Question about KUrl::equals/cmp...

[David Faure]

On Wednesday 04 November 2009, Dawit A. wrote:
> On Wednesday 04 November 2009 10:34:25 David Faure wrote:
> > On Wednesday 04 November 2009, Dawit A. wrote:
> > > Is there a reason why KUrl::equals with its second parameter set to
> > > KUrl::CompareWithoutTrailingSlash or KUrl::cmp's second parameter set
> > > to true returns "http://www.kde.org/" and "http://www.kde.org" are not
> > > equal to each other ?
> >
> > Ah, I think I know. Tricky: with http these urls are equal, while with
> > e.g. FTP they are not, since ftp://user@host redirects to
> >  ftp://user@host/home/user (on a linux server), while ftp://user@host/ is
> >  the root dir. Basically we treat "no path" as "please go to the default
> >  starting directory". But I'm ok with an exception for HTTP, where indeed
> >  people  expect these two urls to be the same.
> 
> Well I complely understand the different requirements of different
>  protocols, but why would one specificy the
>  KUrl::CompareWithoutTrailingSlash flag to make comparisons, like the ftp
>  case you mentioned above, when the trailing slash is important ? In other
>  words, why use the KUrl::CompareWithoutTrailingSlash flag in the first
>  place in such circumstances ?

KDirLister doesn't know about the "no path special case". It simply wants to 
know that, when it's told about /foo/bar/, and it knows about /foo/bar, they 
are the same item. So KDirLister does all directory-url comparisons with 
"CompareWithoutTrailingSlash". Well, technically it uses more often 
adjustPath(removeTrailingSlash), but same idea: that one also doesn't remove 
the single slash in case the path is just a '/'.

I gave FTP as an example above, now let me give "file" as an example.
Would you say that the local path "/" is equal to the local path ""? I would 
say no. The first one is the root dir, the second one is an empty path (which 
in the case of local files, doesn't represent any file or directory).

I think they are equivalent only for protocols that don't know anything about 
directories, like http :-). (But in kde4 we moved away from making KUrl depend 
on KProtocolInfo, so no way to write generic solutions like the previous 
sentence.)

> It does not seem the correct thing to do to make exceptions in the API for
>  a protocol specific need it already accomodates through existing options
>  unless of course I am completely missing something. Perhaps not breaking
>  existing applications that rely on the current behavior ?

Well, it would not seem correct to have to make exceptions in KDirLister and 
other places which compare paths often :-)

I'm a bit surprised that this comes up only now. What do you need this 
comparison for, and how is the same problem handled in the current khtml/konq 
code? (explicit check for empty path (on http) and making it "/", iirc, 
somewhere?)

-- 
David Faure, faure at kde.org, http://www.davidfaure.fr
Sponsored by Nokia to work on KDE, incl. Konqueror (http://www.konqueror.org).