"Relaxed/Fast" sorting optimization

[poutnik]

I have this idea:

Sorted disk is nice, but optimization is time/ resource expensive and stresses a disk.
So does small changes leads to heavy optimisation next time making sorting method unwanted as common task.

What about "Fast sorting optimization" ?

Fast optimization algorithm could be modified in these methods in a way,
what files will be chosen to fill gaps, either in defragment or optimization stage.

Only files that would be moved by fast optimization will be moved by fast sorting optimization.
But gaps near beginning of the disk will be filled by those fitting files,
that would be put near beginning of the disk by full sorting method.
EDIT: well, list of moved files would be little different, but amount of moved data about the same.

By other wording, method is like "let files stay where they are, but if they are moved, move them sorted."

The result of these "relaxed/fast sorting" will be similar move amount as in fast optimization,
but bias toward full sorted disk by each time disk is defragmented.

These method could be used as complementary incremental step,
after previously used initial full sorting optimization of the same kind.

Or, one can change his mind in sorting preference after previously used fast or sorting optimization
without extra prise of resorting.

[jeroen]

Thank you very much for sharing your idea, I appreciate it! The fast optimization at the moment will fill gaps with files from above. It looks for combinations of files that will perfectly fill a gap. When the program finds a combination it will stop searching. But to implement your idea the program would have to go keep searching and go through all possible combinations, and then determine the best. Computing all the possible permutations for just 32 chess pieces can give computers a nervous breakdown. Just imagine how many permutations there are for a few hundred thousand files.... I am sorry, it's a good idea in theory but will be too slow in practice.

An alternative would be to take the first file in the sorted list that will fit in the gap. I have actually tried this in the past, but then unsorted. It results in thousands of tiny little gaps that cannot be filled, because the smallest files are quickly used up. Gaps promote fragmentation, a few gaps are acceptable, but the method leaves too many gaps, especially after having used it for a while.

[poutnik]

I think hunting for the best choice is lose, just good one would be enough.
But I agree the risk of leaving gaps unfilled is higher than at normal fast one.