Missing dependencies after registering Python as external evaluator

Questions
1

Hi all,
I’ve registered Python as external evaluator and when I am going to find the external evaluator with
FindExternalEvaluator[“Python”], I got the correct python version and executable but I have “missing dependencies”.
Any idea how to solve the issue?
Thanks and regards
Paolo

2

You need to specify where to find Python:

 RegisterExternalEvaluator[
    "Python", 
    "C:\\bin\\Anaconda3\\python.EXE"]

Register a Python-Numpy library:

    RegisterExternalEvaluator[
        "Python-NumPy", 
        "C:\\bin\\Anaconda3\\python.EXE"]

How to import a Numpy library

    ExternalEvaluate[
        "Python", 
        "import numpy; x = numpy.pi; x**2"
    ]

Import any libraries:

ExternalEvaluate[
        "Python", 
        "from sympy import isprime; isprime(7)"
    ]

Examples from the following websites: How to call Python from Mathematica, including NumPy.

3

In current versions of the Wolfram Language you can simply start a Python session in a virtual environment as follows (here I am using astropy as an example, but you can do this with any Python package):

session = StartExternalSession[<|
   "System" -> "Python",
   "Evaluator" -> <|"Dependencies" -> {"astropy"}|>,
   "SessionProlog" -> {
     "from astropy.time import Time",
     "from astropy.coordinates import get_body_barycentric",
     "from astropy import units as u"
     }|>]

Note how I specify the dependent astropy package and how I am also running session initialization code. Now we can define a callable function with some python code:

EarthMarsDistance = ExternalFunction[session, "
def distance_earth_mars(date_str):
    try:
        time = Time(date_str)
        earth = get_body_barycentric('earth', time)
        mars = get_body_barycentric('mars', time)
        separation = (mars - earth).norm()
        distance_au = separation.to(u.au)
        return distance_au.value
    except ValueError:
        return None
"]

This function computes the distance from Earth to Mars for a given ISO date string. You can call it directly from the Wolfram Language:

EarthMarsDistance[DateString[Now, "ISODateTime"]]

This gave the answer of 1.29486 (astronomial units) at the time of evaluation.

You can then also use the EarthMarsDistance function in other Wolfram Language code, for example a plot over time:

ListLinePlot[
   Table[
        EarthMarsDistance[DateString[t, "ISODateTime"]], 
        {t, DateRange[Now, Now + Quantity[1000, "Days"]]}]]

Which gives a nice plot:

earth-mars.png
earth-mars.png720×451 16.5 KB

In this case you can do this directly in the Wolfram Language with the AstroDistance function:

AstroDistance[Entity["Planet", "Mars"]]

But if you have Python code that you would like to call from the Wolfram Language, then this is the route I suggest you take. I have attached a notebook with all the inputs from this comment.

astropy-01.nb (116 KB)