Coverage for src/hdmf/data_utils.py: 93%
455 statements
« prev ^ index » next coverage.py v7.3.2, created at 2023-10-04 02:57 +0000
« prev ^ index » next coverage.py v7.3.2, created at 2023-10-04 02:57 +0000
1import copy
2import math
3from abc import ABCMeta, abstractmethod
4from collections.abc import Iterable
5from warnings import warn
6from typing import Tuple, Callable
7from itertools import product, chain
9import h5py
10import numpy as np
12from .utils import docval, getargs, popargs, docval_macro, get_data_shape
15def append_data(data, arg):
16 if isinstance(data, (list, DataIO)):
17 data.append(arg)
18 return data
19 elif type(data).__name__ == 'TermSetWrapper': # circular import
20 data.append(arg)
21 return data
22 elif isinstance(data, np.ndarray):
23 return np.append(data, np.expand_dims(arg, axis=0), axis=0)
24 elif isinstance(data, h5py.Dataset): 24 ↛ 31line 24 didn't jump to line 31, because the condition on line 24 was never false
25 shape = list(data.shape)
26 shape[0] += 1
27 data.resize(shape)
28 data[-1] = arg
29 return data
30 else:
31 msg = "Data cannot append to object of type '%s'" % type(data)
32 raise ValueError(msg)
35def extend_data(data, arg):
36 """Add all the elements of the iterable arg to the end of data.
38 :param data: The array to extend
39 :type data: list, DataIO, np.ndarray, h5py.Dataset
40 """
41 if isinstance(data, (list, DataIO)):
42 data.extend(arg)
43 return data
44 elif type(data).__name__ == 'TermSetWrapper':
45 data.extend(arg)
46 return data
47 elif isinstance(data, np.ndarray):
48 return np.vstack((data, arg))
49 elif isinstance(data, h5py.Dataset): 49 ↛ 56line 49 didn't jump to line 56, because the condition on line 49 was never false
50 shape = list(data.shape)
51 shape[0] += len(arg)
52 data.resize(shape)
53 data[-len(arg):] = arg
54 return data
55 else:
56 msg = "Data cannot extend object of type '%s'" % type(data)
57 raise ValueError(msg)
60@docval_macro('array_data')
61class AbstractDataChunkIterator(metaclass=ABCMeta):
62 """
63 Abstract iterator class used to iterate over DataChunks.
65 Derived classes must ensure that all abstract methods and abstract properties are implemented, in
66 particular, dtype, maxshape, __iter__, ___next__, recommended_chunk_shape, and recommended_data_shape.
68 Iterating over AbstractContainer objects is not yet supported.
69 """
71 @abstractmethod
72 def __iter__(self):
73 """Return the iterator object"""
74 raise NotImplementedError("__iter__ not implemented for derived class")
76 @abstractmethod
77 def __next__(self):
78 r"""
79 Return the next data chunk or raise a StopIteration exception if all chunks have been retrieved.
81 HINT: numpy.s\_ provides a convenient way to generate index tuples using standard array slicing. This
82 is often useful to define the DataChunk.selection of the current chunk
84 :returns: DataChunk object with the data and selection of the current chunk
85 :rtype: DataChunk
86 """
87 raise NotImplementedError("__next__ not implemented for derived class")
89 @abstractmethod
90 def recommended_chunk_shape(self):
91 """
92 Recommend the chunk shape for the data array.
94 :return: NumPy-style shape tuple describing the recommended shape for the chunks of the target
95 array or None. This may or may not be the same as the shape of the chunks returned in the
96 iteration process.
97 """
98 raise NotImplementedError("recommended_chunk_shape not implemented for derived class")
100 @abstractmethod
101 def recommended_data_shape(self):
102 """
103 Recommend the initial shape for the data array.
105 This is useful in particular to avoid repeated resized of the target array when reading from
106 this data iterator. This should typically be either the final size of the array or the known
107 minimal shape of the array.
109 :return: NumPy-style shape tuple indicating the recommended initial shape for the target array.
110 This may or may not be the final full shape of the array, i.e., the array is allowed
111 to grow. This should not be None.
112 """
113 raise NotImplementedError("recommended_data_shape not implemented for derived class")
115 @property
116 @abstractmethod
117 def dtype(self):
118 """
119 Define the data type of the array
121 :return: NumPy style dtype or otherwise compliant dtype string
122 """
123 raise NotImplementedError("dtype not implemented for derived class")
125 @property
126 @abstractmethod
127 def maxshape(self):
128 """
129 Property describing the maximum shape of the data array that is being iterated over
131 :return: NumPy-style shape tuple indicating the maximum dimensions up to which the dataset may be
132 resized. Axes with None are unlimited.
133 """
134 raise NotImplementedError("maxshape not implemented for derived class")
137class GenericDataChunkIterator(AbstractDataChunkIterator):
138 """DataChunkIterator that lets the user specify chunk and buffer shapes."""
140 __docval_init = (
141 dict(
142 name="buffer_gb",
143 type=(float, int),
144 doc=(
145 "If buffer_shape is not specified, it will be inferred as the smallest chunk "
146 "below the buffer_gb threshold."
147 "Defaults to 1GB."
148 ),
149 default=None,
150 ),
151 dict(
152 name="buffer_shape",
153 type=tuple,
154 doc="Manually defined shape of the buffer.",
155 default=None,
156 ),
157 dict(
158 name="chunk_mb",
159 type=(float, int),
160 doc=(
161 "If chunk_shape is not specified, it will be inferred as the smallest chunk "
162 "below the chunk_mb threshold.",
163 "Defaults to 10MB.",
164 ),
165 default=None,
166 ),
167 dict(
168 name="chunk_shape",
169 type=tuple,
170 doc="Manually defined shape of the chunks.",
171 default=None,
172 ),
173 dict(
174 name="display_progress",
175 type=bool,
176 doc="Display a progress bar with iteration rate and estimated completion time.",
177 default=False,
178 ),
179 dict(
180 name="progress_bar_options",
181 type=None,
182 doc="Dictionary of keyword arguments to be passed directly to tqdm.",
183 default=None,
184 ),
185 )
187 @docval(*__docval_init)
188 def __init__(self, **kwargs):
189 """
190 Break a dataset into buffers containing multiple chunks to be written into an HDF5 dataset.
192 Basic users should set the buffer_gb argument to as much free RAM space as can be safely allocated.
193 Advanced users are offered full control over the shape parameters for the buffer and the chunks; however,
194 the chunk shape must perfectly divide the buffer shape along each axis.
