grosskc / hdfdict

Helps h5py to dump and load dictionaries.

hdfdict helps h5py to dump and load python dictionaries

If you have a hierarchical data structure of numpy arrays in a dictionary for example, you can use this tool to save this dictionary into a h5py File() or Group() and load it again. This tool just maps the hdf Groups to dict keys and the Datset to dict values. Only types supported by h5py can be used. The dicitonary-keys need to be strings until now.

A lazy loading option is activated per default. So big h5 files are not loaded at once. Instead a dataset gets only loaded if it is accessed from the LazyHdfDict instance.

Example

import hdfdict
import numpy as np


d = {
    'a': np.random.randn(10),
    'b': [1, 2, 3],
    'c': 'Hallo',
    'd': np.array(['a', 'b']).astype('S'),
    'e': True,
    'f': (True, False),
}
fname = 'test_hdfdict.h5'
hdfdict.dump(d, fname)
res = hdfdict.load(fname)

print(res)

Output: {'a': <HDF5 dataset "a": shape (10,), type "<f8">, 'b': <HDF5 dataset "b": shape (3,), type "<i8">, 'c': <HDF5 dataset "c": shape (), type "|O">, 'd': <HDF5 dataset "d": shape (2,), type "|S1">, 'e': <HDF5 dataset "e": shape (), type "|b1">, 'f': <HDF5 dataset "f": shape (2,), type "|b1">}

This are all lazy loding fields in the result res. Just call res.unlazy() or dict(res) to get all fields loaded. If you only want to load specific fields, just use item access e.g. res['a'] so only field 'a' will be loaded from the file.

print(dict(res))`

Output: {'a': array([-0.47666824, 0.11787749, 0.51405835, -1.49557787, -0.33617182, -0.22381693, 0.25966526, 0.58160661, 0.17019176, 1.3167669 ]), 'b': array([1, 2, 3]), 'c': 'Hallo', 'd': array([b'a', b'b'], dtype='|S1'), 'e': True, 'f': array([ True, False])}

Installation

• pip install hdfdict
• poetry install hdfdict
• git clone https://github.com/SiggiGue/hdfdict.git and python hdfdict/setup.py install

Modifications from original source

Ensure h5 files are opened as read-only when loading data. This keeps the file modification date from changing whenever a file is opened. It also makes h5py happy.