Removing postgres -f command line option

On Thu, Nov 17, 2011 at 4:19 AM, Heikki Linnakangas
<heikki.linnakangas@enterprisedb.com> wrote:

While looking at Shigeru Hanada's foreign join pushdown patch, I noticed a
command line option that I didn't know to exist:

$ postgres --help
...
Developer options:
 -f s|i|n|m|h    forbid use of some plan types

That doesn't include all the options we support, the documentation lists:
s|i|o|b|t|n|m|h. These are aliases for enable_* planner options, e.g -fs is
equal to enable_seqscan=off.

That seems completely useless to me, because you can also do "-c
enable_seqscan=off". Any objections to removing the -f option altogether?

No. Seems useless to me.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

Excerpts from Robert Haas's message of jue nov 17 10:14:21 -0300 2011:

On Thu, Nov 17, 2011 at 4:19 AM, Heikki Linnakangas
<heikki.linnakangas@enterprisedb.com> wrote:

While looking at Shigeru Hanada's foreign join pushdown patch, I noticed a
command line option that I didn't know to exist:

$ postgres --help
...
Developer options:
 -f s|i|n|m|h    forbid use of some plan types

That doesn't include all the options we support, the documentation lists:
s|i|o|b|t|n|m|h. These are aliases for enable_* planner options, e.g -fs is
equal to enable_seqscan=off.

That seems completely useless to me, because you can also do "-c
enable_seqscan=off". Any objections to removing the -f option altogether?

No. Seems useless to me.

Dunno. It seems awfully convenient if you're testing optimizer stuff
that requires turning one of these off, though of course I haven't ever
done that. The alternative requires a lot more keystrokes. It doesn't
matter to me either way, though. (To be honest I wasn't aware of these
options -- not sure if I'll use them in the future)

$ PGOPTIONS=-fs psql
psql (9.2devel)
Digite «help» para obtener ayuda.

alvherre=# show enable_seqscan ;
enable_seqscan
----------------
off
(1 fila)

$ PGOPTIONS="-c enable_seqscan=0" psql
psql (9.2devel)
Digite «help» para obtener ayuda.

alvherre=# show enable_seqscan ;
enable_seqscan
----------------
off
(1 fila)

--
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Heikki,

On 17.11.2011 10:19, Heikki Linnakangas wrote:

$ postgres --help
...
Developer options:
-f s|i|n|m|h forbid use of some plan types

That doesn't include all the options we support, the documentation
lists: s|i|o|b|t|n|m|h. These are aliases for enable_* planner
options, e.g -fs is equal to enable_seqscan=off.

That seems completely useless to me, because you can also do "-c
enable_seqscan=off". Any objections to removing the -f option altogether?

I knew about it. But - I never needed it and I always scroll over it
when I show --help in my trainings.

When I was young - some when in last century - I learned that you never
should remove a feature without pre-announcing it as deprecated.

I think it is better to mark it deprecated in 9.2 and totally remove it
in 9.3.

Susanne

--
Susanne Ebrecht - 2ndQuadrant
PostgreSQL Development, 24x7 Support, Training and Services
www.2ndQuadrant.com

Heikki Linnakangas <heikki.linnakangas@enterprisedb.com> writes:

While looking at Shigeru Hanada's foreign join pushdown patch, I noticed
a command line option that I didn't know to exist:

$ postgres --help
...
Developer options:
-f s|i|n|m|h forbid use of some plan types

Hmm, I thought I'd fixed that help message to match reality recently.

That seems completely useless to me, because you can also do "-c
enable_seqscan=off". Any objections to removing the -f option altogether?

I use it. See also src/test/regress/regressplans.sh, which would become
greatly less wieldy if it had to spell out the switches long-form.

regards, tom lane

Tom Lane wrote:

Heikki Linnakangas <heikki.linnakangas@enterprisedb.com> writes:

While looking at Shigeru Hanada's foreign join pushdown patch, I noticed
a command line option that I didn't know to exist:

$ postgres --help
...
Developer options:
-f s|i|n|m|h forbid use of some plan types

Hmm, I thought I'd fixed that help message to match reality recently.

That seems completely useless to me, because you can also do "-c
enable_seqscan=off". Any objections to removing the -f option altogether?

I use it. See also src/test/regress/regressplans.sh, which would become
greatly less wieldy if it had to spell out the switches long-form.

I think it is hard to argue that such a user-visible flag is
reasonable, even if it helps backend developers avoid some keystrokes.

I think flags used only by backend developers should be things that can
_only_ be done with flags.

--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ It's impossible for everything to be true. +

Bruce Momjian <bruce@momjian.us> writes:

I think it is hard to argue that such a user-visible flag is
reasonable, even if it helps backend developers avoid some keystrokes.

I think flags used only by backend developers should be things that can
_only_ be done with flags.

Huh? By that argument, we should remove every single postmaster command
line switch except "-c var=value", because all the other ones are
shorthands for that --- Peter went to some lengths awhile ago to make
that true.

regards, tom lane

Tom Lane wrote:

Bruce Momjian <bruce@momjian.us> writes:

I think it is hard to argue that such a user-visible flag is
reasonable, even if it helps backend developers avoid some keystrokes.

I think flags used only by backend developers should be things that can
_only_ be done with flags.

Huh? By that argument, we should remove every single postmaster command
line switch except "-c var=value", because all the other ones are
shorthands for that --- Peter went to some lengths awhile ago to make
that true.

I was talking about developer-use-only flags.

--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ It's impossible for everything to be true. +