More power to filebooleans

[Kasuha]

My idea is, syntax "like" regular expressions or regular expressions themselves could be used in filebooleans. For example, instead of

      FileName("*.7z")
      or FileName("*.arj")
      or FileName("*.avi")
      or FileName("*.bak")
      or FileName("*.bup")

MyDefrag could support something like

      FileName("*.7z|*.arj|*.avi|*.bak|*.bup")

evaluating this in a single run (using a single search tree) which would make evaluating such complicated filebooleans much faster. And it would be also nice to have support for full extended regular expression:

      FileNameRegex("\.7z$|\.arj$|\.avi$|\.bak$|\.bup$")

as regular expressions are much more powerful than filenames with wildcards.
There are already free implementations supporting such searches available, too.

[jeroen]

Thanks for sharing your idea, I appreciate it. I have looked at using regular expressions in the beginning of MyDefrag and tried a couple of free implementations. But it is an order of magnitude too slow. Even on a small harddisk there are typically hundreds of thousands of files, and running a regular expression for so many inputs simply takes too much time. Regular expressions work great for comparing a pattern with a single big input, not so great for comparing a pattern with lot's of small inputs, such as filenames. Searching for a solution for the speed problem I arrived at the current filebooleans, which run through all filenames in a tight loop with basic string comparions.

[Kasuha]

That's strange as I'm pretty sure something we are using is able to perform hunders of thousands searches per second. I'll try to check what are we using there.

[Kasuha]

I have found we are using just GNU libc regex library, particularly the utf8regex port of it (http://jwilk.net/software/utf8regex).