Typical case 1:

*IDDate*

*Month_ID*

*Month_Txt*

*DayWeek_ID*

*DayWeek_Txt*

*Year*

2016-03-01

3

March

2

Wendesday

2016

2016-03-02

3

March

3

Thursday

2016

2016-03-02

3

March

4

Friday

2016

IDDate is PK of Dim table and Unique


SOL 1: Uses Hierarchy and Derived from non PK column


*Month_ID*

Hierarchy 2

Normal 1

*Month_Txt*


Derived 1

*DayWeek_ID*

Hierarchy 3

Normal 2

*DayWeek_Txt*


Derived 2

*Year*

Hierarchy 1

Normal 3

Year > Month > Day

Text are derived from ID (in month and Week)

PB1: KYLIN-444 <https://issues.apache.org/jira/browse/KYLIN-444>

PB2: I don't know how create Derived column from non PK with actual UI (Kylin
– 1313 <https://issues.apache.org/jira/browse/KYLIN-1313> v1.5.2 Kylin 1786
<https://issues.apache.org/jira/browse/KYLIN-1786>, v1.5.3)



SOL 2:

*Month_ID*

Hierarchy 2

Join 1

*Month_Txt*


Join 1

*DayWeek_ID*

Hierarchy 3

Join 2

*DayWeek_Txt*


Join 2

*Year*

Hierarchy 1

Normal 3


SOL 2 is this the best solution ??



Typical case 2:

I see the same scenario a lot of times (derived columns with 1:1 Relation)

Product_ID *(PK)*

Product_TXT

TypeProduct_ID

TypeProduct_TXT

Country_TXT

Country_ID

Optimize queries by product / category / country, are mandatory

Perhaps,

Country (lower cardinality) its a good candidate to Join

I don't want put Product_TXT as Join, because is a longgggg text, and can
be affect Row_Key of HBase, but I need Queries like ... where product_TXT =
""iRobot Roomba 650 Robotic Vacuum Cleaner

suggestions ?

Reply via email to