TM1 Forum

Dear all,

We are wondering about two thing;
1) does suppress zero a true indicator of whether a cell has been fed.
2) when do we need feeders from start point to end point, in case of a serial of DB rules.

Explanation for 1)
There is a rule-derived consolidated cell, and its children cells.
What we are quite sure of are;
-its children cells are fed,
-the consolidated cell was a serial DB rules, and fed (cuz we feed at every DB chain, plus the start point to the end point

We use Suppress Zero, and the consolidated cell got suppressed, even it has a number. And this occur when no children has value.
However, when at least one child present value, the conso cell did not get suppressed.

It seems:
possibility 1; feeders statement works (meaning conso cell got fed) only when the children has value.
possibility 2; suppress zero does not accurately indicating whether a cell is fed or not. (a feature defect)
possibility 3; the conso cells are not fed.

What do you think?

Thank you very much.

Additional explanation for 1)
The conso cell, which derives from rules, do not depend on its children; not by rule, nor conso (children sum to conso).

Feeding only affects the N: level. Feeding a consolidated level will in fact feed all N: level cells underneath the consolidation. Writing rules that calculate on the C: level can be tricky as you have found. If all values underneath the consolidated level contain zeroes, then the zero suppression mechanism assumes a zero on the consolidated level - despite you overriding this zero with a C: level rule.
A way to get around this is to have a dummy (N: level) element underneath the C: level element and use this for storing and feeding the calculation on the C: level (giving this dummy element weight 1 in the consolidation and the other elements weight 0) - the C: level element will receive the value through the normal consolidation mechanism.

Michel

Goatship wrote:1) does suppress zero a true indicator of whether a cell has been fed.
2) when do we need feeders from start point to end point, in case of a serial of DB rules.

On point 1). Yes. Provided the rule calculated cell in question is a leaf cell or some of the descendant leaf cells (if it is a consolidated intersection) are both properly fed and evaluate to a non-zero value.

As has been pinted out already feeders are an exclusively leaf cell concept. What you are seeing is therefore exactly what you would expect to see (i.e. if the consolidated ruled cell evaluated to zero then it is suppressed and also it will be supressed even if it would evaluate to non-zero in the case that none of its leaf cell descendents are fed.)

On point 2). It isn't exactly clear what you mean by this. However in general it is better to only feed where you need to and to feed in the most direct way possible. If conceptually a starting value triggers a chain of downstream calculations and it is certain that given there is a startign value then all downstream calcs will also have a value then it is better to feed all end calc points in parallel directly from the source. If however various points of the calculation chain may or may not evaluate to zero and conceptually further calcs only need to happen in the non-zero case then you be better to feed in serial from each point to its immediate downstream dependent as this will limit overfeeding and consume less memory, give better performance etc.

For some calculation types feeding is not required at all. For example:
- C level rules where underlying leaf cells contain real data
- ratio type calculations that will always be queried directly and are not required to display in a zero supressed view
- "feederless" rules where dimension structure is used as an artificial construct in place of feeding (this is really just a special case of C level rules)

I'm not sure if that is what you were getting at or not but hope it helped.

Thanks a lot to both of you !! This is definitely helping our team.

I got one more wonder.

The event of zero suppression suppresses the parent cell when parent has a value and chilren has no value, will still occur, if we do not feed its children cells?

In other word, skipcheck will evaluate parent value to be zero value on a single criterion, that is, when none of its children has value, no matter whether its children are fed or not ?

Thanks