Re: Monotonic WindowFunc support for ntile(), percent_rank() and cume_dist() - Mailing list pgsql-hackers

From Melanie Plageman
Subject Re: Monotonic WindowFunc support for ntile(), percent_rank() and cume_dist()
Date
Msg-id 20230124133858.vw6bi2sot2cbljyh@liskov
Whole thread Raw
In response to Re: Monotonic WindowFunc support for ntile(), percent_rank() and cume_dist()  (David Rowley <dgrowleyml@gmail.com>)
Responses Re: Monotonic WindowFunc support for ntile(), percent_rank() and cume_dist()
List pgsql-hackers
On Tue, Jan 24, 2023 at 02:00:33PM +1300, David Rowley wrote:
> Thanks for having a look at this.
> 
> On Tue, 24 Jan 2023 at 13:26, Melanie Plageman
> <melanieplageman@gmail.com> wrote:
> 
> > Since all three cases are exactly the same code, maybe you could
> > macro-ize it and add a single comment?
> 
> Hmm, I kinda like that it's being spelt out explicitly. To me, it
> seems clean and easy to read. I know we could have fewer lines of code
> with something else, but for me, being able to quickly see what the
> properties of the WindowFunc are without having to look at some other
> function is more important than saving some space in windowfuncs.c
> 
> I'd likely feel differently if the code in question didn't all fit on
> my screen at once, but it does and I can see at a quick glance that
> the function is unconditionally monotonically increasing. Functions
> such as COUNT(*) are conditionally monotonically
> increasing/decreasing, depending on the frame options.
> 
> If you feel strongly about that, then feel free to show me what you
> have in mind in more detail so I can think harder about it.

Nah, I don't feel strongly. I think it was because looking at the patch
in isolation, the repetition stands out but in the context of the rest
of the code it doesn't.

- Melanie



pgsql-hackers by date:

Previous
From: David Geier
Date:
Subject: Re: Reduce timing overhead of EXPLAIN ANALYZE using rdtsc?
Next
From: Aleksander Alekseev
Date:
Subject: Re: to_hex() for negative inputs