Content-Type: text/plain;
	charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3864.400.21\))
Subject: Re: Eliminating SPI / SQL from some RI triggers - take 3
From: Chao Li <li.evan.chao@gmail.com>
In-Reply-To: 
 <CA+HiwqH39QU7vGVx65JH1e0nzVvQc5eVmuY7=qyj0T_+b-HO3A@mail.gmail.com>
Date: Tue, 7 Apr 2026 09:45:42 +0800
Cc: Junwang Zhao <zhjwpku@gmail.com>,
 Haibo Yan <tristan.yim@gmail.com>,
 Pavel Stehule <pavel.stehule@gmail.com>,
 PostgreSQL-development <pgsql-hackers@postgresql.org>,
 Tomas Vondra <tomas@vondra.me>
Content-Transfer-Encoding: quoted-printable
Message-Id: <72AA2663-B642-4FB1-BDC2-5FAFF2D2DF15@gmail.com>
References: 
 <CA+HiwqF4C0ws3cO+z5cLkPuvwnAwkSp7sfvgGj3yQ=Li6KNMqA@mail.gmail.com>
 <CA+HiwqGM6nvAV5O+=Nr+BXMPWOma0oeCRVzVP0XiLE8zX5TVAg@mail.gmail.com>
 <CA+HiwqGMaovCUgDbGxVGnK0Mrivr+ph3YE2Ws+47-ugyPb4f7g@mail.gmail.com>
 <CAFj8pRDaiBe_GOLk_yyYHTtPiDAAaLOM8u1-=Q3ZgXBTH+1igg@mail.gmail.com>
 <CA+HiwqGA5Ay_MR0eJEEbt4j6WrVh4F+AasTp8yCbs5aJLOJn6Q@mail.gmail.com>
 <CAEG8a3JM=NoqiTK0V6S9FNxZPvy1+C5F7rfafTtPKBVWnunL-g@mail.gmail.com>
 <CA+HiwqEyiLCY6MTLbOJXDdLNNQLaURYHvdW797MQgbjEK9od4Q@mail.gmail.com>
 <CAEG8a3+VBpwPf1Rm-ECD90whM9b3YnGhux5CVXdsL6khiBfzRQ@mail.gmail.com>
 <CA+HiwqF2UHzF0sKCp-F2a-U29rqh_9ZPy=f1h+Fh_=M8efj3pg@mail.gmail.com>
 <CAEG8a3L9Ew-WL8sxLROVOcypeaENPmd8qCmMvz4geoGL1TDGCA@mail.gmail.com>
 <CAEG8a3+nUFQo4sdPQF9xy0J73J8RFJ5U9A5+_kMosGDaZ+1sXA@mail.gmail.com>
 <2BE661BA-D909-4093-BF78-DB9B0C099337@gmail.com>
 <CAEG8a3JyKdizWvYsF+z_mA1BKy=dpW11iKVMOG=bk6Tbz6M1Bw@mail.gmail.com>
 <CAEG8a3+Hf4tvvbts29_k_AFhWQmRYfEo_SW4C5FY_140iKghBw@mail.gmail.com>
 <CA+HiwqFV-PY-3BxM6j5TaAiC3AwedDxo-6vwRSbvygg3zF+xAQ@mail.gmail.com>
 <CA+HiwqHpaisS-e+0gAgzh92qZAFxncAJMmmTRZZN=efoeTPgOg@mail.gmail.com>
 <CA+HiwqFwZ6WLRbY8R7VC7JVp5Jot6ktZOkr9wDxTjoK=W1SgdQ@mail.gmail.com>
 <CA+HiwqF_Kz=R8juHJBiOATvabWSOugK4VaGOxoJ_n=E2c7UM9w@mail.gmail.com>
 <CA+HiwqHCB7kcbspkhaLN9enoj5x=ehzhFM4PXDgWUUP8Px41GA@mail.gmail.com>
 <77FA04FE-1F84-4DA1-8855-8BBFD8CC889A@gmail.com>
 <CA+HiwqHpVtP485wEKuXdOkdoZWhvVvfFH40-og07Jp3MPx21eg@mail.gmail.com>
 <CAEG8a3JWHkJSXe9nNcVK7wnYKUEqWuMGFDhy5BynB_9tEjmEUg@mail.gmail.com>
 <CA+HiwqFjfumKrWy03q5M309xJJVYt0WgGfH6AZ8BjFhSwppwsQ@mail.gmail.com>
 <CAEG8a3JjP1LaKSv-r3AMJLRyLMzENJrKshWsDvDouMPM_sizmA@mail.gmail.com>
 <CA+HiwqFQ+ZA7hSOygv4uv_t75B3r0_gosjadetCsAEoaZwTu6g@mail.gmail.com>
 <CA+HiwqHdB0r8z6Mut13BxpYNq2W-os+Arju4mcZbCyU9PeaVog@mail.gmail.com>
 <CAEG8a3K5ayVNkCDnK9OtNb+4OY0chJtW6ObgEOSFjqyymQda8Q@mail.gmail.com>
 <CA+HiwqGJYCgEs_F-LBtrRdx-Y969LMr-OVogjFXU6U-0q5bOwQ@mail.gmail.com>
 <CA+HiwqF2Ma6R_QyjfPBvFreYbezFpGcwASJE0a2bM+Y0jvof+g@mail.gmail.com>
 <C3ACBAE1-CC5F-4688-9828-4668CDE37F83@gmail.com>
 <CA+HiwqGFRAH2O5bGpNMErFopFyn-2-Zu3+5y+BFQim9TE8z+Pw@mail.gmail.com>
 <C2133B47-79CD-40FF-B088-02D20D654806@gmail.com>
 <CA+HiwqFx=aciJYkkaviyTutUm303QXz6GtXSqzG7nfd4MAzddQ@mail.gmail.com>
 <CA+HiwqH39QU7vGVx65JH1e0nzVvQc5eVmuY7=qyj0T_+b-HO3A@mail.gmail.com>
To: Amit Langote <amitlangote09@gmail.com>
Archived-At: 
 <https://www.postgresql.org/message-id/72AA2663-B642-4FB1-BDC2-5FAFF2D2DF15%40gmail.com>
Precedence: bulk


