"Kuiper, Arnout" wrote:
> From: Bill Petheram [mailto:[EMAIL PROTECTED] > > > The main problem with this approach is that it only allows > > > recursion on the end of the pattern, not on the beginning, > > > which is essential for a number of situations Ant has to > > > deal with, like filtering out CVS directories, filtering > > > out test directories, etc. > > > > Normally to exclude CVS directories you would use an exclude pattern. > > I.E don't match any path that contains this pattern. > > Of course it is an exclusion pattern that does the filtering, > but that pattern must match any CVS directory on the end of a path. > Therefore you need "recursion" at the beginning, which cannot be > supplied by any other method that has been proposed. You are using a special syntax inside path specifications. In my view you should have path specification conforming to shell syntax then apply a set of regular expressions to each found lement from the path syntax. > > > > But it is different from anything else. > > Because the other approaches don't give that little bit extra... > Should we refrain from using something more powerful, just because > it's new and a bit different? Maybe you haven't looked at ther approaches properly. tar, find work OK. > > > I think you are confusing two issues. One is pattern matching and the > > other is specifying paths. > > Your proposal does both. Whereas they can be separated. > > I doubt that, I don't think you understand the power of this matching > concept. You match on complete paths instead of just filenames. This can > only be done if your path is part of the pattern. Therefore > you cannot separate paths from patterns. If you can convince me > otherwise, please do. > What about path=test recurse=true ignore="\/CVS\/","~$","tmp$" > > > I say again at the cost of users having to do something new and > > different and unlikely to be used by any other package. > > Who says it won't be picked up by other packages, once they understand > the power. It has to start somewhere;-) > Because there is not a clear separation between paths and regular expressions. > > Arnout cheers bill
