Interoperability

The PythonCall.jl package makes writing and reading ASDF files from the Python implementation of ASDF fairly seamless. For example, here is one way we might reproduce the Creating Files section from the Python asdf documentation without leaving our Julia session:

Accessing the IPython REPL from Julia

We start by creating a fresh Julia environment named asdf in a global location:

julia --proj=@asdf # Or --proj=<path to a custom location>

Tip

To verify that this is a blank environment and to also display its location, run the st command (short for status) from the package REPL:

julia> # Press ] to enter the package mode REPL
(@asdf) pkg> st
Status `~/.julia/environments/asdf/Project.toml` (empty project)

We next install whichever Python packages we may need for our analysis, e.g., asdf and astropy. PythonCall.jl exports CondaPkg.jl, which has nice hooks for accomplishing this:

(@asdf) pkg> add PythonCall

julia> using PythonCall

(@asdf) pkg> conda add asdf astropy

(@asdf) pkg> conda st
CondaPkg Status <path to your julia environment>/CondaPkg.toml
Environment
  <path to your julia environment>/.CondaPkg/.pixi/envs/default
Packages
  asdf v5.3.0
  astropy v7.2.0

(@asdf) pkg> conda run ipython

The last command (conda run ipython) starts an IPython REPL provided in the astropy installation directly from Julia. We can then enter regular Python syntax verbatim to create a sample ASDF file called example.asdf:

from asdf import AsdfFile
import numpy as np

# Create some data
sequence = np.arange(100)
squares = sequence**2
random = np.random.random(100)

# Store the data in an arbitrarily nested dictionary
tree = {
    "foo": 42,
    "name": "Monty",
    "sequence": sequence,
    "powers": {"squares": squares},
    "random": random,
}

# Create the ASDF file object from our data tree
af = AsdfFile(tree)

# Write the data to a new file
af.write_to("example.asdf")

Note

IPython does not come with PythonCall.jl/CondaPkg.jl by default. If working with packages other than astropy that do not bundle ipython, use conda run python instead.

Julia REPL

PythonCall.jl also makes it easy to run Python syntax directly from the Julia REPL. For example, the above IPython commands could also be run directly using the PythonCall.@pyexec macro:

julia --proj=@asdf # Re-use the environment created earlier

using PythonCall

@pyexec """
from asdf import AsdfFile
import numpy as np

# Create some data
sequence = np.arange(100)
squares = sequence**2
random = np.random.random(100)

# Store the data in an arbitrarily nested dictionary
tree = {
    "foo": 42,
    "name": "Monty",
    "sequence": sequence,
    "powers": {"squares": squares},
    "random": random,
}

# Create the ASDF file object from our data tree
af = AsdfFile(tree)

# Write the data to a new file
af.write_to("example.asdf")
"""

The PythonCall.@py macro reduces the barrier between Python and Julia even further by allowing us to work with Python objects directly using Julia syntax:

@py begin
    import asdf: AsdfFile # Julia syntax version of `from asdf import AsdfFile`
    import numpy as np
end

sequence = np.arange(100)

squares = sequence^2 # Note that Julia uses `^` instead of `**`

random = np.random.random(100)

# Create a Python dict with PythonCall.pydict
tree = pydict(
    "foo" => 42,
    "name" => "Monty",
    "sequence" => sequence,
    "powers" => pydict("squares" => squares),
    "random" => random,
);

af = AsdfFile(tree)

af.write_to("example.asdf")

See the PythonCall.jl documentation for more.