Hi Sergei, I understand that having a cursor like syntax we would have to write some new code . What I would have liked to know is that the more details should be discussed which need to be incorporated in the implementation of the aggregate functions. Basically what I am asking is what all details have to be there in the implementation of the aggregate functions. On Fri, Mar 4, 2016 at 10:21 PM, Sergei Golubchik <serg@mariadb.org> wrote:
Hi, Varun!
On Mar 04, Varun Gupta wrote:
With the syntax more or less clear to us, can we have a discussion about how we have to go on with the implementation for the aggregate functions. Firstly I thought I should look at the implementation of CREATE FUNCTION, any other suggestions ?
Sure, why not. But, as you've seen, that suggested syntax is almost identical to the standard one, so hopefully it'll need very little changes to the current CREATE FUNCTION implementation.
More important would be to look at the function execution (not definition). This mainly happens in sql/sp_head.cc but other sql/sp_* files are also used.
This cursor-like syntax introduces an new mode of executing stored routines. Currently, a routine (a procedure or a function) is executed completely from the beginning to the end. But with this cursor-like syntax, the server will need to start executing a stored function, stop at the FETCH statement, postpone the execution, then later continue from that point. There is no code to do it now.
Regards, Sergei Chief Architect MariaDB and security@mariadb.org