Re: RFC: locale-sensitive Emacs functions

[Ted Zlatanov <tzz@lifelogs.com>]

On Tue, 28 Mar 2017 07:59:25 -0400 Gdobbins <gdobbins@protonmail.com> wrote: 

>> I remember concluding that I was glad I didn't have to write code to support the '
>> flag, as it would require messing with the LC_NUMERIC locale setting (currently
>> hardwired to "C" for other good reasons) and would make for some work redoing
>> carefully-calculated output buffer sizes internal to 'format', not to mention
>> the character set conversion that would be required.

G> The way I've proposed should obviate those problems. Since the variables
G> governing the behavior are set by the user, LC_NUMERIC need not be consulted. If
G> it is deemed necessary a function like set-locale-dependent-vars could be
G> created to set these (and potentially other) variables, and the user could put
G> it in their init if they so choose.

Right, so you're *ignoring* the locale, and making "%'d" an
Emacs-specific thing because it doesn't follow the C behavior. I agree
that's a clean approach.

>> For what it's worth, the Common Lisp approach cannot handle the Indian numbering
>> system, which has a comma every two digits except that the last grouping
>> contains three digits (e.g., "12,34,56,789").

G> The literal CL approach can't handle it, but it can be adapted to do so. The
G> attached patch makes it so the format-digit-grouping variable can be a list. The
G> first element of the list controls the grouping size of the least significant
G> figures etc. until the last element controls all remaining groups. Using your
G> example:

G> (let ((format-digit-grouping '(3 2)))
G> (format "%'d" 123456789)) => "12,34,56,789"

I like the patch and the proposed approach, and if there are no
objections, and with some tests, I think it would be good to go.

Ted