Re: Support of sql-92 select statement

Posted in 2003

Topics: Performance & Tuning, SQL Development & Query Writing

Jean Sagi — Thu, 20 Nov 2003 01:06:48 -0500 — source: Usenet: comp.databases.informix

Very interesting... But, Have this approach performance penalties? Chucho! Richard Harnden wrote: > "Fred Prose" <fprose@hotmail.com> wrote in message > news:195a4770.0311190822.49fa561b@posting.google.com... > >>We've been given the SQL source code for an application written to >>sql-92 standards. Some of the syntax looks like this: >> >>select >> c_count, >> count(*) as custdist >>from >> ( >> select > > > FROM TABLE(MULTISET(SELECT ...)) > > or something like that. > > You have to tell Informix that a) a select statement is a bag, > and b) you can make a table from it. > > > > > > > > > > > > -- Atte, Jes's Antonio Santos Giraldo jeansagi@myrealbox.com jeansagi@netscape.net sending to informix-list

↪ replying to Jean Sagi

Richard Harnden — Thu, 20 Nov 2003 15:26:54 -0000 — source: Usenet: comp.databases.informix

[top-posting fixed] "Jean Sagi" <jeansagi@myrealbox.com> wrote in message news:bpicnr$a0o$1@terabinaries.xmission.com... > > Richard Harnden wrote: > > "Fred Prose" <fprose@hotmail.com> wrote in message > > news:195a4770.0311190822.49fa561b@posting.google.com... > > > >>We've been given the SQL source code for an application written to > >>sql-92 standards. Some of the syntax looks like this: > >> > >>select > >> c_count, > >> count(*) as custdist > >>from > >> ( > >> select > > > > > > FROM TABLE(MULTISET(SELECT ...)) > > > > or something like that. > > > > You have to tell Informix that a) a select statement is a bag, > > and b) you can make a table from it. > > > > > Very interesting... > > But, > > Have this approach performance penalties? > > Chucho! > > Over using a tempory table ? I dunno, it's very possible that a temp. table gets created anyway, so probably very little. Personally, I'd stick with the explicit tempory table aproach: portable all the way back to SE, you get the chance to update the stats, create indices, etc.

↪ replying to Richard Harnden

Obnoxio The Clown — Thu, 20 Nov 2003 19:44:16 +0000 — source: Usenet: comp.databases.informix

Richard Harnden wrote: > Personally, I'd stick with the explicit tempory table aproach: portable > all the way back to SE, you get the chance to update the stats, create > indices, etc. Coward! -- "C'est pas parce qu'on n'a rien ' dire qu'il faut fermer sa gueule" - Coluche