Re: Allow to_date() and to_timestamp() to accept localized names

On 2019-Aug-22, Juan José Santamaría Flecha wrote:

> On Sun, Aug 18, 2019 at 10:42 AM Juan José Santamaría Flecha
> <juanjo(dot)santamaria(at)gmail(dot)com> wrote:
> >
> > Going through the current items in the wiki's todo list, I have been
> > looking into: "Allow to_date () and to_timestamp () to accept
> > localized month names".
>
> I have gone through a second take on this, trying to give it a better
> shape but it would surely benefit from some review, so I will open an
> item in the commitfest.

I'm confused why we acquire the MONTH_DIM / etc definitions. Can't we
just use lengthof() of the corresponding array? AFAICS it should work
just as well.

I wonder if the "compare first char" thing (seq_search_localized) really
works when there are multibyte chars in the day/month names. I think
the code compares just the first char ... but what if the original
string uses those funny Unicode non-normalized letters and the locale
translation uses normalized letters? My guess is that the first-char
comparison will fail, but surely you'll want the name to match.
(There's no month/day name in Spanish that doesn't start with an ASCII
letter, but I bet there are some in other languages.) I think the
localized comparison should avoid the first-char optimization, just
compare the whole string all the time, and avoid possible weird issues.

But then, I'm not 100% sure of this code. formatting.c is madness
distilled.

Thanks,

--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services