Replace small reads in central directory parsing with one large read about rs-async-zip HOT 3 CLOSED

This sounds like a good idea. I guess this will then result in the need for a heap allocation, but it seems likely the overhead of this will be far lower than the overhead of individual read calls.

from rs-async-zip.

Great! I added a PR. It uses a slightly different strategy than the one described above.

It uses a BufReader instead of a single buffer. For most reasonable central directory sizes, the BufReader acts as a single buffer of size eocdr.size_cent_dir, but for larger ones it uses a capped buffer size. This helps avoid large allocations if eocdr.size_cent_dir is something like u32::MAX.

from rs-async-zip.

Handled within: #52

from rs-async-zip.