Calculates the compression for a given dandi-set

calculate_nwb_compression.py

import h5py
from dandi.dandiapi import DandiAPIClient
from pynwb import NWBHDF5IO
import pynwb
import numpy as np

dandi_dataset_id = "000166"
dandi_dataset_id = "000213"

with DandiAPIClient() as client:
    dandi_set = client.get_dandiset(dandi_dataset_id, "draft")

# Order assets by size for faster speed
asset_list = list(dandi_set.get_assets())
asset_list.sort(key=lambda x:x.size)

number_of_sessions = 8
asset_size_dic = {}
compression_ratio_dic = {}

for index in range(number_of_sessions):
    print(index)
    asset = asset_list[index]
    asset_size_gb = asset.size / 10 ** 9
    asset_size_dic[index] = asset_size_gb
    print(asset.path)
    
    s3_path = asset.get_content_url(follow_redirects=1, strip_query=True)
    with NWBHDF5IO(s3_path, mode='r', load_namespaces=True, driver='ros3') as io:
        nwbfile = io.read()
        compression_ratio_list = []
        for key1, obj in nwbfile.objects.items():
            for key2, field in obj.fields.items():
                if isinstance(field, h5py.Dataset) and isinstance(obj, pynwb.ecephys.ElectricalSeries):
                    compression_ratio = field.id.get_storage_size() / (field.size * field.dtype.itemsize)
                    compression_ratio_list.append(compression_ratio)
        compression_ratio_dic[index] = compression_ratio_list
        
compression_averages_dic = {index:np.mean(compression) for index, compression in compression_ratio_dic.items()}
overall_average = np.mean(list(compression_averages_dic.values()))
print("""""")
print("Asset size")
print(asset_size_dic)
print("Compression averages")
print(compression_averages_dic)
print("Average of averages")
print(overall_average)