Re: Spotlight -- Metadata importer vs file package containing rtfd files

FROM : Kirk Kerekes
DATE : Wed May 04 19:53:08 2005

Just an update on this topic in case anyone else is interested...

It appears that Spotlight and the rest of the MD system have 
implemented a "path filter" that aggressively prevents the use of MD 
queries inside of Finder-opaque file packages.

Thus, not only does SpotLight not search inside of Finder-opaque 
packages (which I can sort-of understand), but you can't even _force_ 
anything inside such a package to be indexed by using any aspect of 
the public API that I have been able to find.

As a result, an application that encapsulates standard documents 
(like rtf or rtfd files) inside a file-package cannot use 
NSMetadataQuery to retrieve metadata for the file package.

This explains why the default mdimporter for RTFD files fails to 
index attachments to the rtfd file -- it can't use the OS to index 
the files, which could be anything.

To provide a specific example:

If you have an item, say ~/foo.bundle, and you create a query 
constrained to the bundle folder, the _content-indexing_ part of that 
query will always return zilch, whether you perform the query with 
the Spotlight UI or via NSMetadataQuery. Even if you attempt to index/
search ~/foo.bundle/Contents, the query will return zilch for file 
content searches.

You can easily demonstrate this for yourself using rtfd files with 
attachments and the "show package contents" contextual menu item in 
the Finder: If you open a Finder window on the RTFD package contents, 
the window's Find bar is suddenly incapable of finding the contents 
of the TXT.rtf file that contains the rtfd file's text when the 
search is constrained to window's directory. It can find the 
filenames, but ignores file-content.

While I can certainly understand not having most opaque bundles 
search by default, particularly during Spotlight's initial mass-
indexing phase, having the code-level API suffer from the same 
constraint appears to be dubious at best.

Note that there does not appear to be any fundamental technical issue 
here -- the mdimport tool that does the indexing has a -f flag that 
"forces" it to index inside of opaque file packages. The only reason 
for the constraint appears to be that nobody considered this issue 
when the API was laid out, and nobody reconsidered it when the 
default rtfd importer had to be crippled (compared to previous 
versions of content-indexing, which indexed rtfd attachments) because 
of the API limitation.

IMHO, the API should be augmented with a searchOpaqueDirectories flag.

Until then, I would welcome anyone's knowledge of a confirmed work-
around.