Обсуждение: Detailed index predicate with \d on indexes in psql

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

(This is using 7.2b1)

Anyone know why I am getting the "internal form" of the 
partial-index predicate? In other words, instead of 
getting something like this (thanks Tom):

regression=# \d apple  Index "apple"Column  |  Type
- ---------+---------topping | integer
hash
Index predicate: (topping > 99)


I get something like this:
  Index "apple"Column  |  Type
- ---------+---------topping | integer
hash for table "pizza" WHERE (topping > 2000)
Index predicate: ({ EXPR :typeOid 16  :opType op :oper { OPER :opno 521 :opid 14
7 :opresulttype 16 } :args ({ VAR :varno 1 :varattno 3 :vartype 23 :vartypmod -1 :varlevelsup 0 :varnoold 1 :varoattno
3}{ CONST :consttype 23 :constlen 4 :co
 
nstbyval true :constisnull false :constvalue  4 [ 99 0 0 0 ] })})

Doing a:

SELECT c.relname, i.indpred
FROM pg_index i, pg_class c
WHERE c.oid = i.indexrelid
and i.indpred like '(%'

reveals that this is happening to every partial index I 
create.

I'm also wondering if we even need the "Index predicate:" 
section at all? When it works properly, will it ever give 
more information than what the tail end of pg_get_indexdef 
returns? If it gives the same, is one preferred over the other?

Thanks,
Greg Sabino Mullane
greg@turnstep.com
PGP Key: 0x14964AC8 200111131146

-----BEGIN PGP SIGNATURE-----
Comment: http://www.turnstep.com/pgp.html

iQA/AwUBO/FOorybkGcUlkrIEQL9KACgyJu7YFWCjJQPwEL32yjhmegocRYAn1iC
4djb4ZoOkrSDePXJ6rsQcSCW
=M66f
-----END PGP SIGNATURE-----

"Greg Sabino Mullane" <greg@turnstep.com> writes:
> Anyone know why I am getting the "internal form" of the 
> partial-index predicate?

If you look in describe.c, you'll see that what psql is printing is
the result ofSELECT pg_get_expr(i.indpred, i.indrelid) as indpredFROM pg_index i
which should yield the exact same text as what pg_get_indexdef offers
in WHERE.  Have you mucked with this SELECT?  Are you perhaps trying
to run against a pre-7.2 server (pg_get_expr is new in 7.2)?

> I'm also wondering if we even need the "Index predicate:" 
> section at all?

Not if you intend to print the results of pg_get_indexdef instead.
        regards, tom lane