Replace experimental function with parallel step

The Workflows experimental function, experimental.executions.map, starts a workflow execution for each corresponding argument, and waits for all of the executions to finish, returning a list where each element is the result of an execution.

If you are using experimental.executions.map to support parallel work, you can migrate your workflow to use parallel steps instead, executing ordinary for loops in parallel.

A parallel step defines a part of your workflow where two or more steps can execute concurrently. A parallel step waits until all the steps defined within it have completed or are interrupted by an unhandled exception; execution then continues. Like experimental.executions.map, the execution order is not guaranteed. For details, see the syntax reference page for parallel steps.

Note that the use of experimental.executions.map or workflows.executions.run requires additional concurrent executions quota. However, when using parallel steps with calls to connectors inlined (see the translation connector example), no additional execution quota is required.

The following examples are intended to assist you when replacing the use of experimental.executions.map with a parallel step.

Translation workflow

Given a source and target language, the following workflow, named translate, uses the Cloud Translation connector to translate some input text and return the result. Note that the Cloud Translation API must be enabled.

  main:
    params: [args]
    steps:
    - basic_translate:
        call: googleapis.translate.v2.translations.translate
        args:
          body:
            q: ${args.text}
            target: ${args.target}
            format: "text"
            source: ${args.source}
        result: r
    - return_step:
        return: ${r}

  {
    "main": {
      "params":