Please support writing map type directly in SQL

[tigercl]

What would you like to be added or enhanced?

Write directly in SQL to construct the map type, such as this:

SELECT
     {'a': {'b': 1}} as t1,
     map_put('b', {'c': 1}, t1) as t2
FROM "t/#"

Why is this needed?

All data types in the rule engine except map can be written directly in SQL, for example:

'a' as t1,
1 as t2,
[t1, t2] as t3

But for the map type, I can't write like this:

{'a': 1} as t1

This causes the SQL to become difficult to write and read when we want to construct a new map based on the fields extracted above:

SELECT
  '1' as t1,
  '2' as t2,
  '3' as t3,
  map_put('t3', t3, map_put('t2', t2, map_put('t1', t1, map_new()))) as t4
FROM
  "t/#"

or

SELECT
  '1' as t1,
  '2' as t2,
  '3' as t3,
  map_put('t1', t1, map_new()) as t4,
  map_put('t2', t2, t4) as t5,
  map_put('t3', t3, t5) as t6
FROM
  "t/#"

In addition, the introduction of map types in the documentation of built-in functions also seems incompatible with other types:

length([1,2,3,4]) = 4
map_get('b', map_put('b', 1, json_decode('{"a": 1}'))) = 1

# Rather Than:
map_get('a', {'a': {'b': {'c': 1}}}) = {'b': {'c': 1}}

I think there should not be a situation where a certain basic type cannot be written in a complete grammar system.

[zmstone]

string interpolation in general basically?
does this work good enough?

SELECT
  interpolate('{"a": ${payload.a}, "b": ${payload.b}}') as my_json_obj

NOTE: there is no such function yet.

[tigercl]

Actually I'm more looking forward to this format:

SELECT
    {'key': value}