Conversation
4361080 to
ce61f1f
Compare
|
Fired up a new codespace based on this branch - got into a working vs code environment in about a minute where I could write ehrql with full syntax highlighting and autocomplete. The rest of postcreate.sh (which sets up the full Python and R environments) didn't finish until 6 mins later which is quite slow but at least everything else is responsive in the meantime. I've checked that the right version of Python loads (3.10) and all the expected libraries are available ( I've added an auto-starting Rstudio instance via |
Base on MS devcontainer Jammy image to match OpenSAFELY base image. Install R/Python and their dependencies using the respective requirements.txt from the `main` branch on github. This file is not copied to the built docker action images so we cannot copy it from their containers. Use `docker cp` to copy the R/Python library files to the host filesystem which avoids any of the previously seen issues with mounting a running container. However, these library directories don't contain any symlinks (verified with `find . -type l`) so one of the major issues with mounting is not of concern here so commented-out mount-based alternative is provided here which has better performance than the `docker cp` approach.
ce61f1f to
6940b03
Compare
re #147
Base on MS devcontainer Jammy image to match OpenSAFELY base image.
Install R/Python and their dependencies.
Use
docker cpto copy the R/Python library files to the host filesystem which avoids any of the previously seen issues with mounting a running container, however the issues we saw may not apply now we're installing the executables in the host system rather than trying to mount them and their dependencies.Mounting the libraries appears to be faster still, but there may be some more rough edges that are revealed with further testing. A sample implementation is included but commented out.