support key-value backends
version based on Julia 0.5
features
once an Associate type was defined and values were retrievable using Dict indexing, it could be used as a backend of BigArray. So it is highly extensible.
current backends include:
- google cloud storage
- aws s3
test
It was used for pinky40 dataset processing as a big test.
problems
speed
the speed is pretty slow due to single processing reading, should be extensible using multiple threads in the future. Current Julia do not support threading iteration of tuples!
compatibility of neuroglancer
current binary chunks were compressed using blosclz, and config file is config.json
while neuroglancer uses a json file called info
. The basic idea is the same, but implementation is different, so should be strait forward to modify for compatibility of neuroglancer.