naturalCompare Qt5 task

[John Layt]

On 19 July 2013 14:15, Aleix Pol <aleixpol at kde.org> wrote:
> On Fri, Jul 19, 2013 at 1:17 PM, Aleix Pol <aleixpol at kde.org> wrote:
>>
>> Hi,
>> I was looking at that task in the Qt5 epics list and I didn't understand
>> it fully.
>>
>> contribute natural-comparison to Qt5 (see KStringHandler). In Qt there is
>> naturalCompare function but private and not as good as from KStringHandler.
>> Thiago says: add the feature to QCollator.
>>
>> Here's the problems I see:
>> - I don't know what's the goal of the task, so it's hard for me to decide
>> what to do.
>> - The one Qt5 has now, it's in QFileSystemModel
>> - QCollator is also a private class, do we need to have it available from
>> the public API? I guess the only use of it is in any tier1 module. If so,
>> why QCollator?
>> - In the code we have in KStringHandler we can find this comment:
>> // This is based on the natural sort order code code by Martin Pool
>> // http://sourcefrog.net/projects/natsort/
>> // Martin Pool agreed to license this under LGPL or GPL.
>>
>> I don't feel very comfortable with moving this code over to the Qt
>> project.
>>
>> Aleix
>
>
> Hi,
> So after talking a bit with wojtask9.
>
> Apparently the thing is that our itemviews module needs
> KStringHandler::naturalCompare, and we want it to be tier1 (thus can't
> depend on KCoreAddons).
>
> Additionally, he showed me this review [1] that suggests moving QCollator to
> public API. This solves the 3rd point. I guess that then we can add a public
> naturalCompare function in QCollator for easy natural comparison.
>
> For the moment, I'll update the explanation and release the task since
> there's little we can do with a private QCollator.
>
> Aleix
>
> [1] https://codereview.qt-project.org/#change,50372

It's not an area I know much about, but QCollator is still private and
as far as I can see currently needs ICU to actually work.  I guess
that goes on my TODO list then as part of the ICU work.  If we can
find a Windows Win32 api that does the same thing or close enough we
should be able to move it forward more quickly.  Otherwise we will
need our own fallback implementation.

John.