I am running a calculation in a PL/pgSQL function and I want to use the result of that calculation in my C++ code. What's the best way to do that?
I can insert that result into a table and use it from there but I'm not sure how well that fares with best practices. Also, I can send message to
RAISE NOTICE but I don't know can I use that message in my code.
The details here are a bit thin on the ground, so it's hard to say for sure.
Strongly preferable whenever possible is to just get the function's return value directly.
SELECT my_function(args) if it returns a single result, or
SELECT * FROM my_function(args); if it returns a row or set of rows. Then process the result like any other query result. This is part of the basic use of simple SQL and PL/PgSQL functions.
Other options include:
Returning a refcursor. This can be useful in some circumstances where you want to return a dynamic result set or multiple result sets, though it's now mostly superseded by
RETURN QUERY and
RETURN QUERY EXECUTE. You then
FETCH from the refcursorto get the result rows.
LISTENing for an event and having the function
NOTIFY when the work is done, possibly with the result as a notify payload. This is useful when the function isn't necessarily called on the same connection as the program that wants to use its results.
Create a temporary table in the function, then
SELECT from the table from the session that called the function.
Emitting log messages via
RAISE and setting
client_min_messages so you receive them, then processing them. This is a very ugly way to do it and should really be avoided at all costs.
INSERTing the results into an existing non-temporary table, then
SELECTing them out once the transaction commits and the rows become visible to other transactions.
Which is better? It depends entirely on what you're trying to do. In almost all cases the correct thing to do is just call the function and process the return value, but there are exceptions in special cases.