In richelbilderbeek/ormr: Work With Python Installed At A Custom Location

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

ormr

ormr allows to work with Python installed to a custom location, using only a single point of contact.

library(ormr)

We use only one point of contact, ormr_folder_name, which is the folder where ormr installs the Python environment and packages needed:

if (is_conda_installed()) {
  ormr_folder_name <- get_default_ormr_folder_name()
} else {
  ormr_folder_name <- "python3"
}

Note we use the default ormr_folder_name, as this package needs to only use one Python version, in both this vignette and its tests.

Goal

The goal of this vignette is to show how to use ormr to run an example Python script, where Python is installed to a custom location.

Here is the example Python script we want to run:

python_script_path <- system.file("extdata", "scipy_example.py", package = "ormr")
readLines(python_script_path)

As a spoiler, it is demonstrated here that the script maybe cannot be run yet:

try(
  run_python_script(
    ormr_folder_name = ormr_folder_name,
    python_script_path = python_script_path
  ), 
  silent = FALSE
)

When using the default ormr Conda environment, scipy may actually be installed.

Overview

These are the steps to achieve the goal:

1. Install Python packages
2. Run a Python script

Note that there is no need to create a Conda environment: this is what ormr does for you. ormr uses eager loading, which is the opposite of reticulate's lazy loading.

1. Install Python packages within that Conda environment

Currently, these packages are installed:

installed_python_packages <- ormr::get_installed_python_packages(
  ormr_folder_name = ormr_folder_name
)
knitr::kable(installed_python_packages)

We are going to install scipy, which is a small package:

package_name <- "scipy"

We can verify that it may not be installed yet:

is_python_package_installed(
  ormr_folder_name = ormr_folder_name,
  package_name = package_name
)

The package may be installed when using the default ormr Conda environment.

Also, running an Python script that requires scipy may fail now:

try(
  run_python_script(
    ormr_folder_name = ormr_folder_name,
    python_script_path = python_script_path
  ),
  silent = FALSE
)

The script may pass when using the default ormr Conda environment and the needed Python packages are already installed.

Here, we install the package:

ormr::install_python_packages(
  ormr_folder_name = ormr_folder_name,
  package_names = package_name
)

Now the Python package is installed:

testthat::expect_true(
  ormr::is_python_package_installed(
    ormr_folder_name = ormr_folder_name,
    package_name = package_name
  )
)

2. Run a Python script with that Conda environment and packages

Great, with packages in place, we run a Python script with those packages:

testthat::expect_silent(
  ormr::run_python_script(
    ormr_folder_name = ormr_folder_name,
    python_script_path = python_script_path
  )
)

No error, so we can conclude it works!

Cleaning up

The clean up, we throw away the ormr folder in which the magic happened.

if (ormr_folder_name != get_default_ormr_folder_name()) {
  unlink(ormr_folder_name, recursive = TRUE)  
}

richelbilderbeek/ormr documentation built on May 23, 2022, 1:15 p.m.