Re: how does postgresql handle LOB/CLOB/BLOB column data that dies before the query ends

Поиск
Список
Период
Сортировка
От Noel Grandin
Тема Re: how does postgresql handle LOB/CLOB/BLOB column data that dies before the query ends
Дата
Msg-id CAFYHVnUQvsXYo48WEPU6oB9=Rguhv2VP7_TqPmu1WuCzvuTQ1g@mail.gmail.com
обсуждение исходный текст
Ответ на Re: how does postgresql handle LOB/CLOB/BLOB column data that dies before the query ends  (Tom Lane <tgl@sss.pgh.pa.us>)
Ответы Re: how does postgresql handle LOB/CLOB/BLOB column data that dies before the query ends
Список pgsql-hackers
Дерево обсуждения


On Sat, 25 Feb 2023 at 08:33, Tom Lane <tgl@sss.pgh.pa.us> wrote:
Yeah, Postgres has an analogous kind of problem.  Our standard way to
use "large objects" is to store their identifying OIDs in tables,
... 
and in particular they can *not* close the transaction that read the
OID if they'd like to read a matching state of the large object.
So far there's not been a lot of complaints about that ...


OK, so it seems like so far my design is not far off the PostgreSQL design (which is very comforting). 

I wonder if the difference is in the client<->server protocol.

Does PostgreSQL hold the transaction open until the client side has closed the resultset (or the query object possibly, not sure about the PostgreSQL API here).
H2 has a very simple client-server protocol, which means the client simply sends a query and gets back a result-set stream, and there is no explicit acknowledgement of when the client closes the resultset, which means that the MVCC transaction is typically closed by the time the client even starts reading the resultset.

В списке pgsql-hackers по дате отправления:

Предыдущее
От: Julien Rouhaud
Дата:
Сообщение: Re: Doc update for pg_stat_statements normalization
Следующее
От: Amit Kapila
Дата:
Сообщение: Re: [PATCH] Use indexes on the subscriber when REPLICA IDENTITY is full on the publisher