Similarity index improvements

[foxyshadis]

The similar files function is really cool. I've been trying it out, it's already helped find some dupes and neat similar stuff. So, of course I jotted down my thoughts about it. :p

I was thinking, a good way to browse across different sets might be a filmstrip across the top or side, similar to the filmstrip for browsing within a set on the bottom.

I'm unsure whether I like the idea of the filmstrip including the original as well (I think it would make it easier to explicitly delete the first, the current wording is somewhat confusing) and maybe a checkbox for acting on, instead of highlighting the ones you want. The current duplication between "<Action> Selected" and "<Action> Reference+Selected" seems unnecessary and somewhat confusing - does it delete only the original if nothing else is selected? Or does it delete everything? (The answer is the first, thankfully, but I had to check a couple times to be sure.) But the original in there might be somewhat confusing too, unless it was made distinct somehow.

Also, would it be possible to double click (or have an icon) on either the thumbnail or the image to view the full-screen/full-size version? Sometimes that's the only way to tell which is the higher quality variation.

It'd be cool if they were ordered by similarity (or choose name or similarity); having all the near 100%s in front seems more logical to me, but maybe not to others.

Lastly, I'm not sure how the algorithm works, but if it's a bit of data computed for every file and then compared, as opposed to comparing every file against every other one, could this be saved in the thumbnail cache so it isn't recomputed after being run once? And perhaps even optionally computed while making the thumbnails?

V 1.90 B1