Simple Python installation for use with R/fafbseg/FlyWire

Installs Python via an isolated miniconda environment as well as recommended packages for fafbseg. If you absolutely do not want to use miniconda (it is much simpler to get started) please read the Details section.

Usage

simple_python(
  pyinstall = c("basic", "full", "extra", "cleanenv", "blast", "none"),
  pkgs = NULL,
  miniconda = TRUE
)

Arguments

pyinstall: Whether to do a "basic" install (enough for most functionality) a "full" install, which includes tools for accessing fast/simple skeletons and "dotprops" for NBLAST via the fafbseg-py package; "extra" installs neuron packages that enable high resolution skeletonisation that are really not necessary for most users and frankly a pain to install. "cleanenv" will show you how to clean up your Python environment removing all packages. "blast" will show you how to completely remove your dedicated miniconda installation. Choosing what="none" skips update/install of Python and recommended packages only installing extras defined by pkgs.
pkgs: Additional python packages to install.
miniconda: Whether to use the reticulate package's default approach of a dedicated python for R based on miniconda (recommended, the default) or to allow the specification of a different system installed Python via the RETICULATE_PYTHON environment variable.

Details

The recommended Python install procedure installs a miniconda Python distribution. This will not be added to your system PATH by default and can be used exclusively by R. If you do not want to use miniconda, then you should at least a) make a Python virtual environment using virtualenv (or conda if you are managing your own conda install) and b) specify which Python you want to use with the RETICULATE_PYTHON environment variable. You can set RETICULATE_PYTHON with usethis::edit_r_environ().

If you decided to stick with miniconda as recommended, some customisation is still possible. As a halfway house, you can set options('fafbseg.condaenv') to specify a non-standard miniconda virtual environment as an alternative to the default "r-reticulate". Furthermore you can set an environment variable RETICULATE_MINICONDA_PYTHON_VERSION=3.10 to use a newer version of Python than the reticulate package recommends.

If this sounds complicated, we strongly suggest sticking to the default miniconda=TRUE approach.

Note that that after installing miniconda Python for the first time or updating your miniconda install, you will likely be asked to restart R. This is because you cannot restart the Python interpreter linked to an R session. Therefore if Python was already running in this session, you must restart R to use your new Python install.

Examples

if (FALSE) {
# just the basics
simple_python("basic")
# if you want to skeletonise meshes
simple_python("full")

# To install a special package using the recommended approach
simple_python(pkgs="PyChunkedGraph")
# the same but without touching Python itself or the recommended packages
simple_python('none', pkgs='PyChunkedGraph')

# install a specific version of cloud-volume package
simple_python('none', pkgs='cloud-volume~=3.8.0')
# if you really need to insist (e.g. because a newer version is already installed)
reticulate::py_install('cloud-volume==8.15.0', pip = TRUE)

# install the latest version of a package from github
simple_python('none', pkgs="git+git://github.com/schlegelp/skeletor@master")

# install a specific earlier version of a package
simple_python('none', pkgs="git+git://github.com/seung-lab/DracoPy@v0.0.15")

# install all recommended packages but use your existing Python
# only do this if you know what you are doing ...
simple_python("full", miniconda=FALSE)
}