field.py 29.9 KB
Newer Older
1
from enum import Enum
2
from itertools import chain
Martin Bauer's avatar
Martin Bauer committed
3
from typing import Tuple, Sequence, Optional, List
4
5
6
7
import numpy as np
import sympy as sp
from sympy.core.cache import cacheit
from sympy.tensor import IndexedBase
8
from pystencils.alignedarray import aligned_empty
Martin Bauer's avatar
Martin Bauer committed
9
from pystencils.data_types import TypedSymbol, create_type, create_composite_type_from_string, StructType
Martin Bauer's avatar
Martin Bauer committed
10
from pystencils.sympyextensions import is_integer_sequence
11
12


13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
def fields(description=None, index_dimensions=0, layout=None, **kwargs):
    """Creates pystencils fields from a string description.

    Examples:
        Create a 2D scalar and vector field
        >>> s, v = fields("s, v(2): double[2D]")
        >>> assert s.spatial_dimensions == 2 and s.index_dimensions == 0
        >>> assert v.spatial_dimensions == 2 and v.index_dimensions == 1 and v.index_shape == (2,)

        Create an integer field of shape (10, 20)
        >>> f = fields("f : int32[10, 20]")
        >>> f.has_fixed_shape, f.shape
        (True, (10, 20))

        Numpy arrays can be used as template for shape and data type of field
        >>> arr_s, arr_v = np.zeros([20, 20]), np.zeros([20, 20, 2])
        >>> s, v = fields("s, v(2)", s=arr_s, v=arr_v)
        >>> assert s.index_dimensions == 0 and v.index_shape == (2,) and s.dtype.numpy_dtype == arr_s.dtype

        Format string can be left out, field names are taken from keyword arguments.
        >>> fields(f1=arr_s, f2=arr_s)
        [f1, f2]

        The keyword names 'index_dimension' and 'layout' have special meaning and thus can not be used to pass
        numpy arrays:
        >>> f = fields(f=arr_v, index_dimensions=1)
        >>> assert f.index_dimensions == 1

        >>> f = fields("pdfs(19) : float32[3D]", layout='fzyx')
        >>> f.layout
        (2, 1, 0)
    """
    result = []
    if description:
        field_descriptions, dtype, shape = _parse_description(description)
        layout = 'numpy' if layout is None else layout
        for field_name, idx_shape in field_descriptions:
            if field_name in kwargs:
                arr = kwargs[field_name]
                idx_shape_of_arr = () if not len(idx_shape) else arr.shape[-len(idx_shape):]
                assert idx_shape_of_arr == idx_shape
                f = Field.create_from_numpy_array(field_name, kwargs[field_name], index_dimensions=len(idx_shape))
            elif isinstance(shape, tuple):
                f = Field.create_fixed_size(field_name, shape + idx_shape, dtype=dtype,
                                            index_dimensions=len(idx_shape), layout=layout)
            elif isinstance(shape, int):
                f = Field.create_generic(field_name, spatial_dimensions=shape, dtype=dtype,
                                         index_shape=idx_shape, layout=layout)
            elif shape is None:
                f = Field.create_generic(field_name, spatial_dimensions=2, dtype=dtype,
                                         index_shape=idx_shape, layout=layout)
            else:
                assert False
            result.append(f)
    else:
        assert layout is None, "Layout can not be specified when creating Field from numpy array"
        for field_name, arr in kwargs.items():
            result.append(Field.create_from_numpy_array(field_name, arr, index_dimensions=index_dimensions))

    if len(result) == 0:
        return None
    elif len(result) == 1:
        return result[0]
    else:
        return result


80
81
82
83
84
85
86
87
88
89
class FieldType(Enum):
    # generic fields
    GENERIC = 0
    # index fields are currently only used for boundary handling
    # the coordinates are not the loop counters in that case, but are read from this index field
    INDEXED = 1
    # communication buffer, used for (un)packing data in communication.
    BUFFER = 2

    @staticmethod
Martin Bauer's avatar
Martin Bauer committed
90
    def is_generic(field):
91
        assert isinstance(field, Field)
Martin Bauer's avatar
Martin Bauer committed
92
        return field.field_type == FieldType.GENERIC
93
94

    @staticmethod
Martin Bauer's avatar
Martin Bauer committed
95
    def is_indexed(field):
96
        assert isinstance(field, Field)
Martin Bauer's avatar
Martin Bauer committed
97
        return field.field_type == FieldType.INDEXED
98
99

    @staticmethod
Martin Bauer's avatar
Martin Bauer committed
100
    def is_buffer(field):
101
        assert isinstance(field, Field)
Martin Bauer's avatar
Martin Bauer committed
102
        return field.field_type == FieldType.BUFFER
103
104


105
class Field:
106
107
108
109
    """
    With fields one can formulate stencil-like update rules on structured grids.
    This Field class knows about the dimension, memory layout (strides) and optionally about the size of an array.

