In the geospatial_plus stack, python user library is altered to /pseudohome via specification of some environment variables. (/pseudohome doesn't exist and needs to be mounted on startup). This makes it so new packages installed after build are put into an singularity instance/exec-specific folder and don't clash with the default python package installation location in the host home directory (assuming you bind /pseudohome) when calling the image.
However, rstudio server doesn't/respect inherit any system environment variables. In the R REPL itself this can be solved by adding the same environment variables added in #23 to .Renviron (added in 13be89dc9f776527cd41bfca9366c92fdc894957 to Renviron.site), but the Rstudio terminal doesn't source .Renviron. So this means if you install python packages via the rstudio server terminal window they'll install to the default home locaiton on host which will clash with packages installed by other containers (including the host OS).
If you need to pip install to the user directory, I recommend using either singularity shell, or perhaps using a system command in R. Note that you can open a shell of an already-started instance.
In the geospatial_plus stack, python user library is altered to /pseudohome via specification of some environment variables. (/pseudohome doesn't exist and needs to be mounted on startup). This makes it so new packages installed after build are put into an singularity instance/exec-specific folder and don't clash with the default python package installation location in the host home directory (assuming you bind /pseudohome) when calling the image.
However, rstudio server doesn't/respect inherit any system environment variables. In the R REPL itself this can be solved by adding the same environment variables added in #23 to .Renviron (added in 13be89dc9f776527cd41bfca9366c92fdc894957 to Renviron.site), but the Rstudio terminal doesn't source .Renviron. So this means if you install python packages via the rstudio server terminal window they'll install to the default home locaiton on host which will clash with packages installed by other containers (including the host OS).
If you need to pip install to the user directory, I recommend using either singularity shell, or perhaps using a system command in R. Note that you can open a shell of an already-started instance.