Re: [PATCH][BZ #15527] strftime_l.c: Support lowercase output

[Mike Frysinger <vapier at gentoo dot org>]

On 09 Dec 2016 12:00, Rafal Luzynski wrote:
> 9.12.2016 05:07 Mike Frysinger wrote:
> > On 09 Dec 2016 02:14, Rafal Luzynski wrote:
> > > 7.12.2016 16:39 Mike Frysinger wrote:
> > > > [...] might be good to get wider opinion about new
> > > > strftime flags though.
> > >
> > > Now I lean a little against this change. As I wrote before,
> > > I'd love to see a flag to change the first letter to uppercase
> > > (or to titlecase, if that's more appropriate). A flag to change
> > > whole string to lowercase would be a workaround for this problem
> > > if we also converted all months and weekdays names to the
> > > titlecase in all languages, like in English now.
> >
> > i don't think changing the locale data or forcing titlecase everywhere
> > makes sense. the data is already in the standard format that users
> > expect for their locale. having a flag to support that seems like a
> > good way to get developers to naively enforce their own expectations
> > onto users. "my language writes things like 'Dec', therefore every
> > language out there must use titlecase".
> 
> That's really not what I meant. I meant that switching all locale
> data to titlecase would be possible only if there was a flag to
> switch back to lowercase.

i'm saying the flags are irrelevant.  the data should be in the form
that is appropriate for that locale.  that might mean uppercase, or
lowercase, or titlecase.  flags would not help at all, and in fact,
they would make things worse.  en_US authors would see that, think
that it's weird looking, and then use the titlecase flag.  then it'd
be broken for many other locales.

> Yes, in my language (and I think that in most languages) we have
> all months names and weekdays names in lowercase and this is correct
> by default. But we have no flag to switch to titlecase and
> I think it would be useful sometimes.

i have a hard time seeing a valid usecase where the code forcing
titlecase all the time would be desirable.  forcing all lower or
upper case isn't that hard to imagine (e.g. output forms where
*all* content, not just dates, are in the respective cases).

> > forcing to all uppercase or lowercase seems a bit more reasonable in
> > certain outputs (everything is upper/lower case beyond the date).
> >
> > > But in that case, do you guys think that converting to lowercase
> > > is useful if all letters are either lowercase already or should
> > > be always as they are now? Can you explain why would any application
> > > ever need the lowercase which should be provided by a format flag
> > > rather than converting programmatically? It seems to be useful
> > > only if some locales want to convert some data to lowercase and
> > > they don't have it lowercase out of the box.
> >
> > we already have a flag to force it to uppercase. makes sense to have
> > a flag to do the opposite. your arguments here apply to the uppercase
> > flag too.
> 
> No, does not. I didn't say anything about uppercase, if I had
> to say I'd say it's correct so let's leave it as it is now.
> What I wanted to say is that we don't have a flag to switch
> to titlecase.
> 
> My question was: what is the real problem solved by implementing
> a flag to switch to lowercase?

i didn't say you talked about uppercase.  my point is that your argument
isn't specific to lowercase.  you could s/lower/upper/ and have the same
points.  so why do you think having uppercase makes sense but adding
lowercase is wrong ?

> And my suggestion is that it would be useful only if a locale
> allows all-lowercase months or weekdays but locale data deliver
> non-lowercase. For example titlecase. This would solve two problems:
> - provide titlecase,
> - provide an easy way to generate lowercase.

locale data cannot be changed.  adding flags to transform their case
would not help the situation at all, just make it worse.
-mike

Attachment: signature.asc
Description: Digital signature