On 2014-09-18 17:30:38 +0800, Craig Ringer wrote: > On 07/29/2014 10:43 PM, Bruce Momjian wrote: > >> > * When a user tries to run "psql -f some_custom_format_backup", detect > >> > this and emit a useful error message. Ditto stdin. > > > > Uh, good idea, but can we really do that in psql? > > Where stdin is a file, or an explicit -f is given, then yes. > > Just look for PGDMP as the first five bytes and complain. > > To do it more generally we'd need to look at each statement and see if > it seems to begin with PGDMP. I'm not sure that's worth it; handling the > simple cases of > > psql -f mydb.dump > > and > > psql < mydb.dump > > would be quite sufficient. > > I'm not 100% sure if we can easily differentiate the second case from > "pg_dump | psql"; I know there's isatty(...) to test if stdin is a > terminal, but I'm not sure how easy/possible it is to tell if it's a > file not a pipe.
I don't think we need to make any discinction between psql -f mydb.dump, psql < mydb.dump, and whatever | psql. Just check, when noninteractively reading the first line in mainloop.c:MainLoop(), whether it starts with the magic header. That'd also trigger the warning on \i pg_restore_file, but that's hardly a problem. > pg_restore already knows to tell you to use psql if it sees an SQL file > as input. Having something similar for pg_dump would be really useful. Agreed. We could additionally write out a hint whenever a directory is fed to psql -f that psql can't be used to read directory type dumps. Greetings, Andres Freund -- Andres Freund http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training & Services -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
