reticulate: Is there any reason to install *conda when running R/RStudio on Ubuntu 22.04?

After a major problem trying to install a package that didn't want to use a *conda repository, installed automagically by reticulate, I realized (prompted by Dirk Eidenbuettel) that ubuntu 22.04 has a perfectly good python3 already installed. Given that there appears to be a conflict between some R packages and the *conda installed library, I wonder whether one even needs to install the *conda that seems to be the default python library for reticulate. Is it possible just to point reticulate directly to the existing python3 libraries? Would there be any problems with other reticulate functions doing this?
Larry Hunsicker

1 Like

It is not necessary to use conda. See here for examples without it. Installation functions like py_install() do have a method argument to refuse conda and force using pip.

I know a few people who only use their system installation of Python for work, and don't seem to have run into horrible problems. That being said, there is a tendency in Python that some package can have poor forward compatibility and require a specific, older, version of Python/numpy/other. To avoid conda, you can still use virtual environments, that will still give you versioning but you can use pip. Virtualenv is supported by reticulate.

For more details, Stack Overflow has multiple questions of pip vs conda vs virtualenv etc, e.g. 1, 2, 3, ...

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.

Note that as of reticulate v1.31, reticulate no longer prompts users to install conda, nor does it default to using conda if a standard python installation is present.

There is no need to install conda, and as of v1.31, using reticulate with conda is strictly opt-in.