Bug or intentionally under-documented "\c databasename;" behavior?
Hey,
I expected the following to tell me: database "testdb;" does not exist
Instead the connection attempt was successful.
postgres=# create database testdb;
CREATE DATABASE
postgres=# \c testdb;
You are now connected to database "testdb" as user "vagrant".
Specifically, the trailing semi-colon on the testdb is being treated,
apparently, as a second parameter to \c (or just plain ignored which seems
wrong too); which itself is a surprise given the absence of whitespace, and
\c documents that - passed as a parameter is an acceptable way to omit a
parameter so the semi-colon should have been considered as a username.
David J.
"David G. Johnston" <david.g.johnston@gmail.com> writes:
I expected the following to tell me: database "testdb;" does not exist
Instead the connection attempt was successful.
postgres=# create database testdb;
CREATE DATABASE
postgres=# \c testdb;
You are now connected to database "testdb" as user "vagrant".
This is because psql_scan_slash_option was told to strip any trailing
semicolon, and did so. If there's any rhyme or reason to which
psql metacommands pass semicolon=true and which do not, I can't
detect it :-(. And you're right that there's nothing about it in
the documentation.
I wonder if we can get away with removing that quirk. Or else try
to establish some policy about it, and document the policy.
A really brave person might propose nuking the OT_SQLIDHACK parsing
mode as well. That was never supposed to be a long-term fixture.
regards, tom lane