primary key display in psql

Started by Peter Eisentrautalmost 16 years ago4 messages
#1Peter Eisentraut
peter_e@gmx.net

When you look at a table definition with psql \d, one of the arguably
most important pieces of information -- the primary key -- is hidden
somewhere below under "indexes":

Table "public.test2"
Column | Type | Modifiers
--------+---------+-----------
a | integer | not null
b | integer | not null
Indexes:
"test2_pkey" PRIMARY KEY, btree (a, b)

I think we could easily improve that by having it look something like
this instead:

Table "public.test2"
Column | Type | Modifiers
--------+---------+-----------
a | integer | PK
b | integer | PK
Indexes:
"test2_pkey" PRIMARY KEY, btree (a, b)

Since there can only be one primary key, this should be unambiguous.

I don't have time to code this up right now, but maybe someone feels
inspired. What do you think?

#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: Peter Eisentraut (#1)
Re: primary key display in psql

Peter Eisentraut <peter_e@gmx.net> writes:

I think we could easily improve that by having it look something like
this instead:

Table "public.test2"
Column | Type | Modifiers
--------+---------+-----------
a | integer | PK
b | integer | PK
Indexes:
"test2_pkey" PRIMARY KEY, btree (a, b)

Spelling out "primary key" would seem to be more in keeping with existing
entries in that column, eg we have "not null" not "NN".

I think this is a sensible proposal for a single-column PK, but am less
sure that it makes sense for multi-col. The modifiers column is
intended to describe column constraints; which a multi-col PK is not,
by definition.

regards, tom lane

#3Robert Haas
robertmhaas@gmail.com
In reply to: Tom Lane (#2)
Re: primary key display in psql

On Wed, Jan 13, 2010 at 4:47 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Peter Eisentraut <peter_e@gmx.net> writes:

I think we could easily improve that by having it look something like
this instead:

    Table "public.test2"
 Column |  Type   | Modifiers
--------+---------+-----------
 a      | integer | PK
 b      | integer | PK
Indexes:
    "test2_pkey" PRIMARY KEY, btree (a, b)

Spelling out "primary key" would seem to be more in keeping with existing
entries in that column, eg we have "not null" not "NN".

I think this is a sensible proposal for a single-column PK, but am less
sure that it makes sense for multi-col.  The modifiers column is
intended to describe column constraints; which a multi-col PK is not,
by definition.

Yeah, IIRC, MySQL shows PRI for each column of a multi-column primary
key, and I think it's horribly confusing. I wouldn't even be in favor
of doing this just for the single-column case, on the grounds that it
makes the single and multiple column cases asymmetrical. IMO, the \d
output has too many bells and whistles already; the last thing we
should do is add more.

...Robert

#4Ross J. Reedstrom
reedstrm@rice.edu
In reply to: Robert Haas (#3)
Re: primary key display in psql

On Wed, Jan 13, 2010 at 05:03:33PM -0500, Robert Haas wrote:

On Wed, Jan 13, 2010 at 4:47 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Spelling out "primary key" would seem to be more in keeping with existing
entries in that column, eg we have "not null" not "NN".

I think this is a sensible proposal for a single-column PK, but am less
sure that it makes sense for multi-col. �The modifiers column is
intended to describe column constraints; which a multi-col PK is not,
by definition.

Yeah, IIRC, MySQL shows PRI for each column of a multi-column primary
key, and I think it's horribly confusing. I wouldn't even be in favor
of doing this just for the single-column case, on the grounds that it
makes the single and multiple column cases asymmetrical. IMO, the \d
output has too many bells and whistles already; the last thing we
should do is add more.

How about spelling it as so:

� � Table "public.test"
�Column | �Type � | Modifiers
--------+---------+-----------
�a � � �| integer | primary key
�b � � �| integer |
Indexes:
� � "test1_pkey" PRIMARY KEY, btree (a)

� � Table "public.test2"
�Column | �Type � | Modifiers
--------+---------+-----------
�a � � �| integer | primary key (compound)
�b � � �| integer | primary key (compound)
Indexes:
� � "test2_pkey" PRIMARY KEY, btree (a, b)

As to Tom's point that a compound primary key is a table level
restriction, by definition, participating in such a key is still a
restriction on what values that column can take. When introspecting
someone else's schema, with a very wide table, seeing '(compound)'
is a nice strong hint to go looking for the other members of the PK.

Ross
--
Ross Reedstrom, Ph.D. reedstrm@rice.edu
Systems Engineer & Admin, Research Scientist phone: 713-348-6166
The Connexions Project http://cnx.org fax: 713-348-3665
Rice University MS-375, Houston, TX 77005
GPG Key fingerprint = F023 82C8 9B0E 2CC6 0D8E F888 D3AE 810E 88F0 BEDE