list all filenames in the zip archive twice in the worst case
check filename is correct or not by linear search twice in the worst case
Therefore, I'd like to cache filenames with the set structure to avoid redundant construction and improve search performance if the archive is opened in read-only mode.
When I tested https://github.com/pfnet/pfio/pull/316 PR with the ImageNet-1K dataset, I found that Zip(FS).stat() performance is pretty bad because:
Therefore, I'd like to cache filenames with the
setstructure to avoid redundant construction and improve search performance if the archive is opened in read-only mode.