Hive case when 子句效率

Question

Is there difference between:

select
  case some_calculation() 
  when 'a' then 1
  when 'b' then 2
  else 0
  end

and

select
  case
  when some_calculation() ='a' then 1
  when some_calculation() ='b' then 2
  else 0
  end

I assume in the second version, the some_calculation() function will be evaluated twice. I don't know how to verify that. Any input will be appreciated.

Answer 1

This is not well-documented in Hive. But it is well understood in the standard and in other databases.

In the second version, some_calculation() will definitely be evaluated multiple times. A case expression is evaluated sequentially, so the second when is not evaluated until the first evaluates to false.

In the first version, the value should be evaluated once.

You can get a flavor for the difference by using a volatile functions, such as random() . This db<>fiddle illustrates the difference between the two approaches. The first version never returns NULL . The second returns lots of them.