So @josh ... does this mean that if Rstudio hasn't built and cached a package you may not be able to install it? I use a miniCRAN repo hosted on github for extra control over what packages are used (see link below). Is there a way that I can build these packages for Rstudio cloud myself?
Split from topic: I can't install packaged Xgboost - #4
You are welcome to install packages from where ever you like.
We build and cache packages from CRAN that either take a long time to compile, or require more memory than is currently available on rstudio.cloud and therefore are not able to be compiled.
As long as the packages you are installing from other repositories don't exceed the memory limit, you shouldn't have an issue.
Thanks for the reply @josh. Several of the packages I want to use do exceed memory limits and may not be the version that is available pre-built in Rstudio Cloud. Is there a way that I can pre-build these packages for Rstudio cloud myself?
Alternatively, it would be great if there was an option like in python to use a requirements.txt file where you state something like the below and Rstudio Cloud would have different versions of packages available pre-built:
XGBoost == 0.6.4.1
Sorry, currently there is no supported way for users to pre-build packages.
I expect the package building will see an improvement to build specific versions of packages not in the cache, so if one used packrat with a project to capture specific versions of CRAN packages it would be able to take advantage. But I don't have a timeframe to communicate just now.