Allow empty multi-select-list and multi-select-hash

[innovate-invent]

There are a few situations where it would be useful to instantiate an empty array/object.

This is particularly important for reduce() to be able to pass an empty value as the init expression.
This could also be used with the arithmetic operators for a fallback operand, ie: (a || {}) + {foo: 5} in the event that a is undefined or otherwise falsey. A more advanced example: (type(a) == 'object' && a || {}) + {foo: 5}.

I believe there is room in the grammar to allow {} to represent an empty object, but [] currently refers to the flatten operator.

We could have to_array() accept zero arguments and return an empty array.
[`null`][] is an existing and valid way to get an empty array. This could be shortened to [a][] assuming 'a' is undefined in the context.
Are there any better ways to do this?

The proposed grammar change is:

multi-select-hash = "{" [ keyval-expr *( "," keyval-expr ) ] "}" 
multi-select-list = "[" ( expression *( "," expression ) / " " ) "]" 

[springcomp]

• You can instantiate an empty array with an empty multi-select-list, like so: [ ] (note at least a space between the opening and closing brackets, to prevent this from being tokenized to a flatten-expression.
  For instance, search( length([ ]), [1, 2, 3] ) -> 0

• We could indeed extend {} to be an empty hash-wildcard-expression object. Although that is currently possible with raw-string literals.
  For instance, search( `{}`, [1, 2, 3] ) -> {}

[innovate-invent]

Ha!, I don't know why raw string literals completely escaped me. I would prefer to move in a direction that allows us to eliminate them from the grammar though.

The space in the [ ] is a cool hack, I like it. It does violate the grammar rules:

multi-select-list = "[" ( expression *( "," expression ) ) "]" 

The first expression is required. I will add that to the proposal.

[springcomp]

You are right. I wondered if that was correct when I stumbled into it.
We definitely need to standardize this as part of the JEP.

Would you feel like writing the JEP draft here ?

I think it would:

• allow empty multi-select-list.
• specify a standard symmetric empty multi-select-hash object. like {}.

[springcomp]

I think empty multi-select-hash expression can be standardized in this JEP.

[springcomp]

This has been included in Compact Syntax for multi-select-hash