Re: RFC: "-contains" option

[Eddie Kohler]

James Youngman wrote:
> On Sat, Feb 21, 2009 at 10:23 PM, Eddie Kohler <address@hidden> wrote:
> > exclude=$(find . -name .find-ignore -printf '%h\n' | sort -u); find . \! \(
> > -false $(for e in $exclude ; do printf ' -o -name "%s" ' $e; done) \) ,
> > -true > /dev/null
> >   => 3.06s
>
> There is a -prune missing from both find expressions there I think.
> The first can be pruned because once we find a .find-ignore file we
> don't need to search for more .find-ignore files in subdirectories.
> The second needs a -prune because that's the point of putting the
> .find-ignore file there.

I purposefully left off -prune from all expressions, including the -contains
one, so as to test only match overhead.

And how could you use -prune in the first expression?

>> exclude=$(find . -name .find-ignore -printf '%h\n' | sort -u)

By the time you have found the .find-ignore file it is too late to prune, you
are already in the directory.  Depending on how the directory's entries are
ordered you may have already gone through the subdirectories!

> It's fair to say I think that my concern around -contains has to do
> with the fact that it seems to be an extra feature designed for only
> one use case.   I avoid adding new features which only have limited or
> special-case uses.
>
> Can you think of a more generally useful (i.e. more generic) feature
> which might also lend itself to your problem?

Well I'll think about this, but I don't think -contains is special purpose.
A combined "-contains -prune" would only satisfy my use case, but as
defined in the patch, "-contains" is a test, and can be used in many
ways.  For instance here is a find of all directories containing a given file:

find . -contains file -print

All directories containing either file1 or file2:

find . \( -contains file1 -o -contains file2 \) -print

Of course these could be written in other ways now, but more obscurely.

Eddie