[DOCUMENTATION] Detailed installation instructions for non-conda users

[AnthonyOfSeattle]

NOT BLOCKING FOR PUBLICATION

This request is very specific to my situation. The current miniconda installation instructions were very straight forward and I was able to create this Dockerfile to give me an environment for testing.

FROM rocker/r-ver:4.4

RUN R -e "install.packages('reticulate')"
RUN R -e "reticulate::install_miniconda()"

RUN R -e "install.packages('BiocManager')"
RUN R -e "BiocManager::install(version = '3.20')"

RUN R -e "install.packages('remotes')"
RUN R -e "BiocManager::install('RforMassSpectrometry/ProtGenerics', ask=F)"
RUN R -e "BiocManager::install('RforMassSpectrometry/Spectra', ask=F)"
RUN R -e "BiocManager::install('RforMassSpectrometry/SpectriPy', ask=F)"

RUN R -e "install.packages('testthat')"

However, I have a second weird situation. Due to commercial restrictions, I am unable to use miniconda at work. So I was looking for some detailed instructions on how to setup the package using virtualenv instead. I found the toggle between the two setups in the code chunk bellow,

SpectriPy/R/zzz.R

Lines 68 to 98 in c823e95

	.install_python_packages <- function(envname = .spectripy_env(),
	use_conda = .spectripy_use_conda(), ...) {
	if (!py_module_available("matchms")) {
	packageStartupMessage("Installing required 'matchms' library")
	if (use_conda) {
	py_install(c("matchms==0.28.2"),
	envname = envname, method = "conda", pip = FALSE,
	channel = c("bioconda", "conda-forge"), ...)
	} else {
	py_install(c("matchms==0.28.2", "numpy==2.0.2"),
	envname = envname, method = "virtualenv",
	channel = c("conda-forge"), ...)
	}
	packageStartupMessage(
	"\nPlease restart R to load the freshly installed packages.\n")
	}
	if (!py_module_available("spectrum_utils")) {
	packageStartupMessage("Installing required 'spectrum_utils' library")
	if (use_conda) {
	py_install(c("spectrum_utils==0.3.2"),
	envname = envname, method = "conda", pip = FALSE,
	channel = c("bioconda", "conda-forge"), ...)
	} else {
	py_install(c("spectrum_utils==0.3.2", "numpy==2.0.2"),
	envname = envname, method = "virtualenv",
	channel = c("conda-forge"), ...)
	}
	packageStartupMessage(
	"\nPlease restart R to load the freshly installed packages.\n")
	}
	}

And I was able to find a reference to an the R variable that I needed to set here

SpectriPy/R/zzz.R

Lines 55 to 59 in c823e95

	.spectripy_use_conda <- function() {
	as.logical(getOption(
	"spectripy.use_conda",
	Sys.getenv("SPECTRIPY_USE_CONDA", unset = "TRUE")))
	}

But I couldn't seem to get all the way through tests. Could these instructions be added to the Documentation?

[jorainer]

Thanks for pointing this out @AnthonyOfSeattle ! I've added a link to the respective documentation for the Python configuration to the README (see latest commit referenced above).

For the docker image we use for our Metabonaut resource we set

## Use SpectriPy with virtual env to avoid need to install miniconda
ENV SPECTRIPY_USE_CONDA="FALSE"

before installing SpectriPy. See also the respective lines in the Docker file: https://github.com/rformassspectrometry/Metabonaut/blob/devel/Dockerfile#L18-L19

Let me know if your problem still persists.

[AnthonyOfSeattle]

Thanks for your help on this. I will need to circle back to it in the coming week but I am definitely interested in trying again.

[jorainer]

and any feedback is highly appreciated!

[jorainer]

Note: we have changed the default from conda/miniconda-based installation to virtualenv. It works well on unix and Windows - for macOS (at least in our GHA) we still have an issue.

[jorainer]

Note: we have completely changed the installation setup for SpectriPy in PR #82 . We're now using the new reticulate::py_require() functionality to install and manage Python requirements. This is independent of conda and does also not require any manual virtualenv installation. It works very nicely - only - seems there are some issues with reticulate/R/Python on recent macOS.