The Python module JuliaCall

Installation

It's as simple as

pip install juliacall

Developers may wish to clone the repo (https://github.com/JuliaPy/PythonCall.jl) directly and pip install the module in editable mode. You should add "dev":true, "path":"../.." to pysrc/juliacall/juliapkg.json to ensure you use the development version of PythonCall in conjunction with JuliaCall.

Getting started

For interactive or scripting use, the simplest way to get started is:

from juliacall import Main as jl

This loads a single variable jl which represents the Main module in Julia, from which all of Julia's functionality is available:

jl.println("Hello from Julia!")
# Hello from Julia!
x = jl.rand(range(10), 3, 5)
x._jl_display()
# 3×5 Matrix{Int64}:
#  8  1  7  0  6
#  9  2  1  4  0
#  1  8  5  4  0
import numpy
numpy.sum(x, axis=0)
# array([18, 11, 13,  8,  6], dtype=int64)

In this example:

• We called the jl.println function to print a message.
• We called the jl.rand function to generate an array of random integers. Note that the first argument is range(10) which is converted to 0:9 in Julia.
• We called its special _jl_display() to show it using Julia's display mechanism.
• We called the numpy.sum function to sum each column of x. This automatically converted x to a NumPy array. (We could have done jl.sum(x, dims=1) too.)

If you are writing a package which uses Julia, then to avoid polluting the global Main namespace you instead should start with:

import juliacall
jl = juliacall.newmodule("SomeName")

Julia modules have a special method seval which will evaluate a given piece of code given as a string in the module. This is most frequently used to import modules:

from array import array
jl.seval("using Statistics")
x = array('i', [1, 2, 3])
jl.mean(x)
# 2.0
y = array('i', [2,4,8])
jl.cor(x, y)
# 0.9819805060619657

What to read next:

• The main functionality of this package is in AnyValue objects, which represent Julia objects, documented here.
• If you need to install Julia packages, read here.
• When you call a Julia function, such as jl.rand(...) in the above example, its arguments are converted to Julia according to this table and its return value is converted to Python according to this table.

Managing Julia dependencies

JuliaCall manages its Julia dependencies using JuliaPkg.

It will automatically download a suitable version of Julia if required.

A Julia environment is also created, activated and populated with any required packages. If you are in a virtual or Conda environment, the environment is put there. Otherwise a global environment is used at ~/.julia/environments/pyjuliapkg.

If your project requires any Julia packages, or a particular version of Julia itself, then create a file called juliapkg.json in your package. For example: Here is an example:

{
    "julia": "1.5",
    "packages": {
        "Example": {
            "uuid": "7876af07-990d-54b4-ab0e-23690620f79a",
            "version": "0.5, 0.6"
        }
    }
}

Alternatively you can use add, rm, etc. from JuliaPkg to edit this file.

See JuliaPkg for more details.

Configuration

Some features of the Julia process, such as the optimization level or number of threads, may be configured in two ways:

• As an -X argument to Python, such as -X juliacall-optlevel=3; or
• As an environment variable, such as PYTHON_JULIACALL_OPTLEVEL=3.

Multi-threading

From v0.9.22, JuliaCall supports multi-threading in Julia and/or Python, with some caveats.

Most importantly, you can only call Python code while Python's Global Interpreter Lock (GIL) is locked by the current thread. You can use JuliaCall from any Python thread, and the GIL will be locked whenever any JuliaCall function is used. However, to leverage the benefits of multi-threading, you can unlock the GIL while executing any Julia code that does not interact with Python.

The simplest way to do this is using the _jl_call_nogil method on Julia functions to call the function with the GIL unlocked.

from concurrent.futures import ThreadPoolExecutor, wait
from juliacall import Main as jl
pool = ThreadPoolExecutor(4)
fs = [pool.submit(jl.Libc.systemsleep._jl_call_nogil, 5) for _ in range(4)]
wait(fs)

In the above example, we call Libc.systemsleep(5) on four threads. Because we called it with _jl_call_nogil, the GIL was unlocked, allowing the threads to run in parallel, taking about 5 seconds in total.

If we did not use _jl_call_nogil (i.e. if we did pool.submit(jl.Libc.systemsleep, 5)) then the above code will take 20 seconds because the sleeps run one after another.

It is very important that any function called with _jl_call_nogil does not interact with Python at all unless it re-locks the GIL first, such as by using PythonCall.GIL.@lock.

You can also use multi-threading from Julia.

Caveat: Julia's task scheduler

If you try the above example with a Julia function that yields to the task scheduler, such as sleep instead of Libc.systemsleep, then you will likely experience a hang.

In this case, you need to yield back to Julia's scheduler periodically to allow the task to continue. You can use the following pattern instead of wait(fs):

jl_yield = getattr(jl, "yield")
while True:
  # yield to Julia's task scheduler
  jl_yield()
  # wait for up to 0.1 seconds for the threads to finish
  state = wait(fs, timeout=0.1)
  # if they finished then stop otherwise try again
  if not state.not_done:
    break

Set the timeout parameter smaller to let Julia's scheduler cycle more frequently.

Future versions of JuliaCall may provide tooling to make this simpler.

Caveat: Signal handling

We recommend setting PYTHON_JULIACALL_HANDLE_SIGNALS=yes before importing JuliaCall with multiple threads.

This is because Julia intentionally causes segmentation faults as part of the GC safepoint mechanism. If unhandled, these segfaults will result in termination of the process. See discussion here for more information.

Note however that this interferes with Python's own signal handling, so for example Ctrl-C will not raise KeyboardInterrupt.

Future versions of JuliaCall may make this the default behaviour when using multiple threads.