Re: PATCH: _ps1234, _date_formats: Complete strftime formats for %D{}

[Mikael Magnusson <mikachu@xxxxxxxxx>]

On Mon, Jul 6, 2015 at 4:17 PM, Mikael Magnusson <mikachu@xxxxxxxxx> wrote:
> On Mon, Jul 6, 2015 at 1:11 PM, Oliver Kiddle <okiddle@xxxxxxxxxxx> wrote:
>> Mikael Magnusson wrote:
>>>
>>> PS, my strftime(3) says it handles %E and %O, and it does work with date +%Ey,
>>> but zsh just prints the literal %Ey. Is it handled wrong by us or not
>>> by the libc like the manpage claims?
>>
>> Looking at the source, it seems only - is treated as a modifier.
>> It'd be nice for it to be updated if it can remain portable.
>>
>>> +  compset -P "*"
>>> +  _describe -t date-format-specifier 'date format specifier' '(%)' -S ''
>>
>> Using _describe to complete just % seems rather overkill. An _wanted
>> would do. However, it'd be better to use a prefix for the % and use
>> compset -P to cut off complete format specifiers or unrelated
>> characters. The patch below does that and adds a case statement to
>> select some OS specific additions.
>>
>> It'd perhaps be nice to complete format specifiers and modifiers with a
>> different tag and description.
>
> I don't know how most of the things you did work, but as far as I can
> tell it only breaks one thing; pressing tab after %5 now completes all
> modifiers instead of just putting in a . since %. is the only one that
> takes a numeric argument. %- is also handled internally by zsh for
> some modifiers, so it'll work on any OSTYPE as long as $1 is zsh.

Oh, I see I didn't read far enough in strftime(3), and only checked
behaviour in zsh, not date. This is fun too btw:

% print -P %D\{%10A\}
%10A
% print -P %D\{%9A\}
       %9A

I guess this code could use some overall touchups.

It looks like the order of modifiers matter, so for example %-Ey works
in date, but %E-y does not.

Ah, date has a similar bug:
% date +%5Ey
   15
% date +%-Ey
15
% date +%-5Ey
   15
% date +%5-Ey
  %5-Ey
(the string was already five characters, but it still added extra
padding, and didn't expand the number, and if it was expanded, there
should be one more space).

-- 
Mikael Magnusson