checkpointr - A tool to checkpoint in R

Checkpointr is a tool to assist storing intermediate values during long-running calculations. This package will either load a stored result or evaluate the code if any of the following conditions are satisfied:

The expression to be evaluated has changed
Any dependent variables have changed
The force parameter is set to TRUE

See the R documentation using ?checkpointr::checkpoint after installation for more information.

Installation

Installation depends on devtools, so once that's installed, the following command will fetch and install the checkpointr from git:

devtools::install_git("https://github.com/jasonrig/checkpointr.git")

Usage

See the sample code below for a simple demo of checkpointr.

x <- seq(0, 100)
y <- checkpointr::checkpoint({
  m <- mean(x)
  s <- sd(x)
  42
}, "test_checkpoint1")

After running the code above, y = 42, m = 50, and s = 29.30017. These values will be stored in checkpoint.dat. If this code were to be re-run, the expression inside the curly braces would be skipped.

The execution of the code can be further be triggered by changes to dependent variables:

x <- seq(0, 100)
y <- checkpointr::checkpoint({
  m <- mean(x)
  s <- sd(x)
  42
}, "test_checkpoint1", x)

In the sample code above, if the value of x changes from what was used during the initial checkpoint, the code in the braces will be re-evaluated and the checkpoint will be updated. Under the hood, the dependent variables are hashed and compared to the previously stored hash using the digest package. This means that any R objects can be used to define a dependency here.