Identifiers that are reserved keywords should be disallowed.

[claymcleod]

The specification says (link):

The following (case-sensitive) language keywords are reserved and cannot be used to name declarations, calls, tasks, workflows, import namespaces, struct types, or aliases.

However, wdl is completely fine with reserved keywords being used as identifiers.

version 1.2

task run {
    input {
        String in
    }
}

This should be a validation error instead.

[peterhuene]

The reading of declarations in that context is ambiguous to me.

At the time this was implemented, it was assumed the specification meant private declarations and therefore it allows reserved WDL keywords in task/workflow inputs and outputs.

Thus the places we currently accept reserved names is:

• Struct member names and (by extension) literal struct expression key names
• Task and workflow input and output names
• Runtime/hints/requirement key names
• Metadata and parameter metadata key names
• Call inputs (necessary because reserved names are allowed for inputs)
• Name reference expressions (necessary to reference inputs and outputs)

We should get clarification from the spec.

For what it's worth, miniwdl does not fail to validate it either (but fails to parse for other keywords).

[claymcleod]

Yeah that makes sense. When you go look at the spec section on declarations, I feel that it's clear that unbound input/output declarations are included in that phrasing (e.g., the first example there in the spec specifically uses input declarations within their illustrative example). If you still feel that it needs clarification after re-reading that section, I can make an issue.

EDIT: I further see this now in the declarations section right above where the example with input/output declarations is used:

A declaration reserves a name that can be referenced anywhere in the scope where it is declared. A declaration has a type, a name, and an optional initialization. Each declaration must be unique within its scope, and may not collide with a reserved WDL keyword (e.g., workflow, or input).

Now I'm convinced there's really no room for allowing this. Let me know if you feel otherwise.

[peterhuene]

Agreed, that language is not ambiguous. I'll fix this.

The spec also uses "declarations" nomenclature around struct members, so I'm going to extend the prohibition on reserved keywords there as well.

[claymcleod]

Agreed, that language is not ambiguous. I'll fix this.

The spec also uses "declarations" nomenclature around struct members, so I'm going to extend the prohibition on reserved keywords there as well.

Yeah that sounds correct. Thanks