Re: Schema variables - new implementation for Postgres 15 - Mailing list pgsql-hackers

From Pavel Stehule
Subject Re: Schema variables - new implementation for Postgres 15
Date
Msg-id CAFj8pRAmGr7rgWstsuX7zT0nmhD7Lwr8wfOdcYReb_44_EiSMQ@mail.gmail.com
Whole thread Raw
In response to Re: Schema variables - new implementation for Postgres 15  (Erik Rijkers <er@xs4all.nl>)
List pgsql-hackers


st 24. 8. 2022 v 10:04 odesílatel Erik Rijkers <er@xs4all.nl> napsal:
Op 24-08-2022 om 08:37 schreef Pavel Stehule:
>>
>
> I fixed these.
>

 > [v20220824-1-*.patch]

Hi Pavel,

I noticed just now that variable assignment (i.e., LET) unexpectedly
(for me anyway) cast the type of the input value. Surely that's wrong?
The documentation says clearly enough:

'The result must be of the same data type as the session variable.'


Example:

create variable x integer;
let x=1.5;
select x, pg_typeof(x);
  x | pg_typeof
---+-----------
  2 | integer
(1 row)


Is this correct?

If such casts (there are several) are intended then the text of the
documentation should be changed.

yes - the behave is designed like plpgsql assignment or SQL assignment

 (2022-08-25 19:35:35) postgres=# do $$
postgres$# declare i int;
postgres$# begin
postgres$#   i := 1.5;
postgres$#   raise notice '%', i;
postgres$# end;
postgres$# $$;
NOTICE:  2
DO

(2022-08-25 19:38:10) postgres=# create table foo1(a int);
CREATE TABLE
(2022-08-25 19:38:13) postgres=# insert into foo1 values(1.5);
INSERT 0 1
(2022-08-25 19:38:21) postgres=# select * from foo1;
┌───┐
│ a │
╞═══╡
│ 2 │
└───┘
(1 row)

There are the same rules as in SQL.

This sentence is not good - the value should be casteable to the target type.

Regards

Pavel





Thanks,

Erik

pgsql-hackers by date:

Previous
From: Ranier Vilela
Date:
Subject: re: postgres_fdw hint messages
Next
From: Tom Lane
Date:
Subject: V14 and later build the backend with -lpthread