ozankabak commented on PR #7793:
URL: 
https://github.com/apache/arrow-datafusion/pull/7793#issuecomment-1764240790

   Hi @crepererum, thanks for taking a look. We actually had similar thoughts 
and did consider whether we can use `Interval`s in a similar fashion. We ended 
up using intervals whenever we can, but there are cases where something like 
`Precision` in its current form remains necessary. Here is what we do as of 
this PR:
   - Whenever we can elicit lower/upper bounds for a quantity, we use 
`Interval`. No need for another type -- a singleton interval already denotes 
exact information and we can represent the absence of any information with a 
universal interval (an interval containing all values) too.
   - There are cases where we only get an inexact/approximate point estimate, 
and we have no information on bounds. In these cases, a type like `Precision` 
is necessary. There are various ways this can arise, but one good example is as 
follows: Let's say I have two zero-mean quantities *A* and *B*. I know the 
value of *A* exactly (say 10), and I don't have any information on *B*. When 
using intervals, the sum *A* + *B* ends up in the no information state, which 
loses information. In such cases, having a value of `Inexact(10)` works better 
for many purposes.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to