Hello,
On Tue, Sep 22, 2009 at 03:25:39AM +0200, olafbuddenha...@gmx.net wrote:
>
> I still don't understand why you think that the current hardcoded
> "descend into all directories, then match all subdirectories against
> foo" is more efficient than "descend into all directories matching *
> (i.e. all directories), then match all subdirectories against foo"?...
Hm, I think I can understand what exactly I failed to notice in your
questions. Let me try an explicit description of the original
functionality to avoid ambiguity.
Consider a directory stow/ with the following structure:
stow/
a/
foo/
something-else/
foo/
foo1/
b/
foo/
anything/
something/
foo/
The pattern being ``foo*''.
The original implementations does the following
1. List all proper subdirectories of a/ and b/ (i.e. foo/ ,
something-else/ , foo1/ , and foo/ , anything/ );
2. Check which of the obtained subdirectories matches the pattern and
union it (i.e. the following directories will be merged: foo/ ,
foo1/ , foo/).
The original implementation does not care about
stow/b/foo/anything/something/foo/ , although this directory matches
the pattern ``foo*''.
It takes the original implementation a single loop to do the pattern
matching.
In multicomponent matching with the optimization that you don't go
deeper than the number of components you have, matching ``*/foo'' will
indeed require the same number of operations.
The misunderstanding, IMHO, appeared mainly due to the fact that I
didn't universally take into account the fact that you don't need to
go deeper than the number of components in the pattern :-( Sorry :-(
Regards,
scolobb
