Date: Fri, 17 Jan 2020 12:14:03 +0200
From: Eugen Konkov <kes-kes@yandex.ru>
Message-ID: <15010033764.20200117121403@yandex.ru>
To: Bruce Momjian <bruce@momjian.us>
CC: PostgreSQL-development <pgsql-hackers@postgreSQL.org>,
 pgsql-docs@lists.postgresql.org
Subject: Re: Does 'instead of delete' trigger support modification of OLD
In-Reply-To: <20191107222818.GA18712@momjian.us>
References: <919823407.20191029175436@yandex.ru>
 <20191106185935.GE1843@momjian.us>
  <1771593631.20191107112032@yandex.ru> <352093181.20191107112429@yandex.ru>
  <20191107212655.GA487@momjian.us> <20191107222818.GA18712@momjian.us>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hello Bruce,

> Triggers are designed to check and modify input data, and since DELETE
> has no input data, it makes no sense.

Sorry,  I  am  still ambiguous. You say that DELETE has no input data,
but doc says that it has:

https://www.postgresql.org/docs/current/trigger-definition.html
For  a row-level trigger, the input data also includes ... the OLD row for ... DELETE triggers


Also  restricting  DELETE  to  change  the  returned  data  by  DELETE
RETURNING seems as incomplete.

For example if triggers implement some compression.
 -- insert into field ZZZZZ value
 -- compress and actually store Zx5 into field
 -- Delete this insert row
 -- So user should get back that the value ZZZZZ was deleted and not Zx5.
 Correct?

 but currently user will see Zx5, because next code:

 OLD.value =  uncompress( OLD.value );

 does not effect RETURNING =(

-- 
Best regards,
Eugen Konkov