196 HDF5 recommends chunk size in the range of 2 to 16 MB for optimal cloud performance.
197 https://youtu.be/rcS5vt-mKok?t=621
198 """
199 buffer_gb, buffer_shape, chunk_mb, chunk_shape, self.display_progress, progress_bar_options = getargs(
200 "buffer_gb", "buffer_shape", "chunk_mb", "chunk_shape", "display_progress", "progress_bar_options", kwargs
201 )
202 self.progress_bar_options = progress_bar_options or dict()
204 if buffer_gb is None and buffer_shape is None:
205 buffer_gb = 1.0
206 if chunk_mb is None and chunk_shape is None:
207 chunk_mb = 10.0
208 assert (buffer_gb is not None) != (
209 buffer_shape is not None
210 ), "Only one of 'buffer_gb' or 'buffer_shape' can be specified!"
211 assert (chunk_mb is not None) != (
212 chunk_shape is not None
213 ), "Only one of 'chunk_mb' or 'chunk_shape' can be specified!"
215 self._dtype = self._get_dtype()
216 self._maxshape = tuple(int(x) for x in self._get_maxshape())
217 chunk_shape = tuple(int(x) for x in chunk_shape) if chunk_shape else chunk_shape
218 self.chunk_shape = chunk_shape or self._get_default_chunk_shape(chunk_mb=chunk_mb)
219 buffer_shape = tuple(int(x) for x in buffer_shape) if buffer_shape else buffer_shape
220 self.buffer_shape = buffer_shape or self._get_default_buffer_shape(buffer_gb=buffer_gb)
222 # Shape assertions
223 assert all(
224 buffer_axis > 0 for buffer_axis in self.buffer_shape
225 ), f"Some dimensions of buffer_shape ({self.buffer_shape}) are less than zero!"
226 assert all(
227 chunk_axis <= maxshape_axis for chunk_axis, maxshape_axis in zip(self.chunk_shape, self.maxshape)
228 ), f"Some dimensions of chunk_shape ({self.chunk_shape}) exceed the data dimensions ({self.maxshape})!"
229 assert all(
230 buffer_axis <= maxshape_axis for buffer_axis, maxshape_axis in zip(self.buffer_shape, self.maxshape)
231 ), f"Some dimensions of buffer_shape ({self.buffer_shape}) exceed the data dimensions ({self.maxshape})!"
232 assert all(
233 (chunk_axis <= buffer_axis for chunk_axis, buffer_axis in zip(self.chunk_shape, self.buffer_shape))
234 ), f"Some dimensions of chunk_shape ({self.chunk_shape}) exceed the buffer shape ({self.buffer_shape})!"
235 assert all(
236 buffer_axis % chunk_axis == 0
237 for chunk_axis, buffer_axis, maxshape_axis in zip(self.chunk_shape, self.buffer_shape, self.maxshape)
238 if buffer_axis != maxshape_axis
239 ), (
240 f"Some dimensions of chunk_shape ({self.chunk_shape}) do not "
241 f"evenly divide the buffer shape ({self.buffer_shape})!"
242 )
244 self.num_buffers = math.prod(
245 [
246 math.ceil(maxshape_axis / buffer_axis)
247 for buffer_axis, maxshape_axis in zip(self.buffer_shape, self.maxshape)
248 ],
249 )
250 self.buffer_selection_generator = (
251 tuple(
252 [
253 slice(lower_bound, upper_bound)
254 for lower_bound, upper_bound in zip(lower_bounds, upper_bounds)
255 ]
256 )
257 for lower_bounds, upper_bounds in zip(
258 product(
259 *[
260 range(0, max_shape_axis, buffer_shape_axis)
261 for max_shape_axis, buffer_shape_axis in zip(self.maxshape, self.buffer_shape)
262 ]
263 ),
264 product(
265 *[
266 chain(range(buffer_shape_axis, max_shape_axis, buffer_shape_axis), [max_shape_axis])
267 for max_shape_axis, buffer_shape_axis in zip(self.maxshape, self.buffer_shape)
268 ]
269 ),
270 )
271 )
273 if self.display_progress:
274 try:
275 from tqdm import tqdm
277 if "total" in self.progress_bar_options:
278 warn("Option 'total' in 'progress_bar_options' is not allowed to be over-written! Ignoring.")
279 self.progress_bar_options.pop("total")
281 self.progress_bar = tqdm(total=self.num_buffers, **self.progress_bar_options)
282 except ImportError:
283 warn(
284 "You must install tqdm to use the progress bar feature (pip install tqdm)! "
285 "Progress bar is disabled."
286 )
287 self.display_progress = False
289 @docval(
290 dict(
291 name="chunk_mb",
292 type=(float, int),
293 doc="Size of the HDF5 chunk in megabytes. Recommended to be less than 1MB.",
294 default=None,
295 )
296 )
297 def _get_default_chunk_shape(self, **kwargs) -> Tuple[int, ...]:
298 """
299 Select chunk shape with size in MB less than the threshold of chunk_mb.
301 Keeps the dimensional ratios of the original data.
302 """
303 chunk_mb = getargs("chunk_mb", kwargs)
304 assert chunk_mb > 0, f"chunk_mb ({chunk_mb}) must be greater than zero!"
