in reply to Re^3: T-SQL problem - Perl solution?
in thread T-SQL problem - Perl solution?

In this case SQL 2000 produces the exact same execution plan for "SELECT 1" and "SELECT *" -

D:\>copy con: select1.sql
set showplan_text on
go
if exists (select 1 from sysobjects where xtype = 'U' and name = 'jobs
+') print 'exists' else print 'doesn''t exist'
^Z
        1 file(s) copied.

D:\>copy con: selectstar.sql
set showplan_text on
go
if exists (select * from sysobjects where xtype = 'U' and name = 'jobs
+') print 'exists' else print 'doesn''t exist'
^Z
        1 file(s) copied.

D:\>osql -dpubs -n -i selectstar.sql > selectstar.txt

D:\>osql -dpubs -n -i select1.sql > select1.txt

D:\>diff select1.txt selectstar.txt
3c3
<  if exists (select 1 from sysobjects where xtype = 'U' and name = 'j
+obs')
---
>  if exists (select * from sysobjects where xtype = 'U' and name = 'j
+obs')

[download]

And while Win's SQL does compile without error, it may be a disaster in production. For example, this code (including the code you sanction) starts an infinite loop -


use pubs
go
create table Experimental_parameters (c char (1) primary key)
insert into Experimental_parameters (c) values ('a')
go
create table another_table (c char (1) references Experimental_paramet
+ers (c))
insert into another_table (c) values ('a')
go
WHILE (exists(select 1 from INFORMATION_SCHEMA.tables where table_name
+ = 'Experimental_parameters'))
begin
    drop table Experimental_parameters -- fails
end
go
print 'finished' -- never gets here

[download]

 

Replies are listed 'Best First'.
Re^5: T-SQL problem - Perl solution?
by monktim (Friar) on Oct 25, 2004 at 15:36 UTC
    Good point on the execution plans EdwardG. This example may be trivial enough that SELECT * is not better than SELECT 1. I should have stated more clearly that it is good practice to use SELECT * in an EXISTS because it can give performance improvements. Your mileage may vary.

    There is no inherent problem with using the WHILE clause just as I posted it. I certainly didn't sanction the code you posted but it is a good example and well worth noting. You never want to get stuck in an infinite loop. Error checking should also always be done. 

        DROP TABLE Experimental_parameters
    IF @@ERROR <> 0 BEGIN
         --DO SOMETHING HERE (i.e. RAISERROR & RETURN)
    END

    [download]
    Update: I just read the OP. I was replying to Re^2: T-SQL problem - Perl solution? which didn't indicate a temp table or code inside the WHILE. My sanctioned code could yield the behavior EdwardG posted based on the OP. The OP had 
    if (SELECT COUNT(*) FROM #Experimental_parameters) <> 0 begin DROP TAB
+LE Experimental_parameters end

    [download]
    Error hanling code needs to be used inside the WHILE.

    Good call EdwardG. Thanks
[reply]
[d/l]
[select]

In Section Seekers of Perl Wisdom