Read and save zarr files

Preface

zarr A data format , and hdf The file is a bit similar , That is, a file can contain many different dataset
And hdf Different documents , Intuitively hdf It's a single file , and zarr It's a folder , It also contains different subfolders （ amount to hdf Medium dataset）, Data is stored under the subfolder
Another point ,zarr Save data in blocks , Frankly speaking , It is to divide a whole block of data into sub blocks of the same size , Each sub block is saved as a file , Name it *.*.*, The advantage of this processing is that it is very friendly to large-scale medical images

There are also two special files .zarray and .zattrs, There are data characteristics, size information and so on , for example
.zarray

.zattrs

Read

import zarr
import numpy as np
input_name = 'data.zarr'
dataset_name = 'volumes/raw'
f = zarr.open(input_name)
raw = f[dataset_name ]
print(raw.shape)
raw_data = raw[:]

preservation

import zarr
import numpy as np
output_name = 'data.zarr'
dataset_name = 'volumes/raw'
root = zarr.open(output_name , mode='a')
shape = [1000, 1000, 1000] #  Size of the entire data ,3D
chunk_size = [128, 128, 128] #  The size of the blocks 
dtype = np.uint8 #  data type 
compressor = {
    'id': 'gzip', 'level': 5}
compressor = zarr.get_codec(compressor) #  Compression way 
dataset_name = dataset_name.lstrip('/')
ds = root.create_dataset(
         dataset_name,
         shape=shape,
         chunks=chunk_size,
         dtype=dtype,
         compressor=compressor)
ds.attrs['resolution'] = [8, 8, 8]
ds.attrs['offset'] = [0, 0, 0]
ds[:] = np.ones(tuple(shape), dtype=dtype)