Consistent prefixing of Original Name path + Import speedup

[tilusnet]

Hello,

I am combining two suggestions in this note as they are closely related in a use case I am going to describe next.

Assume that I have a path in my import folder a/b/c/d.

When I import from root level, ie. "all files from import folder", images get a nice "a/b/c/d" path prefix in Original Name. Nice and useful.

Next I add some new files under d/, therefore this time I import from under a/b/c/d/ only, to speed things up.
My surprise is that the newly imported files no longer have the "a/b/c/d" path prefix, despite the Import root did not change. This is inconsistent. Likewise, if I reimport with path selected "a/b/c", new images under "a/b/c/d" will only get a d/ prefix under Original Name. The path prefix is relative to the selected import subpath, and not the Import root as I'd expect.
Why was it designed the way it is?

So the workaround regarding Original Name path prefix consistency seems to always import from root. But then why allowing subpath selection for import?
And always reimporting from Import root is very slow: I assume the algo is checksuming all files it encounters, ie. it reads the binary streams of all content during every import. An absolute overkill on hundreds of GB of content.

Therefore I suggest some optimisations:

1. Consistency: Original Name relative path should resolve against Import root, regardless of the subpath Import is started from
2. Speed: Do a smarter scan on files, don't always checksum the binary streams.
   To illustrate, Unix rsync likewise does only a quick check by default, and not checksuming, as this will quickly detect if a subpath has not been touched and it will skip it, speeding up the sync process considerably.