Обсуждение: BUG #17824: PQTRANS_ACTIVE misleading

Поиск
Список
Период
Сортировка

BUG #17824: PQTRANS_ACTIVE misleading

От
PG Bug reporting form
Дата:
The following bug has been logged on the website:

Bug reference:      17824
Logged by:          Andy Schoenberger
Email address:      akschoenberger@gmail.com
PostgreSQL version: 15.2
Operating system:   All
Description:

The documentation for PQTRANS_ACTIVE [1] does not make explicitly clear that
it is returned in *all* cases where there is a command in progress on an
okay connection, and not just in the case that there is a command in
progress inside of a transaction [2]. With a name like PQTRANS_ACTIVE, I
believe it is worth making clear that it can happen when a transaction is
not active. I believe this has led to multiple bugs in other programs that
attempt to display the current status of a transaction, including a possible
one in psql [3].

Assuming I am understanding everything correctly, I think the documentation
should probably be updated to make clear that PQTRANS_ACTIVE does not mean
that a transaction is active.

Best,
Andy


[1]
From:
https://www.postgresql.org/docs/15/libpq-status.html#LIBPQ-PQTRANSACTIONSTATUS

> Returns the current in-transaction status of the server.

> The status can be PQTRANS_IDLE (currently idle), PQTRANS_ACTIVE (a command
is in progress), PQTRANS_INTRANS (idle, in a valid transaction block), or
PQTRANS_INERROR (idle, in a failed transaction block). PQTRANS_UNKNOWN is
reported if the connection is bad. PQTRANS_ACTIVE is reported only when a
query has been sent to the server and not yet completed.

[2]

https://github.com/postgres/postgres/blob/d69cd3a2e29c68a3fdfd4e97167f9a7c6ad29e80/src/interfaces/libpq/fe-connect.c#L6729

[3]
The psql documentation below shows that a `*` character is present when a
transaction is active:

From:
https://www.postgresql.org/docs/current/app-psql.html#APP-PSQL-PROMPTING

> Transaction status: an empty string when not in a transaction block, or *
when in a transaction block, or ! when in a failed transaction block, or ?
when the transaction state is indeterminate (for example, because there is
no connection).

…but a `*` character is also technically used if the transaction status is
PQTRANS_ACTIVE, which does not necessarily happen during a transaction.
https://github.com/postgres/postgres/blob/46d490ac19a7ca93a5c0f47e5a0e759b5385a8ae/src/bin/psql/prompt.c#L236

I’m not sure if this is actually considered a bug or not since I don’t think
psql displays a new prompt until after the command completes, but the logic
in the code being misleading does not help.


Re: BUG #17824: PQTRANS_ACTIVE misleading

От
"David G. Johnston"
Дата:
On Wed, Mar 8, 2023 at 8:28 AM PG Bug reporting form <noreply@postgresql.org> wrote:
The following bug has been logged on the website:

Bug reference:      17824
Logged by:          Andy Schoenberger
Email address:      akschoenberger@gmail.com
PostgreSQL version: 15.2
Operating system:   All
Description:       

The documentation for PQTRANS_ACTIVE [1] does not make explicitly clear that
it is returned in *all* cases where there is a command in progress on an
okay connection, and not just in the case that there is a command in
progress inside of a transaction [2]. With a name like PQTRANS_ACTIVE, I
believe it is worth making clear that it can happen when a transaction is
not active.

It is not possible for the server to execute a command without there being an active transaction in progress.


…but a `*` character is also technically used if the transaction status is
PQTRANS_ACTIVE, which does not necessarily happen during a transaction.
https://github.com/postgres/postgres/blob/46d490ac19a7ca93a5c0f47e5a0e759b5385a8ae/src/bin/psql/prompt.c#L236

I’m not sure if this is actually considered a bug or not since I don’t think
psql displays a new prompt until after the command completes, but the logic
in the code being misleading does not help.

Right, it is difficult to report on/see the implicit transaction that occurs if you send a command to the server without constructing an explicit transaction block, since it goes away as the command execution ends.

David J.

Re: BUG #17824: PQTRANS_ACTIVE misleading

От
Tom Lane
Дата:
PG Bug reporting form <noreply@postgresql.org> writes:
> The documentation for PQTRANS_ACTIVE [1] does not make explicitly clear that
> it is returned in *all* cases where there is a command in progress on an
> okay connection, and not just in the case that there is a command in
> progress inside of a transaction [2]. With a name like PQTRANS_ACTIVE, I
> believe it is worth making clear that it can happen when a transaction is
> not active.

The existing documentation defines it as "a command is in progress",
which seems perfectly clear to me --- adding more words would just
muddy the waters.

Also, if a command is running then the server *does* have a transaction
open -- it might be an auto-started, auto-committed one, but there's
always a transaction.  You seem to be using that word in a way that does
not correspond with common Postgres usage.  (Note that the definitions for
PQTRANS_INTRANS and PQTRANS_INERROR say "transaction block" not
"transaction", to comport with our usual terminology.)

            regards, tom lane