Squeeze out all of the data from an nwb file

skinnify_nwb_file.py

import numpy as np
import h5py
import os

def pass_function(*args): return

def visit(group, group_callback=pass_function, 
                 dataset_callback=pass_function, 
                 softlink_callback=pass_function, 
                 broken_link_callback=pass_function, path='/'):

    for key in group:
        try:
            val = group[key]

            link_type = group.get(key, getlink=True)
            if isinstance(link_type, h5py.SoftLink):
                softlink_callback(val.name, link_type.path)

            elif isinstance(val, h5py._hl.group.Group):
                group_callback(val)
                visit(val, group_callback=group_callback, 
                           dataset_callback=dataset_callback, 
                           softlink_callback=softlink_callback, 
                           broken_link_callback=broken_link_callback, path=os.path.join(path, key))
            else:
                dataset_callback(val)
        except KeyError:
            broken_link_callback(os.path.join(path, key))

f_src = h5py.File('496908818.nwb', 'r')
f_tgt = h5py.File('new.h5', 'w')

def group_callback(group):
    
    assert isinstance(group, h5py._hl.group.Group)
    
    attr_dict = dict(group.attrs)
    new_group = f_tgt.create_group(group.name)

    for dkey, dval in attr_dict.items():
        new_group.attrs[dkey] = dval


visit(f_src, group_callback=group_callback)
f_src.close()
f_tgt.close()



f_src = h5py.File('496908818.nwb', 'r')
f_tgt = h5py.File('new.h5', 'r+')

def dataset_callback(dataset):

    assert isinstance(dataset, h5py._hl.dataset.Dataset)

    curr_shape = dataset.shape
    dtype = dataset.dtype
    attr_dict = dict(dataset.attrs)

    if len(curr_shape) == 0:
        data = dataset
    elif len(curr_shape) == 1:
        data = np.array([1.], dtype=dtype)
    elif len(curr_shape) == 2:
        data = np.array([[1.]], dtype=dtype)
    elif len(curr_shape) == 3:
        data = np.array([[[1.]]], dtype=dtype)
    else:
        print len(curr_shape)
        raise NotImplementedError
    dset = f_tgt.create_dataset(dataset.name, data=data)
    for key, val in attr_dict.items():
        dset.attrs[key] = val

def softlink_callback(source, target):
    f_tgt[source] = h5py.SoftLink(target)

def broken_link_callback(source_path):
    data = np.array([1.])
    f_tgt.create_dataset(source_path, data=data)




visit(f_src, dataset_callback=dataset_callback, 
             softlink_callback=softlink_callback, 
             broken_link_callback=broken_link_callback)
f_src.close()
f_tgt.close()