> As a really basic example: `SELECT <expr> FROM tbl` — can we tell what shape the result is?
Like most statically analyzable code, you have problems if you try to statically analyze part of the code without the relevant definitions.
But, yes, if you have the relevant code (e.g., the DDL for the table), you can.
(Without it, you can in the sense that you can statically determine it as a function of the table definition, which can be sufficient in some cases.)
> in SQL, shapes / types require a lot of context — the result could be a single row in the case of `SUM(foo)`, or it could be every row in the case of `foo, bar`.
Oh, you mean, can we statically determine the shape of an expression’s results without knowing the expression? Well, no, and that’s true in most statically-analyzable languages.
Like most statically analyzable code, you have problems if you try to statically analyze part of the code without the relevant definitions.
But, yes, if you have the relevant code (e.g., the DDL for the table), you can. (Without it, you can in the sense that you can statically determine it as a function of the table definition, which can be sufficient in some cases.)
> in SQL, shapes / types require a lot of context — the result could be a single row in the case of `SUM(foo)`, or it could be every row in the case of `foo, bar`.
Oh, you mean, can we statically determine the shape of an expression’s results without knowing the expression? Well, no, and that’s true in most statically-analyzable languages.