Question: Directory versioning

[Mark]

On 7/2/05, Matt Mackall <mpm at selenic.com> wrote:
> The question was, will Mercurial ever treat directories as first class
> objects? Currently the Mercurial back end has no concept of
> directories _at all_. The manifest only contains files, their full
> path, and an executable flag.
> 
> Directories spring into being when you check things out that path
> components that refer to them. Generally everything works out fine as
> they're entirely represented by their contents.
> 
> Except in the case of empty directories, which have no contents. This
> means there are no files around to refer to it. Thus there is no way
> to represent the concept of an empty directory in Mercurial _at all_.
> 
> So the question then becomes, is it worth complicating everything in
> the back end with the concept of versioned directories so that
> we can check in empty directories?

I don't see empty directories as my primary motivation for asking
about them.  The scenerio  that I described a earlier in this thread
where a "hg update <tag>" is done in an existing working copy leave
behind directories that are from later in the history than the tag. 
This results in a working copy that does not match the tag exactly.

Perhaps there is another way to achieve this goal than directory
versioning, if so great.  My primary motivation though is when I do a
"hg update" to a partucular revision of the repository I don't want to
see artifacts from revisions that have not yet occured.

-Mark