The ario.load(path) function returns a dict-like object which provides
functionality similar to numpy.load() but which works on a number of
archive formats (.zip, .tar with internal or external or no
compression and .npz) and a number of object file formats (.json as
well as .npy).
Supported archives include .zip, .npz and .tar and also tar files with
various compression. The .tar loader also supports the use of files
being internally compressed and the decompression can be handled
transparently.
If the format allows for it, ario will create or use an existing
archive index to allow for fast random access. Files that do not
support indexing are decompressed and loaded in their entirety into
memory.
Compressed file formats supporting efficient random access by ario
have these extensions:
.zip.npz.tarwith internal compressed files.tixaka.tar.pixzaka.tpxztar files withpixzindexes
Compressed file formats leading to their entire contents being greedily loaded have these extensions:
.tar.{xz,gz,bz2}and popular aliases (.txz, .tgz, .tbz2)