306 n_dims = len(self.maxshape)
307 itemsize = self.dtype.itemsize
308 chunk_bytes = chunk_mb * 1e6
310 min_maxshape = min(self.maxshape)
311 v = tuple(math.floor(maxshape_axis / min_maxshape) for maxshape_axis in self.maxshape)
312 prod_v = math.prod(v)
313 while prod_v * itemsize > chunk_bytes and prod_v != 1:
314 non_unit_min_v = min(x for x in v if x != 1)
315 v = tuple(math.floor(x / non_unit_min_v) if x != 1 else x for x in v)
316 prod_v = math.prod(v)
317 k = math.floor((chunk_bytes / (prod_v * itemsize)) ** (1 / n_dims))
318 return tuple([min(k * x, self.maxshape[dim]) for dim, x in enumerate(v)])
320 @docval(
321 dict(
322 name="buffer_gb",
323 type=(float, int),
324 doc="Size of the data buffer in gigabytes. Recommended to be as much free RAM as safely available.",
325 default=None,
326 )
327 )
328 def _get_default_buffer_shape(self, **kwargs) -> Tuple[int, ...]:
329 """
330 Select buffer shape with size in GB less than the threshold of buffer_gb.
332 Keeps the dimensional ratios of the original data.
333 Assumes the chunk_shape has already been set.
334 """
335 buffer_gb = getargs("buffer_gb", kwargs)
336 assert buffer_gb > 0, f"buffer_gb ({buffer_gb}) must be greater than zero!"
337 assert all(chunk_axis > 0 for chunk_axis in self.chunk_shape), (
338 f"Some dimensions of chunk_shape ({self.chunk_shape}) are less than zero!"
339 )
341 k = math.floor(
342 (
343 buffer_gb * 1e9 / (math.prod(self.chunk_shape) * self.dtype.itemsize)
344 ) ** (1 / len(self.chunk_shape))
345 )
346 return tuple(
347 [
348 min(max(k * x, self.chunk_shape[j]), self.maxshape[j])
349 for j, x in enumerate(self.chunk_shape)
350 ]
351 )
353 def __iter__(self):
354 return self
356 def __next__(self):
357 """
358 Retrieve the next DataChunk object from the buffer, refilling the buffer if necessary.
360 :returns: DataChunk object with the data and selection of the current buffer.
361 :rtype: DataChunk
362 """
363 if self.display_progress:
364 self.progress_bar.update(n=1)
365 try:
366 buffer_selection = next(self.buffer_selection_generator)
367 return DataChunk(data=self._get_data(selection=buffer_selection), selection=buffer_selection)
368 except StopIteration:
369 if self.display_progress:
370 self.progress_bar.write("\n") # Allows text to be written to new lines after completion
371 raise StopIteration
373 def __reduce__(self) -> Tuple[Callable, Iterable]:
374 instance_constructor = self._from_dict
375 initialization_args = (self._to_dict(),)
376 return (instance_constructor, initialization_args)
378 @abstractmethod
379 def _get_data(self, selection: Tuple[slice]) -> np.ndarray:
380 """
381 Retrieve the data specified by the selection using minimal I/O.
383 The developer of a new implementation of the GenericDataChunkIterator must ensure the data is actually
384 loaded into memory, and not simply mapped.
386 :param selection: Tuple of slices, each indicating the selection indexed with respect to maxshape for that axis
387 :type selection: tuple of slices
389 :returns: Array of data specified by selection
390 :rtype: np.ndarray
391 Parameters
392 ----------
393 selection : tuple of slices
394 Each axis of tuple is a slice of the full shape from which to pull data into the buffer.
395 """
396 raise NotImplementedError("The data fetching method has not been built for this DataChunkIterator!")
398 @abstractmethod
399 def _get_maxshape(self) -> Tuple[int, ...]:
400 """Retrieve the maximum bounds of the data shape using minimal I/O."""
401 raise NotImplementedError("The setter for the maxshape property has not been built for this DataChunkIterator!")
403 @abstractmethod
404 def _get_dtype(self) -> np.dtype:
405 """Retrieve the dtype of the data using minimal I/O."""
406 raise NotImplementedError("The setter for the internal dtype has not been built for this DataChunkIterator!")
408 def _to_dict(self) -> dict:
409 """Optional method to add in child classes to enable pickling (required for multiprocessing)."""
410 raise NotImplementedError(
411 "The `._to_dict()` method for pickling has not been defined for this DataChunkIterator!"
412 )
414 @staticmethod
415 def _from_dict(self) -> Callable:
416 """Optional method to add in child classes to enable pickling (required for multiprocessing)."""
417 raise NotImplementedError(
418 "The `._from_dict()` method for pickling has not been defined for this DataChunkIterator!"
419 )
421 def recommended_chunk_shape(self) -> Tuple[int, ...]:
422 return self.chunk_shape
424 def recommended_data_shape(self) -> Tuple[int, ...]:
425 return self.maxshape
427 @property
428 def maxshape(self) -> Tuple[int, ...]:
429 return self._maxshape
430 @property
431 def dtype(self) -> np.dtype:
432 return self._dtype
435class DataChunkIterator(AbstractDataChunkIterator):
436 """
437 Custom iterator class used to iterate over chunks of data.
439 This default implementation of AbstractDataChunkIterator accepts any iterable and assumes that we iterate over
440 a single dimension of the data array (default: the first dimension). DataChunkIterator supports buffered read,
441 i.e., multiple values from the input iterator can be combined to a single chunk. This is
442 useful for buffered I/O operations, e.g., to improve performance by accumulating data
443 in memory and writing larger blocks at once.
445 .. note::
447 DataChunkIterator assumes that the iterator that it wraps returns one element along the
448 iteration dimension at a time. I.e., the iterator is expected to return chunks that are
449 one dimension lower than the array itself. For example, when iterating over the first dimension
450 of a dataset with shape (1000, 10, 10), then the iterator would return 1000 chunks of
451 shape (10, 10) one-chunk-at-a-time. If this pattern does not match your use-case then
452 using :py:class:`~hdmf.data_utils.GenericDataChunkIterator` or
453 :py:class:`~hdmf.data_utils.AbstractDataChunkIterator` may be more appropriate.
