targets (version 0.0.0.9000)

tar_workspace: Load a saved workspace and seed for debugging.

Description

Load the packages, workspace, and random number generator seed of an errored target attempted with error = "save". Remove workspace files with tar_undebug() when you are done debugging.

Usage

tar_workspace(name, envir = parent.frame(), packages = TRUE)

Arguments

name

Symbol, name of the target whose workspace to read.

envir

Environment in which to put the objects.

packages

Logical, whether to load the required packages of the target. Uses require() with lib.loc equal to the library setting of the target (from tar_target(library = "PATH_TO_LIB") or tar_option_set(library = "PATH_TO_LIB")).

Value

the function returns no value, but it does load the target's required packages, as well as multiple objects into the environment (envir argument) in order to replicate the workspace where the error happened. These objects include the global objects at the time tar_make() was called, the dependency targets, the random number generator seed (assigned to .targets$seed and set with set.seed()) and the traceback of the error (assigned to .targets$traceback).

Details

If you set error = "save" in tar_option_set() or tar_target(), then if that target throws an error in tar_make(), it will save its workspace to a compressed file in _targets/workspaces/. The workspace includes the target's required packages, the global objects at the time tar_make() was called, the dependency targets, the random number generator seed (assigned to .targets$seed and set with set.seed()) and the traceback of the error (assigned to .targets$traceback). tar_workspace() loads the packages, populates the environment with the objects, and sets the seed to the seed of the target. Workspace files can be large sometimes, so it is good practice to remove them with tar_undebug() when you are done debugging.

Although useful, this behavior does not perfectly replicate what tar_make() does to set up the runtime environment for a target. tar_make() creates a formal hierarchy of environments that inherit from one another in order to organize data and contain side effects.

In addition, sometimes, you may not be able to debug with error = "save". Workspace files are saved with qs::qsave(), so if one of your dependencies cannot be properly serialized this way (e.g. Keras models you store with format = "keras") then the workspace cannot be saved properly. If this happens to you, either avoid non-exportable objects or use interactive debugging. (See the debug argument of tar_option_set().)

Examples

# NOT RUN {
tmp <- sample(1)
tar_script({
  tar_option_set(error = "save") # Required for saving workspaces.
  tar_pipeline(
    tar_target(x, "loaded"),
    tar_target(y, stop(x))
  )
})
try(tar_make())
exists("x") # Should be FALSE.
tail(.Random.seed)
tar_workspace(y)
exists("x") # Should be TRUE.
print(x) # "loaded"
tail(.Random.seed) # Should be different.
tail(.targets$traceback, 1)
# }