alamb commented on issue #8078:
URL: 
https://github.com/apache/arrow-datafusion/issues/8078#issuecomment-1811098404

   >  I think what you are trying to convey is that @berkaysynnada's proposal 
can not represent the triple of (lower bound, guess, upper bound). If this is 
what you mean, I agree with you.
   
   Yes, this is what I was trying to say. Thank you
   
   > I don't understand what you mean by this. In this snippet, lower/upper 
bounds refer to column bounds, while estimate refers to row count. Did you mean 
to write:
   
   Yes, I think so (though I am not 100% sure about the 62.5 probability -- I 
need to double check my math)
   
   
   > Do we need the Exact variant anymore? It seems no. If lower, upper and 
estimate values are the same, it would be an exact estimate. This struct can 
implement a method is_exact to check this.
   
   That is a good point
   
   > Do we need theAbsent variant anymore? I think yes. We may want to 
differentiate between "no information" and "all possible values" (i.e. [MIN, 
MAX]) in certain cases.
   
   I agree this is probably a useful special case. 
   
   > Should we use Interval in the bounded variant? I think yes, but we should 
probably start off without it and migrate when we move the interval library 
outside of the physical expression crate (this week hopefully).
   
   Yes, the more I play with the code, the more I think using Interval is 
important to avoid reimplementing the same thing over and over again
   
   > Should we use the name Precision still? IMO it is not a good name for 
this. We are making estimates after all. I think we should use the right name 
for what we are doing and use the name Estimate.
   
   I agree the term `Precision` is not a good match.  I think `Estimate` is 
better, though I am still not entirely happy with that because in some cases it 
will be "known exactly" rather than an estimate. Though perhaps this case is 
technically just a very precise estimate (no error) 🤔 


-- 
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