[cafard]

Is this really a SQL feature or a relational feature? The essence of the relational model is that names are known up front: an attribute of one relation is not an attribute of another.

[dragonwriter]

> Is this really a SQL feature or a relational feature?

SQL.

There is no reason you couldn't have a composable implementaiton of relational logic.

> The essence of the relational model is that names are known up front

That every relation consist of tuples each member of which consists of a name, a type, and a value is part of the model, sure. That doesn't impose restrictions on composability.

> an attribute of one relation is not an attribute of another.

An attribute of one relation may well be an attribute of another.

There's no reason you couldn't have a relational language that let you store and reuse clauses.

[nendroid]

"attributes" are "generic" properties of relations.

So for example the union operator can take two different relations that have the same columns but different rows and create a new relation.

In this binary operation that takes two different relations.... the attributes of one relation (the columns) is indeed the attributes of the other relation.

See here: https://www.geeksforgeeks.org/basic-operators-in-relational-...

Look for "union" in the link above.

>Union on two relations R1 and R2 can only be computed if R1 and R2 are union compatible (These two relation should have same number of attributes and corresponding attributes in two relations have same domain)

It's the same thing for unary operations like select. A select expression can operate on different relations providing that the relations have the relevant attributes.