Re: Date format (was: How many people need locales?)

[Brian May <bam@debian.org>]

>>>>> "Radovan" == Radovan Garabik <garabik@melkor.dnp.fmph.uniba.sk> writes:

    Radovan> On Tue, Sep 04, 2001 at 09:17:12PM +1000, Martijn van
    Radovan> Oosterhout wrote:
    >> On Tue, Sep 04, 2001 at 08:56:54AM +0200, Radovan Garabik
    >> wrote:

    >> > > > > GNUmeric doesn't display the date in this format (it is
    >> up to the user > > to pick what format is used), BUT it does
    >> require entering dates in > > MM/DD/YYYY format.
    >> > > 
    >> > > - unless of course you have changed the locale...
    >> > 
    >> > I would file a bug against gnumeric in this case.  > This is
    >> broken behaviour. Does in behave the same > when you import
    >> data from text file? Then it > is even more broken.

Yes, it does. This was what started this investigation in the first
place: why doesn't gnumeric import dates as dates, not as strings.
Eventually we found this was the problem.

    Radovan> It should reject commonly used ambiguous formats (such as
    Radovan> NN/NN/NN, NN/NN/NNNN), or at least warn if something like
    Radovan> this is entered.  This should ideally be configurable,
    Radovan> but rejecting should be the default (and independent on
    Radovan> locale).

I agree with you here. Something in the form of NN/NN/NNNN that cannot
be parsed as a date should not get interpreted as a string.

If the user really does want a string (seems unlikely), he/she can
quote the value.

The error/warning message should indicate that the user can change the
format by changing the locale.
-- 
Brian May <bam@debian.org>