clarification on chaining of set operations

Поиск
Список
Период
Сортировка
От PG Doc comments form
Тема clarification on chaining of set operations
Дата
Msg-id 163338891727.12510.3939775743980651160@wrigleys.postgresql.org
обсуждение исходный текст
Ответы Re: clarification on chaining of set operations  (Tom Lane <tgl@sss.pgh.pa.us>)
Список pgsql-docs
The following documentation comment has been logged on the website:

Page: https://www.postgresql.org/docs/12/queries-union.html
Description:

Apologies if this has been raised previously - I searched the archives and
did not find anything.

Regarding section 7.4 of the documentation.  I was curious about the
behavior when combining different set operations in one query, e.g, 
query1 op1 query2 op2 query3;
where op1 and op2 are one of UNION, INTERSECT, EXCEPT.

The documentation suggests that this is equivalent to 
(query1 op1 query2) op2 query3;
but only states it for the case when op1 = op2 = UNION.  (Which really
doesn't matter anyway given that set union is associative.)

In fact, mixing INTERSECT and UNION does not exhibit the expected behavior
(at least in version 12 - I have not tested in later versions).  Here's my
counterexample:

CREATE TABLE test1 (x INTEGER);
CREATE TABLE test2 (x INTEGER);
CREATE TABLE test3 (x INTEGER);

INSERT INTO test1 VALUES (1), (2);
INSERT INTO test2 VALUES (2), (3);
INSERT INTO test3 VALUES (3), (4);

SELECT * FROM test1
UNION
SELECT * FROM test2
INTERSECT
SELECT * FROM test3;

(SELECT * FROM test1
UNION
SELECT * FROM test2)
INTERSECT
SELECT * FROM test3;

The observed behavior suggests that INTERSECT is applied before UNION.  I
haven't tested to figure out other interactions (e.g., with EXCEPT or the
ALL variations).

В списке pgsql-docs по дате отправления:

Предыдущее
От: Bruce Momjian
Дата:
Сообщение: Re: [PATCH] fix ICU explorer link in locale documentation
Следующее
От: Tom Lane
Дата:
Сообщение: Re: clarification on chaining of set operations