On Wed, May 12, 2004 at 05:37:49PM -0700, Keith Herold wrote:
> In MS SQL 2000, through the query analyzer
>
> SELECT '500' = 500
>
> returns 500 .
>
> SELECT 500 = '500'
>
> returns
> Server: Msg 170, Level 15, State 1, Line 1
> Line 1: Incorrect syntax near '='.
Well that's sure inordinately stupid behavior. The test "A = B"
should give the same result as "B = A", at least!
In Oracle 8.1.7.4, string '5' does equal integer 5, but you seem to
have to ask it via a case statement:
SQL> select 5 = 5 from dual;
ERROR at line 1:
ORA-00923: FROM keyword not found where expected
SQL> select case when 5 = '5' then 1 else 0 end as bool from dual;
BOOL
----------
1
SQL> select case when '5' = 5 then 1 else 0 end as bool from dual;
BOOL
----------
1
> > But if that is the case, then clearly, '500' != 500. So
> > unless somebody can come up with a better idea, SQLite
> > version 3.0 will return "0" for the following:
> >
> > SELECT '500'=500;
> >
> > On the other hand, the following two statements will return
> > "1" (or true):
> >
> > SELECT '500'+0=500;
> > SELECT '500'=(500||'');
That sounds ok to me. Wouldn't it be clearer to have explicit cast
statements though, rather than doing weird no-ops like "||''" solely
to force a type conversion?
Also, since you're introducing manifest typing, it would probably be
very handy to have good boolean tests both for "is X currently of type
Y?" and "CAN X be of type Y?". E.g., "CAN this thing be an integer?",
where by "can" I mean, "Is it currently an integer, or if it is
currently stored as a some other type, can it be LOSSLESSLY converted
to an integer?"
Some languages (e.g., S-Plus), make the latter oddly difficult. (Not
THAT difficult; in S you can write that "can be integer" test in 5
lines, but those 5 lines are also somewhat easy to get wrong.) I
think that in any system where objects can automatically change type,
both the those sorts of "can be" boolean tests should also be built
in.
--
Andrew Piskorski <[EMAIL PROTECTED]>
http://www.piskorski.com/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
