Bug in synthetic key generation?

[#1]

Dear all,

during development of an rather unusual app I may have come across a bug in the synthetic key generation. I am not sure, therefore I want to ask if anybody can make sense out of the following behavior.

All below is the minimal example I could come up with exhibiting the strange behavior. The data model looks like:

I load in tables A, B, C a row of ones each and in table D a row of twos.

Qlikview somehow seems to loose the connection between the key and data part of table D during synthetic key generation:

If I check the synthetic tables created it show two lines of null values:

If I load Key_A = 2, Key_C = NULL, KeyD = 2 into table A, everything works as expected:

Am I missing something obvious regarding the synthetic key generation?

Thank you and with best regards,

Oliver Wienand

The load script, in case someone wants to recreate the issue in a personal edition:

SET NullInterpret = NULL;

TableA:

LOAD * INLINE [

    Key_A, Key_C, Key_D

    1, 1, 1

];

// Works if the following row is added to TableA

//  2, NULL, 2

TableB:

LOAD * INLINE [

   Key_A, Key_B, Key_C

   1    , 1    , 1

];

TableC:

LOAD * INLINE [

    Key_A, Key_B

    1    , 1

];

TableD:

LOAD * INLINE [

    Key_A, Key_D, Data_D

    2    , 2    , 2

];

[#2]

I got feedback from QlikTech and they stated the the synthetic key generation is more or less buggy and broken. Therefore no bug fixing will happen in this area and hence the only solution is to manually create id columns for linking, e.g. using auto number or hash functions.