Обсуждение: HOST variable in psql

Hi,

the use of :HOST in psql containing the socket directory for local connections is pretty useless for us, so we set it
tothe hostname in a global psqlrc file: 

\set HOST `hostname -s`

This works great when invoking psql on the command line. But if I perform a \c to connect to a different database in
thedatabase cluster, then :HOST gets reset to the socket directory. On the one hand, this makes sense, on the other,
theglobal psqlrc file does not seem to be executed again, thus leaving :HOST in an unuable state for us: 

postgres=# \echo :HOST
mypghost
postgres=# create database test;
CREATE DATABASE
postgres=# \c test
You are now connected to database "test" as user "postgres".
test=# \echo :HOST
/tmp

The documentations (https://www.postgresql.org/docs/12/app-psql.html) says:

"HOST
The database server host you are currently connected to. This is set every time you connect to a database (including
programstart-up), but can be changed or unset." 

"Files
psqlrc and ~/.psqlrc
Unless it is passed an -X option, psql attempts to read and execute commands from the system-wide startup file (psqlrc)
andthen the user's personal startup file (~/.psqlrc), after connecting to the database but before accepting normal
commands."

Makes sense so far, except the last part "after connecting to the database but before accepting normal commands." This
appliesonly to the first psql startup but not a subsequent \c. If I do a \c, then that is connecting to a database and
thatmeans, the psqlrc *should* be executed again, which it isn't. 

Bug or feature? If feature, is there any remedy to this behavior?

Thanks in advance,
Paul

=?utf-8?Q?Paul_F=C3=B6rster?= <paul.foerster@gmail.com> writes:
> the use of :HOST in psql containing the socket directory for local connections is pretty useless for us, so we set it
tothe hostname in a global psqlrc file: 
> \set HOST `hostname -s`
> This works great when invoking psql on the command line. But if I perform a \c to connect to a different database in
thedatabase cluster, then :HOST gets reset to the socket directory. On the one hand, this makes sense, on the other,
theglobal psqlrc file does not seem to be executed again, thus leaving :HOST in an unuable state for us: 

AFAICS, psql is behaving as documented.  Why are you trying to override
HOST like that, instead of just using some other variable?

            regards, tom lane

Hi Tom,

> On 27. Apr, 2020, at 14:20, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> AFAICS, psql is behaving as documented.  Why are you trying to override
> HOST like that, instead of just using some other variable?

I could but

a) HOST is HOST and should not point to a socket directory. A socket directory is not a HOST. This makes no sense,
especiallysince I see no use for using a socket directory in any psql script. There are tons of use cases for the
hostname,though. 

b) it would require changing some scripts.

c) the docs are IMHO not 100% clear on this. I think the psqlrc should be executed on EVERY new connect, hence
including\c, and this is how I interpret the docs too. But it doesn't... 

also (slightly related) d) PostgreSQL unfortunately does not provide any functionality to determine the server hostname
itruns on. There is an extension but I still don't see why PostgreSQL does not include this natively. 

Cheers,
Paul

=?utf-8?Q?Paul_F=C3=B6rster?= <paul.foerster@gmail.com> writes:
>> On 27. Apr, 2020, at 14:20, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> AFAICS, psql is behaving as documented.  Why are you trying to override
>> HOST like that, instead of just using some other variable?

> I could but

> a) HOST is HOST and should not point to a socket directory. A socket directory is not a HOST. This makes no sense,
especiallysince I see no use for using a socket directory in any psql script. There are tons of use cases for the
hostname,though. 

[ shrug... ]  When connecting via a Unix socket, psql does not have a
relevant "hostname" to report, so it reports what it's got.  You are
ascribing a meaning to HOST that it does not have, never has had, and
never can have in this case.

> c) the docs are IMHO not 100% clear on this. I think the psqlrc should be executed on EVERY new connect, hence
including\c, and this is how I interpret the docs too. But it doesn't... 

There has been some talk of providing a way to execute stuff on reconnect,
if memory serves, but nothing's been done about it.

            regards, tom lane

Hi Tom,

> On 27. Apr, 2020, at 15:13, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> [ shrug... ]  When connecting via a Unix socket, psql does not have a
> relevant "hostname" to report, so it reports what it's got.  You are
> ascribing a meaning to HOST that it does not have, never has had, and
> never can have in this case.

I'm not much of a C programmer, but there's AFAIK a C function to get the hostname. So it should IMHO be easy to put
somethinginto the psql init (and \c) code that calls the function and set the HOST variable accordingly. IMHO... 

> There has been some talk of providing a way to execute stuff on reconnect,
> if memory serves, but nothing's been done about it.

thought so. So I'll have to work around that.

Thanks,
Paul