Confusing semantics when adding if ... to a partial set rule

[anderseknert]

From the perspective of a clueless indivdual — and that would be me :) — it is difficult to understand the semantics of these expressions, and how adding if true to a rule head would alter the value/type of the rule(!)

a["b"]

As expected, this creates a set containing the value "b".

a["b"] if true

This creates a map with the key b and the value true. That is... not what I expected, as to me the if in a["b"] if true would be syntax sugar for a["b"] { true }, which apparently it isn't.

Now, I'm well aware that using contains would be the correct way to work with partial sets in ref-head rules, but given how that came out of backwards compatibility concerns, and if also is a new keyword, it's hard to imagine backwards compatibility being the problem here.

The formatter also rewrites a["b"] if true -> a["b"] = true which to me is not semantically equivalent. I've tried finding anything on this in the docs but I've come up short.

[anderseknert]

A side effect of this — a trick we use in Regal to distinguish:

a { true }

from

a = true { true }

Is to simply check the rule head value for the presence of a location attribute. Only in the latter case will there be a location, since in the former case the implied assignment is "generated" and cannot be traced to a location in the source code.

This "trick" however breaks for the beforementioned case, as a location for the value is added for the true value, despite no value being present on that location in the actual source code:

a["b"] if true

opa parse --format json --json-include locations p.rego

{
  "head": {
    "value": {
      "location": {
        "file": "p.rego",
        "row": 5,
        "col": 1
      },
      "type": "boolean",
      "value": true
    }
  }
}

[stale]

This issue has been automatically marked as inactive because it has not had any activity in the last 30 days. Although currently inactive, the issue could still be considered and actively worked on in the future. More details about the use-case this issue attempts to address, the value provided by completing it or possible solutions to resolve it would help to prioritize the issue.

[tsandall]

Can we close this in favor of #6247?

[anderseknert]

Sure. Just so that I understand, while we will require contains and if, we don't want to require if for a rule without conditions, right? Provided the original example, I'd expect this:

a["b"]

to become this

a contains "b"

Without an if added / required. Is that correct?

[tsandall]

Yes!