Record and Tuple creation should be recursive.

[caikan]

Record and Tuple are both deeply immutable, and there will never be a situation where mutable objects are passed in nested structures. So our syntax can be simplified as follows:

// before
const record = #{
  a: 1,
  b: 2,
  nested: #{
    tuple: #[1,2,3],
  },
}

// after
const record = #{
  a: 1,
  b: 2,
  nested: {
    tuple: [1,2,3],
  },
}

We can make #expression as a syntax suger of createRecordOrTuple(expression),

functoin createRecordOrTuple(data) {
  if (data && data[Symbol.iterator]) {
    return Tuple(...data)
  }
  if (data instanceof Object) {
    return Record(data)
  }
  return data
}

[ljharb]

Why would there never be such a situation, where someone attempts or intends to do that?

[acutmore]

I think this is a duplicate of #375

Applying the construction recursively could slow down creating the values, as the original mutable object would be created and then the 2nd record/tuple gets created. Having explicit syntax at the construction site makes this easier to avoid.

For an example like:

#{ p: [2,4] }

It looks like it is easy to optimise out the creation of the array and infer that the inner value should be a tuple but with indirection it becomes harder:

const getArr =  () => [2,4];
#{ p: getArr() };

More analysis is required to optimise out the array. This could lead to surprising performance differences between very similar code.

It may also result in errors, when invalid values are used; being thrown later due to the delay in converting the array/object to a tuple/record. Making it harder to spot the source of the invalid value.

[caikan]

I think that if recursive creation is not supported, it will not prevent people from trying to create nested Record/Tuple, but it will just make the code writing more cumbersome:

const getArr =  () => [2,4];
#{ p: #[...getArr()] };

[ljharb]

Sure. But it will make the code reading more clear, which is always more important than how hard it is to write.

[caikan]

Since it is deeply immutable, nested data must be converted to immutable. I don't think the more cumbersome way of writing it would make the code clearer to read. Also I wish this syntactic sugar could be used for any expression, not just object/array literals. I actually thought it would be more convenient to use # at the end of the expression:

const [head, ...rest] = [1,2,3]#
console.log(rest# === [2,3]#)

console.log(someObj.getSomeValue()# === targetTuple)