jl")
  load_data(port, {:x_train, :y_train}, "./data/train")
  load_data(port, {:x_test, :y_test}, "./data/test")
  lr_train(port, :beta, :x_train, :y_train)
  port_receive(port, false)
  lr_test(port, :error, :x_test, :y_test, :beta)
  IO.puts port_receive(port, true)
end

• run

iex -S mix
port = JuliaPort.init
# => #Port<0.5310>
JuliaPort.print_version port
# => received data: v"1.0.0"
JuliaPort.simple_test port
# => received data: 3
JuliaPort.complex_test port
# => received data: 5.710327361153192
JuliaPort.script_test port
# => received data: 0.9587912087912088
JuliaPort.terminate port
# => {#PID<0.143.0>, :close}