Loops

Loops in matflow can be used when you want to iterate over one or more parameters.

Loops versus sequences

If you simply want to run a task multiple times with potentially different parametrisations, you can do this with a sequence. Loops are used when there should exist some dependency between the different “repeats” (or iterations) of a given task (or subset of tasks).

A basic loop workflow

Consider a task schema that takes one input parameter, p1, and outputs the same parameter, p1, with it’s value incremented by 1:

objective: my_task
inputs:
  - parameter: p1
outputs:
  - parameter: p1
actions:
  - commands:
      - command: echo $(( <<parameter:p1>> + 1 ))
        stdout: <<int(parameter:p1)>>

Note that this task schema’s output parameter is the same as its input parameter. In other words, it is a parameter-modifying schema, with respect to the parameter p1. This means if we use this schema in a workflow, we can define a loop that iterates the parameter p1:

loops:
  - name: my_loop
    tasks: [0]
    num_iterations: 2
tasks:
  - schema: my_task
    inputs:
      p1: 101

In the example workflow template above, we define a loop called my_loop that loops over the task with index 0 (the only task, in this case) two times. Since no default value for the input parameter p1 is provided in the task schema, we must provide a “local” value in the template. This value will be used for the first iteration of the loop, but for subsequent iterations, the p1 input value will be sourced from the output of the previous iteration.

This workflow will progress in the following way:

Task	loop iteration index	p1 input value	p1 output value
my_task	0	101	102
my_task	1	102	103