Using DO.utils with reticulate and Python

What does Python add?

DO.utils is designed to balance functionality and ease of use, with a goal for reproducibility. Some functionality is just not available in R, or it’s less reproducible or harder to manage. Python fills in these gaps. Specifically using Python powers functions that:

1. Execute SPARQL queries on local files.
   • The R SPARQL engines require database management (virtuoso) or do not work with SPARQL 1.1 (rdflib).
2. Extract information from git repos.
   • To update data from releases and analyze unexpected changes to files.
3. Predict term mappings.
   • Utilizing the biopragmatics stack: bioregistry, biomappings, etc.; pyobo; and INDRA labs’ GILDA

Why use Python via reticulate?

Generally, the R-python connection allows data integration with R where data frames can be more easily managed, explored, and analyzed. More specifically, reticulate should help manage Python and python module dependencies and simplify user setup/use.

How to use python-dependent DO.utils functions?

These functions should work as standard R functions. In some cases internal data extractor functions have been written to eliminate the need for non-standard user data munging.

Python Setup

Although it has not been tested yet, use of reticulate should handle installation of python and/or python modules automatically and require little to no user input. The default is to attempt to identify and use python that is already installed. Should that not work, reticulate should install python and needed modules via miniconda.

Customization

More user control can be achieved over this installation (like using pip and venv instead of miniconda) by using reticulate to install python and create a virtual environment before calling library(DO.utils) for the first time. The basic procedure would be to execute the following:

library(reticulate) # installs with DO.utils
py_version <- "3.9" # or whatever version you prefer
venv_name <- "r-reticulate"

py_path <- install_python(py_version)
virtualenv_create(venv_name, py_path, version = py_version)

A Note on Virtual Environments

reticulate can create conda environments or venv/virtualenv virtual environments. Non-conda virtual environment options are described briefly below but similar options approaches are available for conda environments via reticulate.

r-reticulate is the default virtual environment name promoted by the developers of reticulate for ALL packages that depend on reticulate, with the goals of 1) promoting interoperability between packages that depend on reticulate and python and 2) to better manage python module dependencies, through the use of a single virtual environment. Alternatives to this default environment can be specified by modifying the first argument envname to virtualenv_create(). A name (without slashes) will create a virtual environment of that name in the default location (see virtualenv_root()); a relative or absolute path (with slashes) will create a virtual environment in the specified directory, which can be nice if you want a python virtual environment associated with a specific project.

A Note on Python Module Installation

It is possible to specify python modules to install at the time a virtual environment is created or later by activating the virtual environment and executing virtualenv_install(). If only DO.utils dependencies are desired, this should not be necessary as the first execution of library(DO.utils) in an active virtual environment should install the python module dependencies of DO.utils automatically. Should any python module need to be installed by a user, the modules used by DO.utils can be found in the “Config/reticulate” section of it’s DESCRIPTION file.