454 """
456 __docval_init = (
457 {'name': 'data', 'type': None, 'doc': 'The data object used for iteration', 'default': None},
458 {'name': 'maxshape', 'type': tuple,
459 'doc': 'The maximum shape of the full data array. Use None to indicate unlimited dimensions',
460 'default': None},
461 {'name': 'dtype', 'type': np.dtype, 'doc': 'The Numpy data type for the array', 'default': None},
462 {'name': 'buffer_size', 'type': int, 'doc': 'Number of values to be buffered in a chunk', 'default': 1},
463 {'name': 'iter_axis', 'type': int, 'doc': 'The dimension to iterate over', 'default': 0}
464 )
466 @docval(*__docval_init)
467 def __init__(self, **kwargs):
468 """Initialize the DataChunkIterator.
469 If 'data' is an iterator and 'dtype' is not specified, then next is called on the iterator in order to determine
470 the dtype of the data.
471 """
472 # Get the user parameters
473 self.data, self.__maxshape, self.__dtype, self.buffer_size, self.iter_axis = getargs('data',
474 'maxshape',
475 'dtype',
476 'buffer_size',
477 'iter_axis',
478 kwargs)
479 self.chunk_index = 0
480 # Create an iterator for the data if possible
481 if isinstance(self.data, Iterable):
482 if self.iter_axis != 0 and isinstance(self.data, (list, tuple)):
483 warn('Iterating over an axis other than the first dimension of list or tuple data '
484 'involves converting the data object to a numpy ndarray, which may incur a computational '
485 'cost.')
486 self.data = np.asarray(self.data)
487 if isinstance(self.data, np.ndarray):
488 # iterate over the given axis by adding a new view on data (iter only works on the first dim)
489 self.__data_iter = iter(np.moveaxis(self.data, self.iter_axis, 0))
490 else:
491 self.__data_iter = iter(self.data)
492 else:
493 self.__data_iter = None
494 self.__next_chunk = DataChunk(None, None)
495 self.__next_chunk_start = 0
496 self.__first_chunk_shape = None
497 # Determine the shape of the data if possible
498 if self.__maxshape is None:
499 # If the self.data object identifies its shape, then use it
500 if hasattr(self.data, "shape"):
501 self.__maxshape = self.data.shape
502 # Avoid the special case of scalar values by making them into a 1D numpy array
503 if len(self.__maxshape) == 0: 503 ↛ 504line 503 didn't jump to line 504, because the condition on line 503 was never true
504 self.data = np.asarray([self.data, ])
505 self.__maxshape = self.data.shape
506 self.__data_iter = iter(self.data)
507 # Try to get an accurate idea of __maxshape for other Python data structures if possible.
508 # Don't just call get_data_shape for a generator as that would potentially trigger loading of all the data
509 elif isinstance(self.data, list) or isinstance(self.data, tuple):
510 self.__maxshape = get_data_shape(self.data, strict_no_data_load=True)
512 # If we have a data iterator and do not know the dtype, then read the first chunk
513 if self.__data_iter is not None and self.__dtype is None:
514 self._read_next_chunk()
516 # Determine the type of the data if possible
517 if self.__next_chunk.data is not None:
518 self.__dtype = self.__next_chunk.data.dtype
519 self.__first_chunk_shape = get_data_shape(self.__next_chunk.data)
521 # This should be done as a last resort only
522 if self.__first_chunk_shape is None and self.__maxshape is not None:
523 self.__first_chunk_shape = tuple(1 if i is None else i for i in self.__maxshape)
525 if self.__dtype is None:
526 raise Exception('Data type could not be determined. Please specify dtype in DataChunkIterator init.')
528 @classmethod
529 @docval(*__docval_init)
530 def from_iterable(cls, **kwargs):
531 return cls(**kwargs)
533 def __iter__(self):
534 """Return the iterator object"""
535 return self
537 def _read_next_chunk(self):
538 """Read a single chunk from self.__data_iter and store the results in self.__next_chunk
540 :returns: self.__next_chunk, i.e., the DataChunk object describing the next chunk
541 """
542 from h5py import Dataset as H5Dataset
543 if isinstance(self.data, H5Dataset):
544 start_index = self.chunk_index * self.buffer_size
545 stop_index = start_index + self.buffer_size
546 iter_data_bounds = self.data.shape[self.iter_axis]
547 if start_index >= iter_data_bounds:
548 self.__next_chunk = DataChunk(None, None)
549 else:
550 if stop_index > iter_data_bounds:
551 stop_index = iter_data_bounds
553 selection = [slice(None)] * len(self.maxshape)
554 selection[self.iter_axis] = slice(start_index, stop_index)
555 selection = tuple(selection)
556 self.__next_chunk.data = self.data[selection]
557 self.__next_chunk.selection = selection
558 elif self.__data_iter is not None:
559 # the pieces in the buffer - first dimension consists of individual calls to next
560 iter_pieces = []
561 # offset of where data begins - shift the selection of where to place this chunk by this much
562 curr_chunk_offset = 0
563 read_next_empty = False
564 while len(iter_pieces) < self.buffer_size:
565 try:
566 dat = next(self.__data_iter)
567 if dat is None and len(iter_pieces) == 0:
568 # Skip forward in our chunk until we find data
569 curr_chunk_offset += 1
570 elif dat is None and len(iter_pieces) > 0:
571 # Stop iteration if we hit empty data while constructing our block
572 # Buffer may not be full.
