Hierarchical roll-up: hierarchy

Syntax

hierarchy ( op , hr { condition condComp { , condComp }* } { rule ruleComp } { mode } { input } { output } )

mode ::= non_null | non_zero | partial_null | partial_zero | always_null | always_zero

input ::= dataset | rule | rule_priority

output ::= computed | all

Input parameters

op	the operand Data Set
hr	the hierarchical ruleset to be applied
condComp	condComp is a Component of op to be associated (in positional order) to the conditioning Value Domains or Variables defined in hr (if any)
ruleComp	ruleComp is the Identifier of op to be associated to the rule Value Domain or Variable defined in hr
mode	this parameter specifies how to treat the possible missing Data Points corresponding to the Code Items in the right side of a rule and which Data Points are produced in output. The meaning of the possible values of the parameter is explained below.
input	this parameter specifies the source of the values used as input of the hierarchical rules. The meaning of the possible values of the parameter is explained below.
output	this parameter specifies the content of the resulting Data Set. The meaning of the possible values of the parameter is explained below.

Examples of valid syntaxes

hierarchy ( DS1, HR1 rule Id_1 non_null all )
hierarchy ( DS2, HR2 condition Comp_1, Comp_2 rule Id_3 non_zero rule computed )

Semantics for scalar operations

This operator cannot be applied to scalar values.

Input parameters type

op

dataset { measure<number> _ }

hr

name<hierarchical>

condComp

name<component>

ruleComp

name<identifier>

Result type

result

dataset { measure<number> _ }

Additional Constraints

If hr is defined on Value Domains then it is mandatory to specify the condition (if any) and the rule parameters. Moreover, the Components specified as condComp and ruleComp must belong to the operand op and must take values on the Value Domains corresponding, in positional order, to the ones specified in the condition and rule parameter of hr.

If hr is defined on Variables, the specification of condComp and ruleComp is not needed, but they can be specified all the same if it is desired to show explicitly in the invocation which are the involved Components: in this case, the condComp and ruleComp must be the same and in the same order as the Variables specified in in the condition and rule signatures of hr.

Behaviour

The hierarchy operator applies the rules of hr to op as specified in the parameters. The operator returns a Data Set with the same Identifiers and the same Measure as op. The Attribute propagation rule is applied on the groups of Data Points which contribute to the same Data Points of the result.

The behaviours relevant to the different options of the input parameters are the following.

First, the parameter input is considered to determine the source of the Data Points used as input of the Hierarchy. The possible options of the parameter input and the corresponding behaviours are the following:

dataset	For each Rule of the Ruleset and for each item on the right hand side of the Rule, the operator takes the input Data Points exclusively from the operand op.
rule	For each Rule of the Ruleset and for each item on the right-hand side of the Rule: · if the item is not defined as the result (left-hand side) of another Rule, the current Rule takes the input Data Points from the operand op · if the item is defined as the result of another Rule, the current Rule takes the input Data Points from the computed output of such other Rule
rule_priority	For each Rule of the Ruleset and for each item on the right-hand side of the Rule: · if the item is not defined as the result (left-hand side) of another rule, the current Rule takes the input Data Points from the operand op · if the item is defined as the result of another Rule, then: > if an expected input Data Point exists in the computed output of such other Rule and its Measure is not NULL, then the current Rule takes such Data Point; > if an expected input Data Point does not exist in the computed output of such other Rule or its measure is NULL, then the current Rule takes the Data Point from op (if any) having the same values of the Identifiers;

if the parameter input is not specified then it is assumed to be rule.

Then the parameter mode is considered, to determine the behaviour for missing Data Points and for the Data Points to be produced in the output. The possible options of the parameter mode and the corresponding behaviours are the following:

