If the input variable is an array, we must include an array separator. In WDL 1.0, this is done using the sep= expression placeholder. Every value in the WDL Array[String] will be separated by whatever value is declared via sep. In this example, that is a simple space, as that is one way how to construct a bash variable.

task count_words {
  input {
    Array[String] a_big_sentence
  }
  command <<<
    ARRAY_OF_WORDS=(~{sep=" " a_big_sentence})
    echo ${#ARRAY_OF_FILES[@]} >> length.txt
    # Note how the bash array uses ${} syntax, which could quickly get
    # confusing if we used that syntax for our WDL variables. This is
    # why we recommend using tilde + {} for your WDL variables.
  >>>
}

It's usually unnecessary to declare an Array[String], because a single String can have many words in it. That being said, an Array[String] can sometimes come in handy if it is made up of outputs from other tasks. We'll talk more about chaining tasks together in upcoming chapters.

::: {.notice data-latex="warning"} The WDL 1.1 spec added a new built-in function, sep(), which replaces the sep= expression placeholder for arrays. This same version of the spec also notes that the sep= expression placeholder are deprecated and will be removed from future versions of WDL. For the time being, we recommend sticking with sep= as it is compatible with both WDL 1.0 and WDL 1.1, even though it is technically deprecated in WDL 1.1. :::

If you're not used to working in bash, the syntax for interacting with bash arrays can be unintuitive, but you don't have to write a WDL's command section only using bash. In fact, working in another language besides bash within a WDL can be a great way to write code quickly, or perform tasks that are more advanced than what a typical bash script can handle. Just be sure to set sep properly to ensure that your array is interpreted correctly. In this example, we place quotation marks before and after the variable to ensure that the first and last value of the list are given beginning and ending quotation marks respectively.

task count_words_python {
  input {
    Array[String] a_big_sentence
  }
  command <<<
    python << CODE
    sentence = [ "~{sep='", "' a_big_sentence}" ]
    print(len(sentence))
    CODE
  >>>
  runtime {
    docker: "python:latest"
  }
}