> On Apr 6, 2026, at 17:45, Amit Langote <amitlangote09@gmail.com> =
wrote:
>=20
> On Fri, Apr 3, 2026 at 6:39=E2=80=AFPM Amit Langote =
<amitlangote09@gmail.com> wrote:
>> On Fri, Apr 3, 2026 at 5:58=E2=80=AFPM Chao Li =
<li.evan.chao@gmail.com> wrote:
>>>> On Apr 3, 2026, at 13:52, Amit Langote <amitlangote09@gmail.com> =
wrote:
>>>> On Thu, Apr 2, 2026 at 5:00=E2=80=AFPM Chao Li =
<li.evan.chao@gmail.com> wrote:
>>>>> I plan to spend time testing and tracing this patch tomorrow. But =
I don=E2=80=99t want to block your progress, if I find anything, I will =
report to you.
>>>>=20
>>>> Sure, I didn't want to leave committing this to the weekend or the =
next week.
>>>=20
>>> I spent several hours debugging this patch today, and I found a =
problem where the batch mode doesn't seem to handle deferred RI =
triggers, although the commit message suggests that it should.
>>>=20
>>> I traced this scenario:
>>> ```
>>> CREATE TABLE pk (a int primary key);
>>> CREATE TABLE fk (a int references pk(a) DEFERRABLE INITIALLY =
DEFERRED);
>>> BEGIN;
>>> INSERT INTO fk VALUES (1);
>>> INSERT INTO pk VALUES (1);
>>> COMMIT;
>>> ```
>>>=20
>>> When COMMIT is executed, it reaches RI_FKey_check(), where =
AfterTriggerIsActive() checks whether afterTriggers.query_depth >=3D 0. =
But in the deferred case, afterTriggers.query_depth is -1.
>>>=20
>>> =46rom the code:
>>> ```
>>>        if (ri_fastpath_is_applicable(riinfo))
>>>        {
>>>                if (AfterTriggerIsActive())
>>>                {
>>>                        /* Batched path: buffer and probe in groups =
*/
>>>                        ri_FastPathBatchAdd(riinfo, fk_rel, newslot);
>>>                }
>>>                else
>>>                {
>>>                        /* ALTER TABLE validation: per-row, no cache =
*/
>>>                        ri_FastPathCheck(riinfo, fk_rel, newslot);
>>>                }
>>>                return PointerGetDatum(NULL);
>>>        }
>>> ```
>>>=20
>>> So this ends up falling back to the per-row path for deferred RI =
checks at COMMIT, even though the intent here seems to be only to bypass =
the ALTER TABLE validation case, where batch callbacks would never fire, =
and MyTriggerDepth is 0. So, maybe we can just check MyTriggerDepth>0 in =
AfterTriggerIsActive().
>>>=20
>>> I tried the attached fix. With it, deferred triggers go through the =
batch mode, and all existing tests still pass.
>>=20
>> I think you might be right.  Thanks for the patch.  It looks correct
>> to me at a glance, but I will need to check it a bit more closely
>> before committing.
>=20
> Thinking about this some more, your fix is on the right track but
> needs a bit more work -- MyTriggerDepth > 0 is too broad since it
> fires for BEFORE triggers too. I have a revised version using a new
> afterTriggerFiringDepth counter that I'll push shortly.
>=20
> Added an open item for tracking in the meantime:
> https://wiki.postgresql.org/wiki/PostgreSQL_19_Open_Items#Open_Issues
>=20
> --=20
> Thanks, Amit Langote
> <v2-0001-Fix-deferred-FK-check-batching-introduced-by-comm.patch>

V2 looks good to me. Besides the normal cases, I also traced an abnormal =
case to verify that afterTriggerFiringDepth is always reset to 0:
```
evantest=3D# begin;
BEGIN
evantest=3D*# INSERT INTO fk VALUES (2);
INSERT 0 1
evantest=3D*# commit;
ERROR:  insert or update on table "fk" violates foreign key constraint =
"fk_a_fkey"
DETAIL:  Key (a)=3D(2) is not present in table "pk".
```

Best regards,
--
Chao Li (Evan)
HighGo Software Co., Ltd.
https://www.highgo.com/