← Back to team overview

maria-developers team mailing list archive

Re: Aggregate stored functions [MDEV-7773]


Hi, Varun!

On Mar 02, Varun Gupta wrote:
> Well i have checked out the syntax for the databases for aggregate
> functions. I thought using the syntax similar to that of HSQL would be
> good. In it if we are having N tuples then we are computing the values for
> the N rows and then after that we make another call and return the value
> for the function. Well this seems very reasonable to do. Using PostgreSQL
> <http://www.postgresql.org/docs/8.3/static/sql-createaggregate.html>  we
> need to have two regular functions but if we have an aggregate function
> like Count(column_name) , then this would be done in just a single function
> , I guess we would not need two functions for it .

Yes, I agree about PostgreSQL syntax. An aggregate function there
has "artefacts" - two regular functions that show up in the list of
functions and can be called directly too. I think this is confusing.

HSQL syntax, I mean, exactly as in HSQL, is full of random limitations.

In HSQL an aggregate function can have only *one* argument, and the
state is, always, exactly *two* variables. We, of course, won't have
these limitation, if we'll do HSQL-style syntax.

But this syntax is also kind of hackish. One function that has two
different semantics and the 'flag' argument that selects which one to
use. And many parameters in the function declaration, while only one is
in the function invocation.

Oracle uses an object-oriented syntax.

Did you find any other DBMS that support this feature?

Chief Architect MariaDB
and security@xxxxxxxxxxx

Follow ups