concurrency in stored procedures

Hi,

using constraints on tables I was able to remove some race conditions,
because the unique index prevents the same data to be inserted twice
into the table.

But I still didn't fix all the race conditions, because in some
functions I have to modify more than one table or I just have read and
write data in the same table. So, what is the best way to handle
concurrency in stored procedures?

I read that using locks isn't good because it may lead to deadlocks, so
I was thinking about transactions, but I wan't able to find a good example.

What would you to in order to be sure that one function or a part of it
is atomically executed?

I also read that postgresql is able to detect deadlocks and can try to
solve them. How does this happen in a stored procedure and how can a
procedure know that it was aborted because of the deadlock?

Thank you

--
Non c'e' piu' forza nella normalita', c'e' solo monotonia.