573 read_next_empty = True
574 break
575 else:
576 # Add pieces of data to our buffer
577 iter_pieces.append(np.asarray(dat))
578 except StopIteration:
579 break
581 if len(iter_pieces) == 0:
582 self.__next_chunk = DataChunk(None, None) # signal end of iteration
583 else:
584 # concatenate all the pieces into the chunk along the iteration axis
585 piece_shape = list(get_data_shape(iter_pieces[0]))
586 piece_shape.insert(self.iter_axis, 1) # insert the missing axis
587 next_chunk_shape = piece_shape.copy()
588 next_chunk_shape[self.iter_axis] *= len(iter_pieces)
589 next_chunk_size = next_chunk_shape[self.iter_axis]
591 # use the piece dtype because the actual dtype may not have been determined yet
592 # NOTE: this could be problematic if a generator returns e.g. floats first and ints later
593 self.__next_chunk.data = np.empty(next_chunk_shape, dtype=iter_pieces[0].dtype)
594 self.__next_chunk.data = np.stack(iter_pieces, axis=self.iter_axis)
596 selection = [slice(None)] * len(self.maxshape)
597 selection[self.iter_axis] = slice(self.__next_chunk_start + curr_chunk_offset,
598 self.__next_chunk_start + curr_chunk_offset + next_chunk_size)
599 self.__next_chunk.selection = tuple(selection)
601 # next chunk should start at self.__next_chunk.selection[self.iter_axis].stop
602 # but if this chunk stopped because of reading empty data, then this should be adjusted by 1
603 self.__next_chunk_start = self.__next_chunk.selection[self.iter_axis].stop
604 if read_next_empty:
605 self.__next_chunk_start += 1
606 else:
607 self.__next_chunk = DataChunk(None, None)
609 self.chunk_index += 1
610 return self.__next_chunk
612 def __next__(self):
613 """
614 Return the next data chunk or raise a StopIteration exception if all chunks have been retrieved.
616 .. tip::
618 :py:attr:`numpy.s_` provides a convenient way to generate index tuples using standard array slicing. This
619 is often useful to define the DataChunk.selection of the current chunk
621 :returns: DataChunk object with the data and selection of the current chunk
622 :rtype: DataChunk
624 """
625 # If we have not already read the next chunk, then read it now
626 if self.__next_chunk.data is None:
627 self._read_next_chunk()
628 # If we do not have any next chunk
629 if self.__next_chunk.data is None:
630 raise StopIteration
631 # If this is the first time we see a chunk then remember the size of the first chunk
632 if self.__first_chunk_shape is None: 632 ↛ 633line 632 didn't jump to line 633, because the condition on line 632 was never true
633 self.__first_chunk_shape = self.__next_chunk.data.shape
634 # Keep the next chunk we need to return
635 curr_chunk = DataChunk(self.__next_chunk.data,
636 self.__next_chunk.selection)
637 # Remove the data for the next chunk from our list since we are returning it here.
638 # This is to allow the GarbageCollector to remove the data when it goes out of scope and avoid
639 # having 2 full chunks in memory if not necessary
640 self.__next_chunk.data = None
641 # Return the current next chunk
642 return curr_chunk
644 next = __next__
646 @docval(returns='Tuple with the recommended chunk shape or None if no particular shape is recommended.')
647 def recommended_chunk_shape(self):
648 """Recommend a chunk shape.
650 To optimize iterative write the chunk should be aligned with the common shape of chunks returned by __next__
651 or if those chunks are too large, then a well-aligned subset of those chunks. This may also be
652 any other value in case one wants to recommend chunk shapes to optimize read rather
653 than write. The default implementation returns None, indicating no preferential chunking option."""
654 return None
656 @docval(returns='Recommended initial shape for the full data. This should be the shape of the full dataset' +
657 'if known beforehand or alternatively the minimum shape of the dataset. Return None if no ' +
658 'recommendation is available')
659 def recommended_data_shape(self):
660 """Recommend an initial shape of the data. This is useful when progressively writing data and
661 we want to recommend an initial size for the dataset"""
662 if self.maxshape is not None:
663 if np.all([i is not None for i in self.maxshape]):
664 return self.maxshape
665 return self.__first_chunk_shape
667 @property
668 def maxshape(self):
669 """
670 Get a shape tuple describing the maximum shape of the array described by this DataChunkIterator.
672 .. note::
674 If an iterator is provided and no data has been read yet, then the first chunk will be read
675 (i.e., next will be called on the iterator) in order to determine the maxshape. The iterator
676 is expected to return single chunks along the iterator dimension, this means that maxshape will
677 add an additional dimension along the iteration dimension. E.g., if we iterate over
678 the first dimension and the iterator returns chunks of shape (10, 10), then the maxshape would
679 be (None, 10, 10) or (len(self.data), 10, 10), depending on whether size of the
680 iteration dimension is known.
682 :return: Shape tuple. None is used for dimensions where the maximum shape is not known or unlimited.
683 """
684 if self.__maxshape is None:
685 # If no data has been read from the iterator yet, read the first chunk and use it to determine the maxshape
686 if self.__data_iter is not None and self.__next_chunk.data is None: 686 ↛ 687line 686 didn't jump to line 687, because the condition on line 686 was never true
687 self._read_next_chunk()
689 # Determine maxshape from self.__next_chunk
690 if self.__next_chunk.data is None:
691 return None
692 data_shape = get_data_shape(self.__next_chunk.data)
693 self.__maxshape = list(data_shape)
694 try:
695 # Size of self.__next_chunk.data along self.iter_axis is not accurate for maxshape because it is just a
696 # chunk. So try to set maxshape along the dimension self.iter_axis based on the shape of self.data if
697 # possible. Otherwise, use None to represent an unlimited size
698 if hasattr(self.data, '__len__') and self.iter_axis == 0:
699 # special case of 1-D array
700 self.__maxshape[0] = len(self.data)
701 else:
702 self.__maxshape[self.iter_axis] = self.data.shape[self.iter_axis]
703 except AttributeError: # from self.data.shape
704 self.__maxshape[self.iter_axis] = None
705 self.__maxshape = tuple(self.__maxshape)
707 return self.__maxshape
709 @property
710 def dtype(self):
711 """
712 Get the value data type
714 :return: np.dtype object describing the datatype
715 """
716 return self.__dtype
719class DataChunk:
720 """
721 Class used to describe a data chunk. Used in DataChunkIterator.
722 """
724 @docval({'name': 'data', 'type': np.ndarray,
725 'doc': 'Numpy array with the data value(s) of the chunk', 'default': None},
726 {'name': 'selection', 'type': None,
727 'doc': 'Numpy index tuple describing the location of the chunk', 'default': None})
728 def __init__(self, **kwargs):
729 self.data, self.selection = getargs('data', 'selection', kwargs)
731 def __len__(self):
732 """Get the number of values in the data chunk"""
733 if self.data is not None:
734 return len(self.data)
735 else:
736 return 0
738 def __getattr__(self, attr):
739 """Delegate retrieval of attributes to the data in self.data"""
740 return getattr(self.data, attr)
742 def __copy__(self):
743 newobj = DataChunk(data=self.data,
744 selection=self.selection)
745 return newobj
747 def __deepcopy__(self, memo):
748 result = DataChunk(data=copy.deepcopy(self.data),
749 selection=copy.deepcopy(self.selection))
750 memo[id(self)] = result
751 return result
753 def astype(self, dtype):
754 """Get a new DataChunk with the self.data converted to the given type"""
755 return DataChunk(data=self.data.astype(dtype),
756 selection=self.selection)
758 @property
759 def dtype(self):
760 """
761 Data type of the values in the chunk
763 :returns: np.dtype of the values in the DataChunk
764 """
765 return self.data.dtype
767 def get_min_bounds(self):
768 """
769 Helper function to compute the minimum dataset size required to fit the selection of this chunk.
771 :raises TypeError: If the the selection is not a single int, slice, or tuple of slices.
773 :return: Tuple with the minimum shape required to store the selection
774 """
775 if isinstance(self.selection, tuple): 775 ↛ 778line 775 didn't jump to line 778, because the condition on line 775 was never false
776 # Determine the minimum array dimensions to fit the chunk selection
777 max_bounds = tuple([x.stop or 0 if isinstance(x, slice) else x+1 for x in self.selection])
778 elif isinstance(self.selection, int):
779 max_bounds = (self.selection+1, )
780 elif isinstance(self.selection, slice):
781 max_bounds = (self.selection.stop or 0, )
782 else:
783 # Note: Technically any numpy index tuple would be allowed, but h5py is not as general and this case
784 # only implements the selections supported by h5py. We could add more cases to support a
785 # broader range of valid numpy selection types
786 msg = ("Chunk selection %s must be a single int, single slice, or tuple of slices "
787 "and/or integers") % str(self.selection)
788 raise TypeError(msg)
789 return max_bounds
792def assertEqualShape(data1,
793 data2,
794 axes1=None,
795 axes2=None,
796 name1=None,
797 name2=None,
798 ignore_undetermined=True):
799 """
800 Ensure that the shape of data1 and data2 match along the given dimensions
802 :param data1: The first input array
803 :type data1: List, Tuple, np.ndarray, DataChunkIterator etc.
804 :param data2: The second input array
805 :type data2: List, Tuple, np.ndarray, DataChunkIterator etc.
806 :param name1: Optional string with the name of data1
807 :param name2: Optional string with the name of data2
808 :param axes1: The dimensions of data1 that should be matched to the dimensions of data2. Set to None to
809 compare all axes in order.
810 :type axes1: int, Tuple of ints, List of ints, or None
811 :param axes2: The dimensions of data2 that should be matched to the dimensions of data1. Must have
812 the same length as axes1. Set to None to compare all axes in order.
813 :type axes1: int, Tuple of ints, List of ints, or None
814 :param ignore_undetermined: Boolean indicating whether non-matching unlimited dimensions should be ignored,
815 i.e., if two dimension don't match because we can't determine the shape of either one, then
816 should we ignore that case or treat it as no match
818 :return: Bool indicating whether the check passed and a string with a message about the matching process
819 """
820 # Create the base return object
821 response = ShapeValidatorResult()
822 # Determine the shape of the datasets
823 response.shape1 = get_data_shape(data1)
824 response.shape2 = get_data_shape(data2)
825 # Determine the number of dimensions of the datasets
826 num_dims_1 = len(response.shape1) if response.shape1 is not None else None
827 num_dims_2 = len(response.shape2) if response.shape2 is not None else None
828 # Determine the string names of the datasets
829 n1 = name1 if name1 is not None else ("data1 at " + str(hex(id(data1))))
830 n2 = name2 if name2 is not None else ("data2 at " + str(hex(id(data2))))
831 # Determine the axes we should compare
832 response.axes1 = list(range(num_dims_1)) if axes1 is None else ([axes1] if isinstance(axes1, int) else axes1)
833 response.axes2 = list(range(num_dims_2)) if axes2 is None else ([axes2] if isinstance(axes2, int) else axes2)
834 # Validate the array shape
835 # 1) Check the number of dimensions of the arrays
836 if (response.axes1 is None and response.axes2 is None) and num_dims_1 != num_dims_2: 836 ↛ 837line 836 didn't jump to line 837, because the condition on line 836 was never true
837 response.result = False
838 response.error = 'NUM_DIMS_ERROR'
839 response.message = response.SHAPE_ERROR[response.error]
840 response.message += " %s is %sD and %s is %sD" % (n1, num_dims_1, n2, num_dims_2)
841 # 2) Check that we have the same number of dimensions to compare on both arrays
842 elif len(response.axes1) != len(response.axes2):
843 response.result = False
844 response.error = 'NUM_AXES_ERROR'
845 response.message = response.SHAPE_ERROR[response.error]
846 response.message += " Cannot compare axes %s with %s" % (str(response.axes1), str(response.axes2))
847 # 3) Check that the datasets have sufficient number of dimensions
848 elif np.max(response.axes1) >= num_dims_1 or np.max(response.axes2) >= num_dims_2:
849 response.result = False
850 response.error = 'AXIS_OUT_OF_BOUNDS'
851 response.message = response.SHAPE_ERROR[response.error]
852 if np.max(response.axes1) >= num_dims_1:
853 response.message += "Insufficient number of dimensions for %s -- Expected %i found %i" % \
854 (n1, np.max(response.axes1) + 1, num_dims_1)
855 elif np.max(response.axes2) >= num_dims_2: 855 ↛ 891line 855 didn't jump to line 891, because the condition on line 855 was never false
856 response.message += "Insufficient number of dimensions for %s -- Expected %i found %i" % \
857 (n2, np.max(response.axes2) + 1, num_dims_2)
858 # 4) Compare the length of the dimensions we should validate
859 else:
860 unmatched = []
861 ignored = []
862 for ax in zip(response.axes1, response.axes2):
863 if response.shape1[ax[0]] != response.shape2[ax[1]]:
864 if ignore_undetermined and (response.shape1[ax[0]] is None or response.shape2[ax[1]] is None):
865 ignored.append(ax)
866 else:
867 unmatched.append(ax)
868 response.unmatched = unmatched
869 response.ignored = ignored
871 # Check if everything checked out
872 if len(response.unmatched) == 0:
873 response.result = True
874 response.error = None
875 response.message = response.SHAPE_ERROR[response.error]
876 if len(response.ignored) > 0:
877 response.message += " Ignored undetermined axes %s" % str(response.ignored)
878 else:
879 response.result = False
880 response.error = 'AXIS_LEN_ERROR'
881 response.message = response.SHAPE_ERROR[response.error]
882 response.message += "Axes %s with size %s of %s did not match dimensions %s with sizes %s of %s." % \
883 (str([un[0] for un in response.unmatched]),
884 str([response.shape1[un[0]] for un in response.unmatched]),
885 n1,
886 str([un[1] for un in response.unmatched]),
887 str([response.shape2[un[1]] for un in response.unmatched]),
888 n2)
889 if len(response.ignored) > 0: 889 ↛ 890line 889 didn't jump to line 890, because the condition on line 889 was never true
890 response.message += " Ignored undetermined axes %s" % str(response.ignored)
891 return response
894class ShapeValidatorResult:
895 """Class for storing results from validating the shape of multi-dimensional arrays.
897 This class is used to store results generated by ShapeValidator
899 :ivar result: Boolean indicating whether results matched or not
900 :type result: bool
901 :ivar message: Message indicating the result of the matching procedure
902 :type messaage: str, None
903 """
904 SHAPE_ERROR = {None: 'All required axes matched',
905 'NUM_DIMS_ERROR': 'Unequal number of dimensions.',
906 'NUM_AXES_ERROR': "Unequal number of axes for comparison.",
907 'AXIS_OUT_OF_BOUNDS': "Axis index for comparison out of bounds.",
908 'AXIS_LEN_ERROR': "Unequal length of axes."}
909 """
910 Dict where the Keys are the type of errors that may have occurred during shape comparison and the
911 values are strings with default error messages for the type.
912 """
914 @docval({'name': 'result', 'type': bool, 'doc': 'Result of the shape validation', 'default': False},
915 {'name': 'message', 'type': str,
916 'doc': 'Message describing the result of the shape validation', 'default': None},
917 {'name': 'ignored', 'type': tuple,
918 'doc': 'Axes that have been ignored in the validaton process', 'default': tuple(), 'shape': (None,)},
919 {'name': 'unmatched', 'type': tuple,
920 'doc': 'List of axes that did not match during shape validation', 'default': tuple(), 'shape': (None,)},
921 {'name': 'error', 'type': str, 'doc': 'Error that may have occurred. One of ERROR_TYPE', 'default': None},
922 {'name': 'shape1', 'type': tuple,
923 'doc': 'Shape of the first array for comparison', 'default': tuple(), 'shape': (None,)},
924 {'name': 'shape2', 'type': tuple,
925 'doc': 'Shape of the second array for comparison', 'default': tuple(), 'shape': (None,)},
926 {'name': 'axes1', 'type': tuple,
927 'doc': 'Axes for the first array that should match', 'default': tuple(), 'shape': (None,)},
928 {'name': 'axes2', 'type': tuple,
929 'doc': 'Axes for the second array that should match', 'default': tuple(), 'shape': (None,)},
930 )
931 def __init__(self, **kwargs):
932 self.result, self.message, self.ignored, self.unmatched, \
933 self.error, self.shape1, self.shape2, self.axes1, self.axes2 = getargs(
934 'result', 'message', 'ignored', 'unmatched', 'error', 'shape1', 'shape2', 'axes1', 'axes2', kwargs)
936 def __setattr__(self, key, value):
937 """
938 Overwrite to ensure that, e.g., error_message is not set to an illegal value.
939 """
940 if key == 'error':
941 if value not in self.SHAPE_ERROR.keys():
942 raise ValueError("Illegal error type. Error must be one of ShapeValidatorResult.SHAPE_ERROR: %s"
943 % str(self.SHAPE_ERROR))
944 else:
945 super().__setattr__(key, value)
946 elif key in ['shape1', 'shape2', 'axes1', 'axes2', 'ignored', 'unmatched']: # Make sure we sore tuples
947 super().__setattr__(key, tuple(value))
948 else:
949 super().__setattr__(key, value)
951 def __getattr__(self, item):
952 """
953 Overwrite to allow dynamic retrieval of the default message
954 """
955 if item == 'default_message': 955 ↛ 957line 955 didn't jump to line 957, because the condition on line 955 was never false
956 return self.SHAPE_ERROR[self.error]
957 return self.__getattribute__(item)
960@docval_macro('data')
961class DataIO:
962 """
963 Base class for wrapping data arrays for I/O. Derived classes of DataIO are typically
964 used to pass dataset-specific I/O parameters to the particular HDMFIO backend.
965 """
967 @docval({'name': 'data',
968 'type': 'array_data',
969 'doc': 'the data to be written',
970 'default': None},
971 {'name': 'dtype',
972 'type': (type, np.dtype),
973 'doc': 'the data type of the dataset. Not used if data is specified.',
974 'default': None},
975 {'name': 'shape',
976 'type': tuple,
977 'doc': 'the shape of the dataset. Not used if data is specified.',
978 'default': None})
979 def __init__(self, **kwargs):
980 data, dtype, shape = popargs('data', 'dtype', 'shape', kwargs)
981 if data is None:
982 if (dtype is None) ^ (shape is None):
983 raise ValueError("Must specify 'dtype' and 'shape' if not specifying 'data'")
984 else:
985 if dtype is not None:
986 warn("Argument 'dtype' is ignored when 'data' is specified")
987 dtype = None
988 if shape is not None:
989 warn("Argument 'shape' is ignored when 'data' is specified")
990 shape = None
991 self.__data = data
992 self.__dtype = dtype
993 self.__shape = shape
995 def get_io_params(self):
996 """
997 Returns a dict with the I/O parameters specified in this DataIO.
998 """
999 return dict()
1001 @property
1002 def data(self):
1003 """Get the wrapped data object"""
1004 return self.__data
1006 @data.setter
1007 def data(self, val):
1008 """Set the wrapped data object"""
1009 if self.__data is not None:
1010 raise ValueError("cannot overwrite 'data' on DataIO")
1011 if not (self.__dtype is None and self.__shape is None):
1012 raise ValueError("Setting data when dtype and shape are not None is not supported")
1013 self.__data = val
1015 @property
1016 def dtype(self):
1017 """Get the wrapped data object"""
1018 return self.__dtype or self.__getattr__("dtype")
1020 @property
1021 def shape(self):
1022 """Get the wrapped data object"""
1023 return self.__shape or self.__getattr__("shape")
1025 def __copy__(self):
1026 """
1027 Define a custom copy method for shallow copy..
1029 This is needed due to delegation of __getattr__ to the data to
1030 ensure proper copy.
1032 :return: Shallow copy of self, ie., a new instance of DataIO wrapping the same self.data object
1033 """
1034 newobj = DataIO(data=self.data)
1035 return newobj
1037 def append(self, arg):
1038 self.__data = append_data(self.__data, arg)
1040 def extend(self, arg):
1041 self.__data = extend_data(self.__data, arg)
1043 def __deepcopy__(self, memo):
1044 """
1045 Define a custom copy method for deep copy.
1047 This is needed due to delegation of __getattr__ to the data to
1048 ensure proper copy.
1050 :param memo:
1051 :return: Deep copy of self, i.e., a new instance of DataIO wrapping a deepcopy of the
1052 self.data object.
1053 """
1054 result = DataIO(data=copy.deepcopy(self.__data))
1055 memo[id(self)] = result
1056 return result
1058 def __len__(self):
1059 """Number of values in self.data"""
1060 if self.__shape is not None:
1061 return self.__shape[0]
1062 if not self.valid:
1063 raise InvalidDataIOError("Cannot get length of data. Data is not valid.")
1064 return len(self.data)
1066 def __bool__(self):
1067 if self.valid: 1067 ↛ 1071line 1067 didn't jump to line 1071, because the condition on line 1067 was never false
1068 if isinstance(self.data, AbstractDataChunkIterator): 1068 ↛ 1069line 1068 didn't jump to line 1069, because the condition on line 1068 was never true
1069 return True
1070 return len(self) > 0
1071 return False
1073 def __getattr__(self, attr):
1074 """Delegate attribute lookup to data object"""
1075 if attr == '__array_struct__' and not self.valid:
1076 # np.array() checks __array__ or __array_struct__ attribute dep. on numpy version
1077 raise InvalidDataIOError("Cannot convert data to array. Data is not valid.")
1078 if not self.valid:
1079 raise InvalidDataIOError("Cannot get attribute '%s' of data. Data is not valid." % attr)
1080 return getattr(self.data, attr)
1082 def __getitem__(self, item):
1083 """Delegate slicing to the data object"""
1084 if not self.valid: 1084 ↛ 1085line 1084 didn't jump to line 1085, because the condition on line 1084 was never true
1085 raise InvalidDataIOError("Cannot get item from data. Data is not valid.")
1086 return self.data[item]
1088 def __array__(self):
1089 """
1090 Support conversion of DataIO.data to a numpy array. This function is
1091 provided to improve transparent interoperability of DataIO with numpy.
1093 :return: An array instance of self.data
1094 """
1095 if not self.valid: 1095 ↛ 1096line 1095 didn't jump to line 1096, because the condition on line 1095 was never true
1096 raise InvalidDataIOError("Cannot convert data to array. Data is not valid.")
1097 if hasattr(self.data, '__array__'):
1098 return self.data.__array__()
1099 elif isinstance(self.data, DataChunkIterator):
1100 raise NotImplementedError("Conversion of DataChunkIterator to array not supported")
1101 else:
1102 # NOTE this may result in a copy of the array
1103 return np.asarray(self.data)
1105 def __next__(self):
1106 """Delegate iteration interface to data object"""
1107 if not self.valid:
1108 raise InvalidDataIOError("Cannot iterate on data. Data is not valid.")
1109 return self.data.__next__()
1111 def __iter__(self):
1112 """Delegate iteration interface to the data object"""
1113 if not self.valid:
1114 raise InvalidDataIOError("Cannot iterate on data. Data is not valid.")
1115 return self.data.__iter__()
1117 @property
1118 def valid(self):
1119 """bool indicating if the data object is valid"""
1120 return self.data is not None
1123class InvalidDataIOError(Exception):
1124 pass