Pydons is a collection of manipulation add-ons for hierarchichal numerical data.

MatStruct class

MatStruct is an ordered dict with string-only keys, which are accessible also as properties. This makes the notation easier (obj.group.subgroup.variable instead of obj['group']['subgroup']['variable']) and enables IPython's auto complete.

MatStruct can be serialized to HDF5 or Matlab files using the excellent hdf5storage package.

LazyDataset class

A lazy evaluate proxy class for data sets in HDF5 or netCDF4 files.

FileBrowser class

FileBrowser employs MatStruct and LazyDataset to enable easy and fast browsing of netCDF4 or HDF5 files.

Examples

Items can be added using either ['keys'] or .properties:

import pydons
import numpy as np
struct = pydons.MatStruct()
struct['string'] = 'A string'
struct.list = ['list', 0, [1, 2]]
struct.numpy = np.random.rand(3,3)

IPython representation is customized:

In [12]: struct
Out[12]:
string: A string
  list: ['list', 0, [1, 2]]
 numpy: [[ 0.71539338  0.69970494  0.19328026]
 [ 0.28645949  0.15262059  0.23362895]
 [ 0.14518748  0.79911631  0.22522526]]

MatStruct can be serialized to HDF5 or Matlab files using saveh5 and savemat methods:

In [15]: struct.saveh5('struct.h5')
In [16]: pydons.MatStruct.loadh5('struct.h5')
Out[16]:
  list: ['list', 0, [1, 2]]
 numpy: [[ 0.71539338  0.69970494  0.19328026]
 [ 0.28645949  0.15262059  0.23362895]
 [ 0.14518748  0.79911631  0.22522526]]
string: A string

(the field order is not maintained---to be fixed soon).

This software is distributed under the MIT license (see the LICENSE file).

Documentation

pydons.readthedocs.org