Date: 2015-10-29 07:58:30 +0100
From: darkbluehacker
To: SQL devs <>
Version: 11.21.11 (Jul2015-SP1)
CC: chenkaijiang, darkbluehacker, @mlkersten, @njnes

Last updated: 2016-01-15 11:38:18 +0100

Comment 21418

Date: 2015-10-29 07:58:30 +0100
From: darkbluehacker

User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:41.0) Gecko/20100101 Firefox/41.0
Build Identifier:

using column of table in user defined fuction will cause execution failure and bogus error message reported.

Reproducible: Always

Steps to Reproduce:

1. unzip the attached fn_bug.tgz
2. create tables 'test_fn' and 'chg' using tables.sql
3. import data in test_fn.out to 'test_fn' table and import data in chg.out to 'chg' table;
4. create user defined function Get_r using monetdb.udf.sql
5. select count(*) from test_fn where get_r(c1, '01', '2015-08-31 23:59:59')>1.0;

Actual Results:

get error messages like: Function '..P*' not defined. (NOTE: the function name is bogus and keeps changing every time)

Expected Results:

get the count of the qualified rows.

Comment 21419

Date: 2015-10-29 08:00:57 +0100
From: darkbluehacker

Created attachment 364
sql and data to reproduce the bug

sql and data to reproduce the bug can be found in attached file.

Attached file: fn_bug.tgz (application/x-compressed, 630950 bytes)
Description: sql and data to reproduce the bug

Comment 21477

Date: 2015-11-08 11:01:29 +0100
From: @mlkersten

CREATE INDEX "chg_t_effc_tm" ON "sys"."ods_chg" ("t_effc_tm");

is not defined in the test scripts.

For the trimmed down version (4 tuples in each table) the problem
reported is:

Function 'calc.>=' not defined

This hints at an error in the UDF, which indeed attempts to execute:
X_141:bat[:oid,:bit] := mal.multiplex("calc","<=",X_132,X_140);

X_132 and X_140 are timestamps and the system only supports mtime.<=.

The issue is that this is not properly handled during the optimizer phase,
which leads to calling the default implementation, that throws an error.

Bug added as file BugTracker-2015/Tests/missed-error.Bug-3842.sql

Comment 21506

Date: 2015-11-11 09:22:21 +0100
From: @sjoerdmullender

(In reply to darkbluehacker from comment 0)

get error messages like: Function '..P*' not defined. (NOTE: the function
name is bogus and keeps changing every time)

This part of the bug has been fixed in changeset 982c835b3f2b

Comment 21601

Date: 2015-11-26 15:55:06 +0100
From: Kaijiang <>

Is there any progress on this bug?

Comment 21625

Date: 2015-12-02 07:36:29 +0100
From: Kaijiang <>

According to our debugging, it requires multiplex function "<=" for timestamps. But the function MANIFOLDremapMultiplex is actually a dummy function and thus it reports a function(In reply to Martin Kersten from comment 2)

CREATE INDEX "chg_t_effc_tm" ON "sys"."ods_chg" ("t_effc_tm");

is not defined in the test scripts.

For the trimmed down version (4 tuples in each table) the problem
reported is:

Function 'calc.>=' not defined

This hints at an error in the UDF, which indeed attempts to execute:
X_141:bat[:oid,:bit] := mal.multiplex("calc","<=",X_132,X_140);

X_132 and X_140 are timestamps and the system only supports mtime.<=.

The issue is that this is not properly handled during the optimizer phase,
which leads to calling the default implementation, that throws an error.

Bug added as file BugTracker-2015/Tests/missed-error.Bug-3842.sql

We found that MANIFOLDremapMultiplex is actually a dummy function and thus the 'calc.>=' is not found. But MANIFOLDevaluate (also in manifold.c) seems to have proper implementation of the 'calc.>='.

To fix this bug, can we just function mapping in manifold.mal as below?

pattern mal.multiplex(mod:str,fcn:str,a:any...):any...
address MANIFOLDremapMultiplex; --> address MANIFOLDevaluate

pattern batmal.multiplex(mod:str,fcn:str,a:any...):any...
address MANIFOLDremapMultiplex; --> address MANIFOLDevaluate

Comment 21686

Date: 2015-12-23 10:17:35 +0100
From: @njnes

for sql inline functions the optimization stack wasn't run, this is now fixed.

Comment 21689

Date: 2015-12-23 14:57:58 +0100
From: Kaijiang <>

(In reply to Niels Nes from comment 6)

for sql inline functions the optimization stack wasn't run, this is now
fixed.

Can I get the version with the fix and learn from it?

More, https://www.monetdb.org/bugzilla/show_bug.cgi?id=3882 looks similar to this bug. Could you also check it?