piktests runs computations like preprocessings for various models in an isolated runtime environment. The most basic piktests run can be started with:
This will create a so called runfolder which contains everything
related to this piktests run. To isolate the package library a new renv is created for each
piktests run. A new mainfolder is created to isolate from the system
wide madrat setup, but
the system wide sourcefolder and mappingfolder are still used. By
default the system wide cachefolder is not used, but this can be changed
by passing localCache = FALSE. After setup, by default the
magpie and remind preprocessing are run, via SLURM if available,
otherwise in a separate R session.
A common use case for piktests is running a computation using a development version of a package. Assuming the development version to be used is located on GitHub at https://github.com/pfuehrlich-pik/madrat:
piktests::run(renvInstallPackages = "pfuehrlich-pik/madrat",
computations = piktests::baseComputations["madratExample"])This will run renv::install("pfuehrlich-pik/madrat")
during setup and then run the madratExample computation using this
madrat version. The renvInstallPackages argument is passed
directly to renv::install, which also allows passing
multiple package specifications,
e.g. c("pfuehrlich-pik/madrat", "[email protected]").
Often it is helpful to compare the logs of a piktests run using the default package versions to the logs of a piktests run with alternate package versions. To do so:
piktests::runWithComparison(renvInstallPackages = "pfuehrlich-pik/madrat",
computations = piktests::baseComputations["madratExample"])This will create two piktests runs: One using default packages and
another one for the alternate packages, in this case
pfuehrlich-pik/madrat. For each computation a shell script
is created that can be used to easily compare/diff the logs of both runs
after they are finished: