123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554 |
- # -*- coding: utf-8 -*-
- """
- Module for reading and writing NSDF files
- Author: Mieszko Grodzicki
- This module support both reading and writing NDSF files.
- Note: Read file must be written using this IO
- """
- from __future__ import absolute_import
- import numpy as np
- import quantities as pq
- from uuid import uuid1
- import pickle
- from datetime import datetime
- import os
- try:
- import nsdf
- except ImportError as err:
- HAVE_NSDF = False
- NSDF_ERR = err
- else:
- HAVE_NSDF = True
- NSDF_ERR = None
- from neo.io.baseio import BaseIO
- from neo.core import Block, Segment, AnalogSignal, ChannelIndex
- class NSDFIO(BaseIO):
- """
- Class for reading and writing files in NSDF Format.
- It supports reading and writing: Block, Segment, AnalogSignal, ChannelIndex, with all relationships and metadata.
- """
- is_readable = True
- is_writable = True
- supported_objects = [Block, Segment, AnalogSignal, ChannelIndex]
- readable_objects = [Block, Segment]
- writeable_objects = [Block, Segment]
- has_header = False
- is_streameable = False
- name = 'NSDF'
- extensions = ['h5']
- mode = 'file'
- def __init__(self, filename=None):
- """
- Initialise NSDFIO instance
- :param filename: Path to the file
- """
- if not HAVE_NSDF:
- raise Exception("Failed to import NSDF.")
- if filename is None:
- raise ValueError("Must provide an input file.")
- BaseIO.__init__(self)
- self.filename = filename
- self.dt_format = '%d/%m/%Y %H:%M:%S'
- self.modeltree_path = '/model/modeltree/neo/'
- def write_all_blocks(self, blocks):
- """
- Write list of blocks to the file
- :param blocks: List of blocks to be written
- """
- writer = self._init_writing()
- neo_model, blocks_model, segments_model = self._prepare_model_tree(writer)
- name_pattern = self._name_pattern(len(blocks))
- for i, block in enumerate(blocks):
- self.write_block(block, name_pattern.format(i), writer, blocks_model)
- def write_block(self, block = None, name='0', writer=None, parent=None):
- """
- Write a Block to the file
- :param block: Block to be written
- :param name: Name for block representation in NSDF model tree (optional)
- :param writer: NSDFWriter instance (optional)
- :param parent: NSDF ModelComponent which will be the parent of block NSDF representation (optional)
- """
- if not isinstance(block, Block):
- raise ValueError("Must provide a Block to write.")
- if writer is None:
- writer = self._init_writing()
- if parent is None:
- neo_model, parent, segments_model = self._prepare_model_tree(writer)
- block_model = nsdf.ModelComponent(name, uid=uuid1().hex, parent=parent)
- self._write_container_metadata(block, block_model)
- self._write_model_component(block_model, writer)
- self._write_block_children(block, block_model, writer)
- self._clean_nsdfio_annotations(block)
- def _write_block_children(self, block, block_model, writer):
- segments_model = nsdf.ModelComponent(name='segments', uid=uuid1().hex, parent=block_model)
- self._write_model_component(segments_model, writer)
- name_pattern = self._name_pattern(len(block.segments))
- for i, segment in enumerate(block.segments):
- self.write_segment(segment=segment, name=name_pattern.format(i),
- writer=writer, parent=segments_model)
- channel_indexes_model = nsdf.ModelComponent(name='channel_indexes', uid=uuid1().hex, parent=block_model)
- self._write_model_component(channel_indexes_model, writer)
- name_pattern = self._name_pattern(len(block.channel_indexes))
- for i, channelindex in enumerate(block.channel_indexes):
- self.write_channelindex(channelindex=channelindex, name=name_pattern.format(i),
- writer=writer, parent=channel_indexes_model)
- def write_segment(self, segment = None, name='0', writer=None, parent=None):
- """
- Write a Segment to the file
- :param segment: Segment to be written
- :param name: Name for segment representation in NSDF model tree (optional)
- :param writer: NSDFWriter instance (optional)
- :param parent: NSDF ModelComponent which will be the parent of segment NSDF representation (optional)
- """
- if not isinstance(segment, Segment):
- raise ValueError("Must provide a Segment to write.")
- if writer is None:
- writer = self._init_writing()
- single_segment = False
- if parent is None:
- neo_model, blocks_model, parent = self._prepare_model_tree(writer)
- single_segment = True
- model = nsdf.ModelComponent(name, uid=uuid1().hex, parent=parent)
- self._write_container_metadata(segment, model)
- self._write_model_component(model, writer)
- self._write_segment_children(model, segment, writer)
- if single_segment:
- self._clean_nsdfio_annotations(segment)
- def _write_segment_children(self, model, segment, writer):
- analogsignals_model = nsdf.ModelComponent(name='analogsignals', uid=uuid1().hex, parent=model)
- self._write_model_component(analogsignals_model, writer)
- name_pattern = self._name_pattern(len(segment.analogsignals))
- for i, signal in enumerate(segment.analogsignals):
- self.write_analogsignal(signal=signal, name=name_pattern.format(i),
- parent=analogsignals_model, writer=writer)
- def write_analogsignal(self, signal, name, writer, parent):
- """
- Write an AnalogSignal to the file
- :param signal: AnalogSignal to be written
- :param name: Name for signal representation in NSDF model tree
- :param writer: NSDFWriter instance
- :param parent: NSDF ModelComponent which will be the parent of signal NSDF representation
- """
- uid = uuid1().hex
- model = nsdf.ModelComponent(name, uid=uid, parent=parent)
- if signal.annotations.get('nsdfio_uid') is not None:
- model.attrs['reference_to'] = signal.annotations['nsdfio_uid']
- self._write_model_component(model, writer)
- return
- self._write_basic_metadata(model, signal)
- signal.annotations['nsdfio_uid'] = uid
- r_signal = np.swapaxes(signal, 0, 1)
- channels_model, channels, source_ds = self._create_signal_data_sources(model, r_signal, uid, writer)
- self._write_signal_data(model, channels, r_signal, signal, source_ds, writer)
- self._write_model_component(model, writer)
- self._write_model_component(channels_model, writer)
- for channel_model in channels:
- self._write_model_component(channel_model, writer)
- def write_channelindex(self, channelindex, name, writer, parent):
- """
- Write a ChannelIndex to the file
- :param channelindex: ChannelIndex to be written
- :param name: Name for channelindex representation in NSDF model tree
- :param writer: NSDFWriter instance
- :param parent: NSDF ModelComponent which will be the parent of channelindex NSDF representation
- """
- uid = uuid1().hex
- model = nsdf.ModelComponent(name, uid=uid, parent=parent)
- self._write_basic_metadata(model, channelindex)
- self._write_model_component(model, writer)
- self._write_channelindex_arrays(model, channelindex, writer)
- self._write_channelindex_children(channelindex, model, writer)
- def _write_channelindex_children(self, channelindex, model, writer):
- analogsignals_model = nsdf.ModelComponent(name='analogsignals', uid=uuid1().hex, parent=model)
- self._write_model_component(analogsignals_model, writer)
- name_pattern = self._name_pattern(len(channelindex.analogsignals))
- for i, signal in enumerate(channelindex.analogsignals):
- self.write_analogsignal(signal=signal, name=name_pattern.format(i),
- parent=analogsignals_model, writer=writer)
- def _init_writing(self):
- return nsdf.NSDFWriter(self.filename, mode='w')
- def _prepare_model_tree(self, writer):
- neo_model = nsdf.ModelComponent('neo', uid=uuid1().hex)
- self._write_model_component(neo_model, writer)
- blocks_model = nsdf.ModelComponent('blocks', uid=uuid1().hex, parent=neo_model)
- self._write_model_component(blocks_model, writer)
- segments_model = nsdf.ModelComponent('segments', uid=uuid1().hex, parent=neo_model)
- self._write_model_component(segments_model, writer)
- return neo_model, blocks_model, segments_model
- def _number_of_digits(self, n):
- return len(str(n))
- def _name_pattern(self, how_many_items):
- return '{{:0{}d}}'.format(self._number_of_digits(max(how_many_items - 1, 0)))
- def _clean_nsdfio_annotations(self, object):
- nsdfio_annotations = ('nsdfio_uid', )
- for key in nsdfio_annotations:
- object.annotations.pop(key, None)
- if hasattr(object, 'children'):
- for child in object.children:
- self._clean_nsdfio_annotations(child)
- def _write_model_component(self, model, writer):
- if model.parent is None:
- nsdf.add_model_component(model, writer.model['modeltree/'])
- else:
- nsdf.add_model_component(model, model.parent.hdfgroup)
- def _write_container_metadata(self, container, container_model):
- self._write_basic_metadata(container_model, container)
- self._write_datetime_attributes(container_model, container)
- self._write_index_attribute(container_model, container)
- def _write_basic_metadata(self, model, object):
- self._write_basic_attributes(model, object)
- self._write_annotations(model, object)
- def _write_basic_attributes(self, model, object):
- if object.name is not None:
- model.attrs['name'] = object.name
- if object.description is not None:
- model.attrs['description'] = object.description
- def _write_datetime_attributes(self, model, object):
- if object.rec_datetime is not None:
- model.attrs['rec_datetime'] = object.rec_datetime.strftime(self.dt_format)
- def _write_index_attribute(self, model, object):
- if object.index is not None:
- model.attrs['index'] = object.index
- def _write_annotations(self, model, object):
- if object.annotations is not None:
- model.attrs['annotations'] = pickle.dumps(object.annotations)
- def _write_signal_data(self, model, channels, r_signal, signal, source_ds, writer):
- dataobj = nsdf.UniformData('signal', unit=str(signal.units.dimensionality))
- dataobj.dtype = signal.dtype
- for i in range(len(channels)):
- dataobj.put_data(channels[i].uid, r_signal[i])
- dataobj.set_dt(float(signal.sampling_period.magnitude),
- str(signal.sampling_period.dimensionality))
- rescaled_tstart = signal.t_start.rescale(signal.sampling_period.dimensionality)
- writer.add_uniform_data(source_ds, dataobj,
- tstart=float(rescaled_tstart.magnitude))
- model.attrs['t_start_unit'] = str(signal.t_start.dimensionality)
- def _create_signal_data_sources(self, model, r_signal, uid, writer):
- channels = []
- channels_model = nsdf.ModelComponent(name='channels', uid=uuid1().hex, parent=model)
- name_pattern = '{{:0{}d}}'.format(self._number_of_digits(max(len(r_signal) - 1, 0)))
- for i in range(len(r_signal)):
- channels.append(nsdf.ModelComponent(name_pattern.format(i),
- uid=uuid1().hex,
- parent=channels_model))
- source_ds = writer.add_uniform_ds(uid, [channel.uid for channel in channels])
- return channels_model, channels, source_ds
- def _write_channelindex_arrays(self, model, channelindex, writer):
- group = model.hdfgroup
- self._write_array(group, 'index', channelindex.index)
- if channelindex.channel_names is not None:
- self._write_array(group, 'channel_names', channelindex.channel_names)
- if channelindex.channel_ids is not None:
- self._write_array(group, 'channel_ids', channelindex.channel_ids)
- if channelindex.coordinates is not None:
- self._write_array(group, 'coordinates', channelindex.coordinates)
- def _write_array(self, group, name, array):
- if isinstance(array, pq.Quantity):
- group.create_dataset(name, data=array.magnitude)
- group[name].attrs['dimensionality'] = str(array.dimensionality)
- else:
- group.create_dataset(name, data=array)
- def read_all_blocks(self, lazy=False, cascade=True):
- """
- Read all blocks from the file
- :param lazy: Enables lazy reading
- :param cascade: Read nested objects or not?
- :return: List of read blocks
- """
- reader = self._init_reading()
- blocks = []
- blocks_path = self.modeltree_path + 'blocks/'
- for block in reader.model[blocks_path].values():
- blocks.append(self.read_block(lazy, cascade, group=block, reader=reader))
- return blocks
- def read_block(self, lazy=False, cascade=True, group=None, reader=None):
- """
- Read a Block from the file
- :param lazy: Enables lazy reading
- :param cascade: Read nested objects or not?
- :param group: HDF5 Group representing the block in NSDF model tree (optional)
- :param reader: NSDFReader instance (optional)
- :return: Read block
- """
- block = Block()
- group, reader = self._select_first_container(group, reader, 'block')
- if group is None:
- return None
- attrs = group.attrs
- if cascade:
- self._read_block_children(lazy, block, group, reader)
- block.create_many_to_one_relationship()
- self._read_container_metadata(attrs, block)
- return block
- def _read_block_children(self, lazy, block, group, reader):
- for child in group['segments/'].values():
- block.segments.append(self.read_segment(lazy=lazy, group=child, reader=reader))
- for child in group['channel_indexes/'].values():
- block.channel_indexes.append(self.read_channelindex(lazy=lazy, group=child, reader=reader))
- def read_segment(self, lazy=False, cascade=True, group=None, reader=None):
- """
- Read a Segment from the file
- :param lazy: Enables lazy reading
- :param cascade: Read nested objects or not?
- :param group: HDF5 Group representing the segment in NSDF model tree (optional)
- :param reader: NSDFReader instance (optional)
- :return: Read segment
- """
- segment = Segment()
- group, reader = self._select_first_container(group, reader, 'segment')
- if group is None:
- return None
- attrs = group.attrs
- if cascade:
- self._read_segment_children(lazy, group, reader, segment)
- self._read_container_metadata(attrs, segment)
- return segment
- def _read_segment_children(self, lazy, group, reader, segment):
- for child in group['analogsignals/'].values():
- segment.analogsignals.append(self.read_analogsignal(lazy=lazy, group=child, reader=reader))
- def read_analogsignal(self, lazy=False, cascade=True, group=None, reader=None):
- """
- Read an AnalogSignal from the file (must be child of a Segment)
- :param lazy: Enables lazy reading
- :param cascade: Read nested objects or not?
- :param group: HDF5 Group representing the analogsignal in NSDF model tree
- :param reader: NSDFReader instance
- :return: Read AnalogSignal
- """
- attrs = group.attrs
- if attrs.get('reference_to') is not None:
- return self.objects_dict[attrs['reference_to']]
- uid = attrs['uid']
- data_group = reader.data['uniform/{}/signal'.format(uid)]
- t_start = self._read_analogsignal_t_start(attrs, data_group)
- signal = self._create_analogsignal(data_group, lazy, group, t_start, uid, reader)
- self._read_basic_metadata(attrs, signal)
- self.objects_dict[uid] = signal
- return signal
- def read_channelindex(self, lazy=False, cascade=True, group=None, reader=None):
- """
- Read a ChannelIndex from the file (must be child of a Block)
- :param lazy: Enables lazy reading
- :param cascade: Read nested objects or not?
- :param group: HDF5 Group representing the channelindex in NSDF model tree
- :param reader: NSDFReader instance
- :return: Read ChannelIndex
- """
- attrs = group.attrs
- channelindex = self._create_channelindex(group)
- if cascade:
- self._read_channelindex_children(lazy, group, reader, channelindex)
- self._read_basic_metadata(attrs, channelindex)
- return channelindex
- def _read_channelindex_children(self, lazy, group, reader, channelindex):
- for child in group['analogsignals/'].values():
- channelindex.analogsignals.append(self.read_analogsignal(lazy=lazy, group=child, reader=reader))
- def _init_reading(self):
- reader = nsdf.NSDFReader(self.filename)
- self.file_datetime = datetime.fromtimestamp(os.stat(self.filename).st_mtime)
- self.objects_dict = {}
- return reader
- def _select_first_container(self, group, reader, name):
- if reader is None:
- reader = self._init_reading()
- if group is None:
- path = self.modeltree_path + name + 's/'
- if len(reader.model[path].values()) > 0:
- group = reader.model[path].values()[0]
- return group, reader
- def _read_container_metadata(self, attrs, container):
- self._read_basic_metadata(attrs, container)
- self._read_datetime_attributes(attrs, container)
- self._read_index_attribute(attrs, container)
- def _read_basic_metadata(self, attrs, signal):
- self._read_basic_attributes(attrs, signal)
- self._read_annotations(attrs, signal)
- def _read_basic_attributes(self, attrs, object):
- if attrs.get('name') is not None:
- object.name = attrs['name']
- if attrs.get('description') is not None:
- object.description = attrs['description']
- object.file_origin = self.filename
- def _read_datetime_attributes(self, attrs, object):
- object.file_datetime = self.file_datetime
- if attrs.get('rec_datetime') is not None:
- object.rec_datetime = datetime.strptime(attrs['rec_datetime'], self.dt_format)
- def _read_annotations(self, attrs, object):
- if attrs.get('annotations') is not None:
- object.annotations = pickle.loads(attrs['annotations'])
- def _read_index_attribute(self, attrs, object):
- if attrs.get('index') is not None:
- object.index = attrs['index']
- def _create_analogsignal(self, data_group, lazy, group, t_start, uid, reader):
- if lazy:
- data_shape = data_group.shape
- data_shape = (data_shape[1], data_shape[0])
- signal = self._create_lazy_analogsignal(data_shape, data_group, uid, t_start)
- else:
- dataobj = reader.get_uniform_data(uid, 'signal')
- data = self._read_signal_data(dataobj, group)
- signal = self._create_normal_analogsignal(data, dataobj, uid, t_start)
- return signal
- def _read_analogsignal_t_start(self, attrs, data_group):
- t_start = float(data_group.attrs['tstart']) * pq.Quantity(1, data_group.attrs['tunit'])
- t_start = t_start.rescale(attrs['t_start_unit'])
- return t_start
- def _read_signal_data(self, dataobj, group):
- data = []
- for channel in group['channels/'].values():
- channel_uid = channel.attrs['uid']
- data += [dataobj.get_data(channel_uid)]
- return data
- def _create_normal_analogsignal(self, data, dataobj, uid, t_start):
- return AnalogSignal(np.swapaxes(data, 0, 1), dtype=dataobj.dtype, units=dataobj.unit,
- t_start=t_start, sampling_period=pq.Quantity(dataobj.dt, dataobj.tunit))
- def _create_lazy_analogsignal(self, shape, data, uid, t_start):
- attrs = data.attrs
- signal = AnalogSignal([], dtype=data.dtype, units=attrs['unit'],
- t_start=t_start, sampling_period=pq.Quantity(attrs['dt'], attrs['tunit']))
- signal.lazy_shape = shape
- return signal
- def _create_channelindex(self, group):
- return ChannelIndex(index=self._read_array(group, 'index'),
- channel_names=self._read_array(group, 'channel_names'),
- channel_ids=self._read_array(group, 'channel_ids'),
- coordinates=self._read_array(group, 'coordinates'))
- def _read_array(self, group, name):
- if group.__contains__(name) == False:
- return None
- array = group[name][:]
- if group[name].attrs.get('dimensionality') is not None:
- return pq.Quantity(array, group[name].attrs['dimensionality'])
- return array
|