Dictionaries with no sense of self

[copumpkin]

I find that most of the time I'm writing jsonnet, having an object bring a magic self into scope is a bit of a pain. I usually end up having to define a local realSelf = self in the outer scope to retain the self that I actually wanted.

Can we have raw dictionaries with no magic self in scope for those times we just want to operate on a pure datastructure?

This is almost the opposite of the rec keyword in Nix. For context, rec brings a dictionary's keys into scope within that dictionary, and the default is to have no access to the equivalent of self. I'm not proposing changing the default at this stage, but it would be nice to have a way to opt out.

cc @roconnor, since whenever I mention Nix @sparkprime cc's him anyway 😄

[sparkprime]

I usually give them descriptive names so instead of local realSelf = self I use local service = self or whatever it is.

To be clear, what you're proposing is a syntax sugar:

{ inner: {{ x: self.y, y: 3 }}, y: 10 }

desugared via this rule:
{{ [f]: e, ... }} -> local fresh = self; { [f]: e[fresh / self] }

to

{ inner: local fresh = self; { x: fresh.y, y: 3 }, y: 10 }

which would manifest as:

{ "inner": { "x": 10, "y": 3 }, "y": 10 }

[sparkprime]

Assuming {{ }} is the syntax, which is probably unwise because of templating languages and their predilection for {{ }}

[igorpeshansky]

Note that this is exactly how ES6 arrow functions work (as compared to regular JS functions).

[sparkprime]

Technically, yes, but the presence of self binding in an object is significantly more legitimate / expected than in a closure.

[igorpeshansky]

Right. I meant that JS functions have a "this" binding in them (similar to JSonnet objects), which is very inconvenient (one has to use f.bind(this)). Your proposed sugar is similar to what ES6 tries to achieve with ()=> functions, which don't have a "this" binding, and thus capture one from the outer scope.
On an unrelated note, I wonder if you should even call those sugared blocks "objects" anymore.

[benley]

I would secretly be pleased if jsonnet used {{}} in its grammar, because it would discourage the use of text templating tools to generate jsonnet code ;-)