On Wednesday 22 August 2001 07:21 am, you wrote:
> Jeff Davis <list-pgsql-general@dynworks.com> writes:
> > I made some simple functions that create the operators + and - for any
> > combination of int4's and bool's.
> >
> > My question is two-part:
> > 1) Is this a sane approach?
>
> It'd be less work if you made a bool->int conversion function and relied
> on the regular integer operators. Safer, too: this way is likely to
> cause the system to accept queries that do things you didn't intend.
>
> In fact, I'd personally want the conversion to not be applied by
> default, which means you *shouldn't* call the conversion function int4().
> Functions named after the destination type are assumed to represent
> implicit type coercions, and will be applied without being mentioned
> explicitly. OTOH, if you don't care about type safety, you could
> achieve the same result as your operator set with just such an implicit
> coercion function:
>
> test71=# select 't'::bool + 44;
> ERROR: Unable to identify an operator '+' for types 'bool' and 'int4'
> You will have to retype this query using an explicit cast
>
> test71=# select 't'::bool + 't'::bool;
> ERROR: Unable to identify an operator '+' for types 'bool' and 'bool'
> You will have to retype this query using an explicit cast
>
> test71=# create function int4(bool) returns int4 as '
> test71'# select case when $1 then 1 else 0 end;' language 'sql';
> CREATE
>
> test71=# select 't'::bool + 44;
> ?column?
> ----------
> 45
> (1 row)
>
> test71=# select 't'::bool + 't'::bool;
> ?column?
> ----------
> 2
> (1 row)
>
> test71=#
>
> But as I said, I'd prefer to name the function something else (maybe
> "integerize") and have to invoke it explicitly. I'm an old Pascal
> programmer and believe strongly that bool and int ought not be
> considered interchangeable: that masks too many programming errors.
>
> > 2) Might enough other people find a use that I should make a
> > contribution somewhere (and would it be appropriate to send it to
> > pgsql-hackers or pgsql-patches)?
>
> As you can see, there's not a lot to it, at least not for low-volume
> applications. If you expected to invoke integerize() zillions of
> times, it might be worth the trouble to prepare a C-coded version of
> it. That would be worth contributing, since other people have asked
> for this same functionality. But then you'd have to get people to
> agree on a name for the function, and believe me that'll be the
> hardest part ;-)
>
> regards, tom lane
>
> ---------------------------(end of broadcast)---------------------------
> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org
I have come to the conclusion that such a C function is about the same speed
as an inlined SQL "CASE" clause, but both are much faster than a SQL
function. Basically, I ran a test in which I created 3 views:
create view test_view_sql_fcn as select
*,(as_int_sql(col_a='foo')+as_int_sql(col_b='bar')) as relevance FROM
test_table;
create view test_view_c as select *,(as_int(col_a='foo')+as_int(col_b='bar'))
as relevance FROM test_table;
create view test_view_sql as select *,(case when col_a = 'foo' then 1 else 0
end + case when col_b = 'bar' then 1
else 0 end) as relevance FROM test_table;
test_table(col_a varchar(50),col_b varchar(50)) is a table with 14044 tuples,
mostly repetitive (test data).
as_int is a C function:
int4 bool_as_int(bool b) { return (int4)(b); }
as_int_sql is a SQL function:
create function int4(bool) returns int4 as 'select case when $1 then 1 else 0
end;' language 'sql';
I ran my client (written in C) which iterated 100 times through 3 loops, each
time executing a query on one of those views:
SELECT * FROM <aforementioned view> WHERE relevance > 0;
so a total of 300 queries, 100 for each view, each loop of 100 queries in
it's own transaction.
results:
SQL function: 59 seconds
C function: 17 seconds
SQL inline: 17 seconds
Note that I was not aiming for super-accuracy; just for a general estimate.
So, I concluded:
1) using a function might be cleaner to write in some situations, although
views will ususally eliminate any strong need.
2) SQL function is out of the question if performance is a concern
3) I see little reason to not add the function, since it should be simple
enough not to cause problems/confusion.
4) I agree with Tom about implicit casting, I actually didn't know PostgreSQL
would do that (I assumed a "::int4" suffix would be required when using "+").
I named the function "as_int" for lack of a better term (I am not sure
"integerize" would flow as well from the keyboard). However, I really don't
care what the name is.
I hope this infomation helps someone.
Regards,
Jeff Davis