non_null	the result Data Point is produced when its computed Measure value is not NULL (i.e., when no Data Point corresponding to the Code Items of the right side of the rule is missing or has NULL Measure value); in the calculation, the possible missing Data Points corresponding to the Code Items of the right side of the rule are considered existing and having a Measure value equal to NULL;
non_zero	the result Data Point is produced when its computed Measure value is not equal to 0 (zero); the possible missing Data Points corresponding to the Code Items of the right side of the rule are considered existing and having a Measure value equal to 0;
partial_null	the result Data Point is produced if at least one Data Point corresponding to the Code Items of the right side of the rule is found (whichever is its Measure value); the possible missing Data Points corresponding to the Code Items of the right side of the rule are considered existing and having a NULL Measure value;
partial_zero	the result Data Point is produced if at least one Data Point corresponding to the Code Items of the right side of the rule is found (whichever is its Measure value); the possible missing Data Points corresponding to the Code Items of the right side of the rule are considered existing and having a Measure value equal to 0 (zero);
always_null	the result Data Point is produced in any case; the possible missing Data Points corresponding to the Code Items of the right side of the rule are considered existing and having a Measure value equal to NULL;
always_zero	the result Data Point is produced in any case; the possible missing Data Points corresponding to the Code Items of the right side of the rule are considered existing and having a Measure value equal to 0 (zero);

If the parameter mode is not specified, then it is assumed to be non_null.

The following table summarizes the behaviour of the options of the parameter “mode”:

OPTION of the MODE PARAMETER:	Missing Data Points are considered:	Null Data Points are considered:	Condition for evaluating the rule	Returned Data Points
Non_null	NULL	NULL	If all the involved Data Points are not NULL	Only not NULL Data Points (Zeros are returned too)
Non_zero	Zero	NULL	If at least one the involved Data Points is <> zero	Only not zero Data Points (NULL are returned too)
Partial_null	NULL	NULL	If at least the involved Data Points is not NULL	Data Points of any value (NULL or not NULL and zero too)
Partial_zero	Zero	NULL	If at least the involved Data Points is not NULL	Data Points of any value (NULL or not NULL and zero too)
Always_null	NULL	NULL	Always	Data Points of any value (NULL or not NULL and zero too)
Always_zero	Zero	NULL	Always	Data Points of any value (NULL or not NULL and zero too)

Finally the parameter output is considered, to determine the content of the resulting Data Set. The possible options of the parameter output and the corresponding behaviours are the following:

computed	the resulting Data Set contains only the set of Data Points computed according to the Ruleset
all	the resulting Data Set contains the union between the set of Data Points “R” computed according to the Ruleset and the set of Data Points of op that have different combinations of values for the Identifiers. In other words, the result is the outcome of the following (virtual)expression: union ( setdiff (op , R) , R )

If the parameter output is not specified then it is assumed to be computed.

Examples

Given the following hierarchical ruleset:

define hierarchical ruleset HR_1 ( valuedomain rule VD_1 ) is
    A = J + K + L
    ; B = M + N + O
    ; C = P + Q
    ; D = R + S
    ; E = T + U + V
    ; F = Y + W + Z
    ; G = B + C
    ; H = D + E
    ; I = D + G
end hierarchical ruleset;

And given the operand dataset DS_1 (where At_1 is viral and the propagation rule says that the alphabetic order prevails the NULL prevails on the alphabetic characters and the Attribute value for missing Data Points is assumed as NULL):

Input DS_1 (see structure)

Id_1	Id_2	Me_1	At_1
2010	M	2	Dx
2010	N	5	Pz
2010	O	4	Pz
2010	P	7	Pz
2010	Q	-7	Pz
2010	S	3	Ay
2010	T	9	Bq
2010	U		Nj
2010	V	6	Ko

Example 1

DS_r := hierarchy ( DS_1, HR_1 rule Id_2 non_null );

results in (see structure):

DS_r
Id_1	Id_2	Me_1	At_1
2010	B	11	Dx
2010	C	0	Pz
2010	G	19	Dx

Example 2

DS_r := hierarchy ( DS_1, HR_1 rule Id_2 non_zero );

results in (see structure):

DS_r
Id_1	Id_2	Me_1	At_1
2010	B	11	Dx
2010	D	3
2010	E		Bq
2010	G	11	Dx
2010	H
2010	I	14

Example 3

DS_r := hierarchy ( DS_1, HR_1 rule Id_2 partial_null );

results in (see structure):

DS_r
Id_1	Id_2	Me_1	At_1
2010	B	11	Dx
2010	C	0	Pz
2010	D
2010	E		Bq
2010	G	11	Dx
2010	H
2010	I