Improper tasks

In the previous chapter, we talked about task ("proper tasks") which consist of sys commands (i.e. shell commands). As we mentioned, these tasks can be executed on different systems: local, cluster, cloud, etc.

It is also possible to execute arbitraty bds code within a task, we call these "improper tasks". For example:

a := 42
task(cpus := 1) {
    # This can be any bds code
    println "Hello, a=$a"
}

How it works

In case of an "improper task", bds will create a checkpoint containing an absolute serialization of the whole program execution state. This works similar to a checkpoint statement. Then a task will be scheduled to execute bds restoring from this checkpoint. As a result, the program executing the task will "continue" in the next instruction, i.e. executing the code within the task statments.

Note: An exit statement is implicitly added in the checkpoint to avoid the task to execute beyond the statements within task. This means that the code saved to the checkpoint file created by the improper task would be:

a := 42
task(cpus := 1) {
    # This can be any bds code
    println "Hello, a=$a"   # Note: Variable's "a" value is visible because we inherit the full program state from the checkpoint
    exit                    # Note: This 'exit' statment is added only for the checkpoint to be executed by the task
}

Features

Improper task can have all the same feature as "regular" tasks:

• Task dependencies using files
• Task dependencies using task IDs
• Task parameters, such as cpus, mem, system, etc.
• Task names
• Multiple task dependencies, even on different systems
• dep & goal: You can also create improper dep and goal