Re: [PATCHES] Function's LEAST, GREATEST and DECODE - Mailing list pgsql-hackers

From Robert Treat
Subject Re: [PATCHES] Function's LEAST, GREATEST and DECODE
Date
Msg-id 1119624967.22831.1438.camel@camel
Whole thread Raw
In response to Re: [PATCHES] Function's LEAST, GREATEST and DECODE (Oracle vararg polymorphic functions)  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
On Fri, 2005-06-24 at 09:21, Tom Lane wrote:
> [ moving to -hackers for a wider audience ]
> 
> Today's issue: should the GREATEST/LEAST functions be strict (return
> null if any input is null) or not (return null only if all inputs are
> null, else return the largest/smallest of the non-null inputs)?
> 
> Pavel Stehule <stehule@kix.fsv.cvut.cz> writes:
> > On Thu, 23 Jun 2005, Tom Lane wrote:
> >> Pavel Stehule <stehule@kix.fsv.cvut.cz> writes:
> >> +             /* If any argument is null, then result is null (for GREATEST and LEAST)*/
> >> 
> >> Are you sure about that?  The only reference I could find says that
> >> these functions are not strict in Oracle:
> >> 
> >> http://download-east.oracle.com/otn_hosted_doc/rdb/pdf/sql_ref_v71_vol1.pdf
> >> on page 2-185:
> >> 
> >>> The NULL keyword can appear in the list but is ignored. However, not all 
> >>> value expressions can be specified as NULL. That is, a non-NULL value 
> >>> expression must be in the list so that the data type for the expression
> >>> can be determined. 
> >>> The GREATEST and LEAST functions can result in NULL only if at run time 
> >>> all value expressions result in NULL. 
> >> 
> >> The strict interpretation is mathematically cleaner, no doubt, but
> >> offhand it seems less useful.
> >> 
> 
> > I know it, But when moustly PostgreSQL function is strict I desided so 
> > greatest and least will be strict. There is two analogy:
> 
> > one, normal comparing which implicate strinct
> > aggregate function which ignore NULL.
> 
> > Tom I don't know, what is better. Maybe Oracle,
> 
> > because
> 
> > least(nullif(col2, +max), nullif(col2, +max)) isn't really readable, but 
> > it's "precedens" for PostgreSQL. I selected more conservative solution, 
> > but my patches are only start points for discussion (really) :).
> 
> > Please, if You think, so Oracle way is good, correct it.
> 
> I'm still favoring non-strict but it deserves more than two votes.
> Anybody else have an opinion?
> 

If the sql spec has nothing to say on it, then we should probably
support Oracles take, since this seems like an Oracleism anyway. 


Robert Treat
-- 
Build A Brighter Lamp :: Linux Apache {middleware} PostgreSQL



pgsql-hackers by date:

Previous
From: "John Hansen"
Date:
Subject: Re: [PATCHES] Function's LEAST, GREATEST and DECODE (Oracle vararg polymorphic functions)
Next
From: "Jim C. Nasby"
Date:
Subject: Re: [PATCHES] O_DIRECT for WAL writes