Thread: How to determine PRIMARY KEYS
Hi all,
I'm the author of sdsq/tksql an interface to postgreSQL I released under GPL some weeks ago.
Tksql tries as much as possible to use PRIMARY KEYS instead of oids in any operation so that logs are much more
readableand so that I can use those logs to do a replica of the database to keep two databases in sync (in a very
simplesituation thought...)
Now, I'd like suggestion and a comment on the algorithm I'm using to find the primary key.
I really think that this is more than needed, becouse if you have a table w/ a primary key and unique index NOT NULL
onan other attribute both these attributes would be choosen not just the minimum needed (the primary key OR an
attributeNOT NULL w/ unique index).
Here is what I do:
1. find all unique indexes on table ($tbl)
SELECT i.indexrelid FROM pg_class c, pg_index i \ WHERE c.relname = '$tbl' and c.oid = i.indrelid \
andi.indisunique = 't'
2. select in pg_attribute the attributes of these indexes... 3. loop over these attrs ($at) and select those that
areNOT NULL.
SELECT attname,attnotnull from pg_attribute \ where attrelid = (SELECT oid from pg_class \ where
relname= '$tbl') and attname = '$at' \ and attnotnull = 't'
May I as an alternative just relay on the pattern *_pkey and look at the
attribute of relation *_pkey to get the PRIMARY KEY?
Is there a document that explains system tables?
Thanks in advance
sandro
*:-)
PS. For those who whant to look at tksql or sdsql (the tcl/tk package): http://pcco2.mi.infn.it/sd/soft. If you
encounterproblems installing or running it I'd really appreciate to be informed (and/or if you like it...;-) .
Thanks
Note: sorry if this message shows up 2 times. I had problems w/ config of sendmail
--
Sandro Dentella *:-)
e-mail: sandro@ermit.it sandro.dentella@mi.infn.it
Sandro Dentella <sandro.dentella@tin.it> writes:
> Here is what I do:
Wouldn't it be easier just to look for the pg_index row (if any) that
has indisprimary = true?
regards, tom lane
On Mon, 11 Jun 2001, Sandro Dentella wrote: > I really think that this is more than needed, becouse if you have a table > w/ a primary key and unique index NOT NULL on an other attribute both these > attributes would be choosen not just the minimum needed (the primary key > OR an attribute NOT NULL w/ unique index). There's an attribute indisprimary in pg_index which you can use to find what's the primary key on the table. I think what you should do is this: a) if there's a primary key, use that. b) if there isn't, find a unique/notnull index. > 1. find all unique indexes on table ($tbl) > > SELECT i.indexrelid FROM pg_class c, pg_index i \ > WHERE c.relname = '$tbl' and c.oid = i.indrelid \ > and i.indisunique = 't' > > > 2. select in pg_attribute the attributes of these indexes... > > 3. loop over these attrs ($at) and select those that are NOT NULL. > > SELECT attname,attnotnull from pg_attribute \ > where attrelid = (SELECT oid from pg_class \ > where relname = '$tbl') and attname = '$at' \ > and attnotnull = 't' > > May I as an alternative just relay on the pattern *_pkey and look at the > attribute of relation *_pkey to get the PRIMARY KEY? > > Is there a document that explains system tables? > > > Thanks in advance > > sandro > *:-) > > > > PS. For those who whant to look at tksql or sdsql (the tcl/tk package): > http://pcco2.mi.infn.it/sd/soft. If you encounter problems installing or > running it I'd really appreciate to be informed (and/or if you like > it...;-) . Thanks > > Note: sorry if this message shows up 2 times. I had problems w/ config of > sendmail > > > > -- -- Alex Pilosov | http://www.acecape.com/dsl CTO - Acecape, Inc. | AceDSL:The best ADSL in Bell Atlantic area 325 W 38 St. Suite 1005 | (Stealth Marketing Works! :) New York, NY 10018 |