If you don’t yet have Mac OSX command line developer tools, please install it using:

  xcode-select --install

If your R does not find the correct python version, it will complain that umap-learn is not installed and ask you to install it. Here are some tips on how to find the correct python version that was installed in the conda environment.

Try selecting the correct conda env in R

In this example the conda environment is named scRNAseq2021.

  library(reticulate)
  reticulate::use_conda("scRNAseq2021")

Then check what python you have in R:

  reticulate::py_config()
  # should read at top:
  python:         /Users/asbj/miniconda3/envs/scRNAseq2021/bin/python

If that still is not right, you may have an r-reticulate python installation as well and need to perform the steps below.

Restart R and select python version

OBS! Before doing anything else you need to select python version.

First, find out what path you have to your conda python (in TERMINAL):

  which python
  /Users/asbj/miniconda3/envs/scRNAseq2021/bin/python

Then in R (after restarting):

  reticulate::use_python("/Users/asbj/miniconda3/envs/scRNAseq2021/bin/python", required=T)

Then check again with py_config if correct version of python is used:

  reticulate::py_config()

If you have the correct version now, you should be able to run UMAP without issues.

You can install stringi in R using:

  install.packages('stringi')

This is a problem with the MacOSX compiler, in which conda is unable to find it.

  #Download MacOSX10.9.sdk from Github
  curl -o MacOSX10.9.sdk.tar.gz "https://github.com/phracker/MacOSX-SDKs/releases/download/11.3/MacOSX10.9.sdk.tar.xz"

  #extract
  sudo tar -xzf MacOSX10.9.sdk.tar.xz

  #copy
  sudo cp -r MacOSX10.9.sdk /opt/

  #give executable permissions
  sudo chmod -R a+rX /opt

  #Link the path where conda looks to where the file is
  ln -s /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk /opt/MacOSX10.9.sdk

This error happens when running code inline.

One possible solution is to restart Rstudio and type.

  if(interactive()) { options(error = utils::recover)}

Please try other solutions listed here: https://github.com/rstudio/rstudio/issues/4723

If none of those work, you can click on the wheel engine symbol and check Chunk output in console

If R crashes due to memory issues, it may be a good idea to increase the vector size R_MAX_VSIZE. Put in the file .Renviron either in your home directory or the folder you are launching Rstudio from:

  R_MAX_VSIZE=70Gb

Or to whatever value matches your computer, the default size is 16Gb.