It's impossible to handle a constant parameter in an aggregate function

[msiomkin]

It's impossible to handle a constant parameter in an aggregate function. I hope it's a bug, not the by design behaviour. (E.g. say, I want to evaluate a truncated mean and need to pass the upper and lower bounds as constants).

To Reproduce

Create an aggregate using the attached source file.

Execute the following statements:

create table test(id int, value int);
insert into test(id, value) values(1, 1);
select myagg(id, 1) from test group by value;

Expected behavior

+-------------------------------------------------------------------------------------------+
| %1                                                                                        |
+===========================================================================================+
| 00010203                                                                                  |
+-------------------------------------------------------------------------------------------+

Actual behavior

TypeException:user.main[15]:'aggr.submyagg' undefined in:     X_0:bat[:blob] := aggr.submyagg(X_1:bat[:int], X_2:bat[:int], X_3:bat[:oid], X_4:bat[:oid], true:bit);

So I have to do some workaround: accept the second parameter as bat*, create an iterator for it and do BUNtail(it, 0). I think it's nor graceful neither very efficient.

And it does work for the ungrouped case, but I can't get the proper parameter value. It looks like there is some random number there.

select myagg(id, 1) from test;

Expected behavior (in merovingian.log)

#DFLOWworker1019: myagg: ERROR: COL2: 1

Actual behavior (in merovingian.log)

#DFLOWworker1019: myagg: ERROR: COL2: 1298

Software versions

• MonetDB v11.47.3 (hg id: 79dbf838f044 88683)
• Almalinux 8
• self-installed and compiled

[njnes]

For now only for qantile* we allow a second (const) argument. See the code in is_const_func (rel_bin.c). Should be generalized.