An even better implementation of 22.2.6.2.2 ??

[Nathan Myers]

On Tue, Feb 19, 2002 at 10:19:21AM +0100, Paolo Carlini wrote:
> if I understand well what results from the discussion of yesterday, we
> should not only revert my "fix" for libstdc++/5708, but also
> unconditionally output a "plain" space for the space field (this is
> required by the presence of the space field itself and has nothing to do
> with the filling with the fill char). That is:
> 
>   case money_base::space:
>     // At least one space is required, but if internal
>     // formatting is required, an arbitrary number of
>     // fill spaces will be necessary.
>     __res += __ctype.widen(' ');
>     if (__testipad)
>       __res += string_type(__width - __len - 1, __fill);
>     break;

More or less.  (I hate calling widen over and over again, and doing 
all kinds of memory allocations during I/O, but that's a whole 'nother 
issue.)  We need to think about whether the fill characters go before 
or after the space, because the standard is entirely unclear on the 
point.  Generally, they probably shouldn't go between the space and 
the currency symbol, because the space is there to separate the 
currency symbol from other stuff.  Thus,

  1000****** DMK
  ITL ******1000

This is purely a QOI matter, but once we choose we're probably
stuck with it.

Nathan Myers
ncm at cantrip dot org