'input' and 'data' should be reserved identifiers

[jpeach]

Expected Behavior

If I define a function whose name is data, the compiler should probably emit an error that directly specifies that's not going to work.

Actual Behavior

I ended up with this compiler error:

 rego_type_error: undefined function data.contour.configmap.data.contour.configmap.data.contour.resources.get

Steps to Reproduce the Problem

package contour.resources.configmap

import  data.contour.resources

data(name) = obj {
  config := resources.get("configmaps", name)
  obj := object.get(config, "data", {})
}

Additional Info

I'm using OPA v0.19.2 as a library.

[tsandall]

We currently allow policies to override built-in functions as well as the input and data global variables.

input overriding can actually be quite useful during development/debug cycles if you want to quickly replace the value of input inside of a package for test purposes. data overriding is less useful because you're unlikely to replace the value of data correctly.

In this case, I'm assuming the import data.countour.resources was resolved to the local data function (so the reference data.contour.resources became data.contour.resources.configmap.data.contour.resources like in the error message.) I would have expected imports to be fully qualified so it's arguable this is a bug and should be fixed. That said, I'm not sure how I feel about disallowing overriding entirely.

cc @timothyhinrichs WDYT?

[timothyhinrichs]

Thanks for the reminder @tsandall that supporting overriding for functions is necessary to ensure backwards compatibility of policies. I could see prohibiting overriding data (especially) and input (possibly). Tooling around the playground, vscode plugin, etc. has improved to the point where I don't personally use input overriding anymore for debugging.