This enhancement is related to optimizing In-Memory query to improve performance by removing the unnecessary join in the final pass. This optimization does not have impact on the data observed on the visualization.
SYMPTOM:
Here is an example, which uses Average function to calculate metric "Avg (Row Index)":
CSI Before Optimization:
Intelligent Cube SQL Statements:
Tables Accessed:
Table14 [F_FFSQL_ea_fact_date_fan_engagement05C4213611E81D910D940080EF551C60]: Fan Type Id, Date Id, Engagement Type Id, Page Id, Row Index, Fan Id, Row Count - FFSQL_ea_fact_date_fan_engagement, Frequency, FACT_TABLE
select [Date Id]@[date_id],
sum([Row Index]@[row_index])@{[Date Id]},
count([Row Index]@[row_index])@{[Date Id]} as [<dim metric>_1]
from FB_FFSQL_test
with Table Join Tree: [F_FFSQL_ea_fact_date_fan_engagement05C4213611E81D910D940080EF551C60]
Save As TempTable16
select [Date Id]@[date_id],
([TempTable16.<dim metric>] / [TempTable16.<dim metric>_1]) as [Avg (Row Index)]
from FB_FFSQL_test
with Table Join Tree: TempTable16
Join [F_FFSQL_ea_fact_date_fan_engagement05C4213611E81D910D940080EF551C60] with output level Tuple([Date Id]@[date_id], [Row Index]@[row_index])
Intelligent Cube SQL Statements:
Tables Accessed:
Table14 [F_FFSQL_ea_fact_date_fan_engagement05C4213611E81D910D940080EF551C60]: Fan Type Id, Date Id, Engagement Type Id, Page Id, Row Index, Fan Id, Row Count - FFSQL_ea_fact_date_fan_engagement, Frequency, FACT_TABLE
select [Date Id]@[date_id],
sum([Row Index]@[row_index])@{[Date Id]},
count([Row Index]@[row_index])@{[Date Id]} as [<dim metric>_1]
from FB_FFSQL_4Param_smaller
with Table Join Tree: [F_FFSQL_ea_fact_date_fan_engagement05C4213611E81D910D940080EF551C60]
Save As TempTable16
select [Date Id]@[date_id],
([TempTable16.<dim metric>] / [TempTable16.<dim metric>_1]) as [Avg (Row Index)]
from FB_FFSQL_4Param_smaller
with Table Join Tree: TempTable16