Why does postgres distinct assign a type to strings?

Question

If you execute SELECT 'test', 123 the output will be two columns, the first with an unknown data type and the second with a data type of integer.

If you execute SELECT DISTINCT 'test', 123 the output will be two columns, the first with a data type of text and the second with a data type of integer.

Why does adding DISTINCT function different from running the query without DISTINCT in regards to the data type?

Answer 1

This behaviour is described in Chapter 10. Type Conversion.

In 10.1. Overview you can find:

If a type is not specified for a string literal, then the placeholder type unknown is assigned initially, to be resolved in later stages.

To select distinct values Postgres has to convert string literals to a type with equality operator. The case is analogous to union:

select 'abc', 1
union
select 'def', 1

where the first column is resolved as text. The rule is described in 10.5. UNION, CASE, and Related Constructs:

3. If all inputs are of type unknown, resolve as type text (the preferred type of the string category).