MIME-Version: 1.0
References: 
 <CAE+cgNgVWChqF-f-s4zT18V+oK1y9UOvORq9JX7jFKj4=r_taw@mail.gmail.com>
 <20260311.191117.172824867943979935.ishii@postgresql.org>
 <CAE+cgNgSG6oiAXT=FL+gK71Squ_eacyTCxSVXaAUTwODrkzqFg@mail.gmail.com>
 <20260318.204139.106185544494720182.ishii@postgresql.org>
In-Reply-To: <20260318.204139.106185544494720182.ishii@postgresql.org>
From: Zsolt Parragi <zsolt.parragi@percona.com>
Date: Wed, 18 Mar 2026 23:06:59 +0000
Message-ID: 
 <CAN4CZFOw4X27zgYdUgbVMmT3x_22a9kdGnTe6U+2TV1jJv-1ow@mail.gmail.com>
Subject: Re: Row pattern recognition
To: Tatsuo Ishii <ishii@postgresql.org>
Cc: assam258@gmail.com, sjjang112233@gmail.com, vik@postgresfriends.org,
	er@xs4all.nl, jacob.champion@enterprisedb.com, david.g.johnston@gmail.com,
	peter@eisentraut.org, pgsql-hackers@postgresql.org
Content-Type: text/plain; charset="UTF-8"
Archived-At: 
 <https://www.postgresql.org/message-id/CAN4CZFOw4X27zgYdUgbVMmT3x_22a9kdGnTe6U%2B2TV1jJv-1ow%40mail.gmail.com>
Precedence: bulk

Hello

I found one more bug, the following testcase crashes the server:

CREATE TEMP TABLE t (id int, val text);
INSERT INTO t VALUES
  (1,'A'),(2,'B'),
  (3,'A'),(4,'B'),
  (5,'A'),(6,'B'),
  (7,'A'),(8,'B'),
  (9,'X');

SELECT id, val, count(*) OVER w AS match_count
FROM t
WINDOW w AS (
  ORDER BY id
  ROWS BETWEEN CURRENT ROW AND UNBOUNDED FOLLOWING
  INITIAL
  PATTERN (A B (A B){1,2} A B)
  DEFINE
    A AS val = 'A',
    B AS val = 'B'
);

Where the code does

+ child->min += 1;

It should also increment max if it's not infinity.

if (child->max != RPR_QUANTITY_INF)
    child->max += 1;


+ expr = te->expr;
...
+ attno_map((Node *) expr);

It it okay to mutate an object in the plan cache in the executor?
Wouldn't it be better to first copy it?

> Adds stock trading scenario tests using realistic synthetic data
> (stock.data with 1632 rows). Tests V-shape recovery, W-shape,
> consecutive rises, and other common pattern matching use cases.

The data file is not included in the patchset

+ if (wc->frameOptions & FRAMEOPTION_GROUPS)
+ ereport(ERROR,
+ (errcode(ERRCODE_SYNTAX_ERROR),
+ errmsg("FRAME option GROUP is not permitted when row pattern
recognition is used"),
+ errhint("Use: ROWS instead"),
+ parser_errposition(pstate,
+ windef->frameLocation >= 0 ?
+ windef->frameLocation : windef->location)));

Here and at a few other places, shouldn't this use a different error
code, like ERRCODE_WINDOWING_ERROR?

Also a typo, the error message should say GROUPS.

+ elog(ERROR, "PREV/NEXT must have 1 argument but function %d has %d args",
+ func->funcid, nargs);

Isn't this a user triggerable error, which should have a proper error code?