Managing Python environments within Bioconductor

Environment	Status
BioC-release
BioC-devel

basilisk provides a standardized mechanism for handling Python dependencies within Bioconductor packages. It does so by automatically provisioning and managing one or more Conda environments per BioC package, ensuring that the end-user is not burdened with the responsibility of meeting any Python-based SystemRequirements. We integrate with reticulate to allow intuitive calling of Python code within R, with additional protection to ensure that multiple Python environments can be called within the same R session.

Most "users" of this package are expected to be Bioconductor package developers; end users should not need to interact with the basilisk machinery, all going well. Users can follow the typical installation process for Bioconductor packages:

install.packages("BiocManager") # if not already installed
BiocManager::install("basilisk")

# Bioconductor package developers may prefer to use the devel version:
BiocManager::install("basilisk", version="devel")

The vignette provides instructions on how to adapt a client package to use basilisk. A minimal example is provided in the inst/example directory and contains code like:

# Provision an environment.
my_env <- BasiliskEnvironment(envname="my_env_name",
    pkgname="name.of.package",
    packages=c("pandas==0.25.1")
)

# Run reticulate code using that environment.
res <- basiliskRun(env=my_env, fun=function(args) {
    out <- reticulate::import("pandas")
    # Do something with pandas
    return(some_r_object)
})

Detailed documentation for each function is available through the usual methods, i.e., ?basiliskRun. See the Bioconductor landing page for more links; some examples of basilisk client packages include crisprScore and velociraptor.

Bugs can be posted to the Issues of this repository. Pull requests are welcome.

bvieth/basilisk

Managing Python environments within Bioconductor