RFC: bitmap storage for hidden

[Yuya Nishihara]

On Fri, 2 Sep 2016 22:05:58 +0100, Jun Wu wrote:
> Given the fact that the plain bitmap is just 1/512 the size of changelog
> revlog index (which is small) and we don't compress the latter, I won't
> worry too much about the file size at this time. If we use the plain bitmap,
> we can mmap it so the memory usage is easily optimized.
> 
> What we need here are a) "testing a bit" and b) "setting a bit". With a
> plain bitmap, they are both O(1), ignoring file system fragmentation.

Compressed bitmaps would be better if we have to c) iterate all high bits
or d) intersect two sets, though I have no idea how important these operations
are.