@xccui if you think using `<i>` is better I'm fine with that, I don't mind one 
way or the other.

However I don't understand:

> before reaching a consensus about the format and the internal implementation, 
> I suggest to keep the current doc style, i.e., without mentioning the return 
> type, for now.

We are already following a convention that for all of the math functions we are 
dropping precision and returning `double`. What's the point of not documenting 
it?

If you could guarantee me that we will document those return types lets say 
within a month, I would be fine with postponing it. However I don't believe 
something like that would happen, so I prefer to incrementally start adding 
this return type information right now. Later if someone will want, we can 
reformat the page. I do not see any downside to such incremental approach until 
that moment.

> why still indicating the return type for each numeric function explicitly? 
> How about adding a single note for that?

since some of them can be implemented precisely (`abs`, `min`, `max`, `ceil`, 
...), others can be for some variants (`power(DECIMAL, INT)`). We can state 
this general rule that functions returning `DOUBLE` are approximations, but we 
still need specific information for every function, whether it returns 
approximated (`DOUBLE`) or precise (`DECIMAL`) result. 


[ Full content available at: https://github.com/apache/flink/pull/6700 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to