Thread: Re: [PATCH] Rename trim_array() for consistency with the rest of array_* functions

On 29/10/2024 13:01, Aleksander Alekseev wrote:
> Hi hackers,
>
> Currently most of the functions dealing with arrays ( except for
> unnest() and cardinality() ) start with `array_` prefix [1] -
> array_length(), array_fill(), etc. Also this is how we name the new
> functions, e.g. recently proposed array_sort() [2] and array_reverse()
> [3]. The only exception from this rule is trim_array() which might be
> somewhat misleading. For instance, it is not show in the output of:
>
> ```
> \df array_*
> ```
>
> The proposed patch renames trim_array() to array_trim(). The old
> function is kept for backward compatibility but is marked as
> deprecated and is left undocumented. It can be removed in 10 years
> from now or so. To my knowledge this is how we typically deprecate old
> functions and operators.
>
> Thoughts?


While unfortunately named, we cannot deprecate TRIM_ARRAY() because it 
is mandated by the standard.

-- 

Vik Fearing

Hi Vik,

> While unfortunately named, we cannot deprecate TRIM_ARRAY() because it
> is mandated by the standard.

I didn't know that, thanks.

Still PostgreSQL doesn't follow the standard precisely in more than
one respect. Perhaps we should add array_trim() and keep both? Or
(less likely) remove trim_array() and document this deviation from the
standard? Or at least document why it's named this way.

What do you think?

-- 
Best regards,
Aleksander Alekseev

Hi Michael,

> On Tue, Oct 29, 2024 at 03:23:15PM +0300, Aleksander Alekseev wrote:
> >> While unfortunately named, we cannot deprecate TRIM_ARRAY() because it
> >> is mandated by the standard.
> >
> > I didn't know that, thanks.
>
> Wow.  Neither did I.
>
> > Still PostgreSQL doesn't follow the standard precisely in more than
> > one respect. Perhaps we should add array_trim() and keep both? Or
> > (less likely) remove trim_array() and document this deviation from the
> > standard? Or at least document why it's named this way.
>
> I suspect that it is not the only function in this case, and that we
> don't document that it is in the standard.  I would suggest to let
> things the way they are on HEAD and withdraw the patch.

Thanks for your feedback. The patch is withdrawn.

-- 
Best regards,
Aleksander Alekseev