使用 Big Query SQL 我可以实现以下减法（减去两行值并将其替换为该值）？

Question

application	buildingid	dashboarding_tag	value
Abacus	BLD01	Heating_System_Energy_Consumption	3000.00
Abacus	BLD02	Heating_System_Energy_Consumption	1000.00
Abacus	BLD04	Heating_System_Energy_Consumption	1000.00
Abacus	BLD05	Heating_System_Energy_Consumption	500.00
Abacus	BLD02	MVHR_Energy_Consumption	300.00
Abacus	BLD03	MVHR_Energy_Consumption	900.00
Abacus	BLD04	MVHR_Energy_Consumption	50.00
Abacus	BLD05	MVHR_Energy_Consumption	80.00
Abacus	BLD06	MVHR_Energy_Consumption	200.00
Forell	BLD07	Heating_System_Energy_Consumption	0.00
Forell	BLD08	Heating_System_Energy_Consumption	50.00
Forell	BLD09	Heating_System_Energy_Consumption	-5.00
Forell	BLD10	Heating_System_Energy_Consumption	-30.00

to

application	buildingid	dashboarding_tag	value
Abacus	BLD01	Heating_System_Energy_Consumption	3000.00
Abacus	BLD02	Heating_System_Energy_Consumption	700.00
Abacus	BLD04	Heating_System_Energy_Consumption	950.00
Abacus	BLD05	Heating_System_Energy_Consumption	420.00
Abacus	BLD02	MVHR_Energy_Consumption	300.00
Abacus	BLD03	MVHR_Energy_Consumption	900.00
Abacus	BLD04	MVHR_Energy_Consumption	50.00
Abacus	BLD05	MVHR_Energy_Consumption	80.00
Abacus	BLD06	MVHR_Energy_Consumption	200.00
Forell	BLD07	Heating_System_Energy_Consumption	0.00
Forell	BLD08	Heating_System_Energy_Consumption	50.00
Forell	BLD09	Heating_System_Energy_Consumption	-5.00
Forell	BLD10	Heating_System_Energy_Consumption	-30.00

Notice how the values in row 2,3 and 4 changed on table 2 based on values in row 5,7,8.

The logic is, if the application is Abacus, then always replace the value of the Heating_System_Energy_Consumption such that:

Heating_System_Energy_Consumption=Heating_System_Energy_Consumption-MVHR_Energy_Consumption for each building

Answer 1

Consider below approach

select * except(value), 
  if(dashboarding_tag = 'Heating_System_Energy_Consumption', 
    value - sum(if(dashboarding_tag = 'MVHR_Energy_Consumption', value, 0)) over(partition by application, buildingid), 
  value) as value
from your_table

if applied to sample data in your question - output is