Request: Add Optional() function for Objects and Maps.

Hi there.

I would like to request an Optional() function for map-like objects. If a key is not present in an object, normally a keyError returns. Optional(object.key) will return None when the key is not present. I could not find such a thing in the current WDL implementation (but I may have overlooked).

That would be useful in these types of situations:

task exampleTask {
  String requiredVariable
  String? optionalString = "sensible default here"
  Int? optionalInt = 3
  Boolean? optionalBool = false
  command {
  # insertcommandhere
  }
}
workflow exampleWf {
  Array[Object] objectList
  scatter(exampleObject in objectList) {
    call exampleTask {
      input: 
        requiredVariable=exampleObject.requiredVariable
        optionalString=Optional(exampleObject.optionalString)
        optionalInt=Optional(exampleObject.optionalInt)
        optionalBool=Optional(exampleObject.optionalBool)
     }
  }
}

Originally (without optional) your inputs.json would look like this:

{
  "exampleWf.objectList":
  [
    {
      "requiredVariable": "Sample1",
      "optionalString": "someOption",
      "optionalInt": 3,
      "optionalBool": false
    },
    {
      "requiredVariable": "Sample2",
      "optionalString":  "sensible default here",
      "optionalInt": 5,
      "optionalBool": false
    }
  ]
}

But with Optional() it can look like this:

{
  "exampleWf.objectList":
  [
    {
      "requiredVariable": "Sample1",
      "optionalString": "someOption"
    },
    {
      "requiredVariable": "Sample2",
      "optionalInt": 5
    }
  ]
}

When there are 10 optional values, this save a lot of typing and unnecessary duplication in the JSON. Why specify all these values if they have sensible defaults? There are quite some command line tools with many optional parameters that this feature seems very useful to me. Real word example:

task centrifugeDownload {
    String centrifugeOutput
    Array[String] domain
    String? seqTaxMap
    String? database = "refseq"
    String? centrifugeDownloadExecutable = "centrifuge-download"
    String? assemblyLevel = "Complete Genome"
    String? refseqCategory =  "any"
    Array[String]? taxIds = ["any"]
    Boolean? filterUnplaced = false
    Boolean? maskLowComplexRegions = false
    Boolean? downloadRnaSeqs = false
    Boolean? modifyHeader = false
    Boolean? downloadGiMap = false

    command {
        ${centrifugeDownloadExecutable} \
        -o ${centrifugeOutput} \
        -d ${sep=',' domain} \
        -a "${assemblyLevel}" \
        -c ${refseqCategory} \
        -t ${sep=',' taxIds} \
        ${true='-r' false='' downloadRnaSeqs} \
        ${true='-u' false='' filterUnplaced} \
        ${true='-m' false='' maskLowComplexRegions} \
        ${true='-l' false='' modifyHeader} \
        ${true='-g' false='' downloadGiMap} \
        ${database} \
        ${">> " + seqTaxMap}
    }
 }

I happen to have some experience with programming in Scala, so if you give me some pointers, I could possibly implement this myself.

[patmagee]

By convention map values are optional, as there are no restrictions on what the key value can be. The key error is really an access error which is what you would expect to receive in the event of a missing value.

What I think would make a better approach would be a single engine function or modifying an existing one, to test for a key in a map.

task {
  Map[String,Int] my_values
  Int code = if contains_key(my_values,"key") then my_values["key"] else 0
  Int? code_2 = if contains_key(my_values,"key_2") then my_values["key"] else null
}

@geoffjentry @cjllanwarne has there ever been any talk of including a check like this?

[freeseek]

The following works a bit better:

scatter (key in keys(map)) { Boolean? is_key = if key == "xxx" then true else None }
if (length(select_all(is_key))>0) { ... }