KDE/kdevplatform/plugins/subversion

[dukju ahn]

2007/7/29, Andreas Pakulat <apaku at gmx.de>:
> On 29.07.07 06:54:16, dukju ahn wrote:
> > 2007/7/29, Andreas Pakulat <apaku at gmx.de>:
> > > On 29.07.07 04:58:06, dukju ahn wrote:
> > > > 2007/7/28, Andreas Pakulat <apaku at gmx.de>:
> > > > > On 28.07.07 18:30:14, Dukju Ahn wrote:
> > > > > > --- trunk/KDE/kdevplatform/plugins/subversion/svn_revision.h #693699:693700
> > > > > > @@ -24,6 +24,7 @@
> > > > > >  #define SVN_REVERT   (KDevelop::VcsJob::Revert)
> > > > > >  #define SVN_COPY     (KDevelop::VcsJob::Copy)
> > > > > >  #define SVN_MOVE     (KDevelop::VcsJob::Move)
> > > > > > +#define SVN_CAT      (KDevelop::VcsJob::Cat)
> > > > >
> > > > > Whats this? Apart from the fact that you didn't commit the change in
> > > > > vcsjob and thus broke the build, why these defines? Whats the use case
> > > > > for that?
> > > >
> > > > Because svn has its unique operations that are not counted
> > > > by our common interface. In most cases its ok but, think about
> > > > "svn switch", "svn info". The VcsJob will not define enums for
> > > > these operations, but subversion plugin still needs this.
> > > >
> > > > So I had no choice but to define job types again.
> > >
> > > I'm not questioning wether you need to invent new enum values, but I'm
> > > questioning the use of the #define's here. #define is _not_ a proper
> > > enum and I don't see a reason why you have them and not use
> > > KDevelop::VcsJob::Cat and Co directly - or SvnJob::Info, SvnJob::Switch.
> >
> > But why is the #define too bad?
>
> Because they are unneeded and IMHO its just the wrong tool for the job.
> enum is a far better tool.

It's needed for uniformity. Suppose that in somewhere we use
VcsJob::Add and in other place we use SvnJob::Info. They are different
type. So I can't, at lease, use VcsJob enums directly.
I need to redefine every enums in a single type.

> > It doesn't do any harm as far as the values are defined correctly.
> > What is the benefit if I redefine everything into enum values such as
> > SvnJob::commit ..?
>
> For example: proper typing and type checking by the compiler, not
> polluting the global namespace because the enum values would be
> "namespaced" by the class name in which they are defined.

In point of syntax it is true.  But I'm not clear why it should be
confined to namespace. Also, the type is regarded just "int". Then
there is no problem is typechecking.