Martin Bauer's avatar
Martin Bauer committed
110
111
112
113
    Creating Fields:

        To create a field use one of the static create* members. There are two options:

114
        1. create a kernel with fixed loop sizes i.e. the shape of the array is already known. This is usually the
Martin Bauer's avatar
Martin Bauer committed
115
           case if just-in-time compilation directly from Python is done. (see :func:`Field.create_from_numpy_array`)
116
        2. create a more general kernel that works for variable array sizes. This can be used to create kernels
Martin Bauer's avatar
Martin Bauer committed
117
           beforehand for a library. (see :func:`Field.create_generic`)
118
119
120
121

    Dimensions:
        A field has spatial and index dimensions, where the spatial dimensions come first.
        The interpretation is that the field has multiple cells in (usually) two or three dimensional space which are
Martin Bauer's avatar
Martin Bauer committed
122
123
        looped over. Additionally  N values are stored per cell. In this case spatial_dimensions is two or three,
        and index_dimensions equals N. If you want to store a matrix on each point in a two dimensional grid, there
Martin Bauer's avatar
Martin Bauer committed
124
        are four dimensions, two spatial and two index dimensions: ``len(arr.shape) == spatial_dims + index_dims``
125
126
127

    Indexing:
        When accessing (indexing) a field the result is a FieldAccess which is derived from sympy Symbol.
Martin Bauer's avatar
Martin Bauer committed
128
        First specify the spatial offsets in [], then in case index_dimension>0 the indices in ()
Martin Bauer's avatar
Martin Bauer committed
129
        e.g. ``f[-1,0,0](7)``
130
131
132

    Example without index dimensions:
        >>> a = np.zeros([10, 10])
Martin Bauer's avatar
Martin Bauer committed
133
        >>> f = Field.create_from_numpy_array("f", a, index_dimensions=0)
134
135
136
        >>> jacobi = ( f[-1,0] + f[1,0] + f[0,-1] + f[0,1] ) / 4

    Example with index dimensions: LBM D2Q9 stream pull
Martin Bauer's avatar
Martin Bauer committed
137
        >>> from pystencils import Assignment
138
        >>> stencil = np.array([[0,0], [0,1], [0,-1]])
Martin Bauer's avatar
Martin Bauer committed
139
140
        >>> src = Field.create_generic("src", spatial_dimensions=2, index_dimensions=1)
        >>> dst = Field.create_generic("dst", spatial_dimensions=2, index_dimensions=1)
141
        >>> for i, offset in enumerate(stencil):
142
143
144
145
        ...     Assignment(dst[0,0](i), src[-offset](i))
        Assignment(dst_C^0, src_C^0)
        Assignment(dst_C^1, src_S^1)
        Assignment(dst_C^2, src_N^2)
146
    """
147
148

    @staticmethod
Martin Bauer's avatar
Martin Bauer committed
149
150
    def create_generic(field_name, spatial_dimensions, dtype=np.float64, index_dimensions=0, layout='numpy',
                       index_shape=None, field_type=FieldType.GENERIC) -> 'Field':
151
152
153
        """
        Creates a generic field where the field size is not fixed i.e. can be called with arrays of different sizes

Martin Bauer's avatar
Martin Bauer committed
154
155
156
157
158
159
160
        Args:
            field_name: symbolic name for the field
            dtype: numpy data type of the array the kernel is called with later
            spatial_dimensions: see documentation of Field
            index_dimensions: see documentation of Field
            layout: tuple specifying the loop ordering of the spatial dimensions e.g. (2, 1, 0 ) means that
                    the outer loop loops over dimension 2, the second outer over dimension 1, and the inner loop
Martin Bauer's avatar
Martin Bauer committed
161
                    over dimension 0. Also allowed: the strings 'numpy' (0,1,..d) or 'reverse_numpy' (d, ..., 1, 0)
Martin Bauer's avatar
Martin Bauer committed
162
163
164
165
166
            index_shape: optional shape of the index dimensions i.e. maximum values allowed for each index dimension,
                        has to be a list or tuple
            field_type: besides the normal GENERIC fields, there are INDEXED fields that store indices of the domain
                        that should be iterated over, and BUFFER fields that are used to generate
                        communication packing/unpacking kernels
167
        """
168
        if isinstance(layout, str):
Martin Bauer's avatar
Martin Bauer committed
169
170
171
172
173
174
            layout = spatial_layout_string_to_tuple(layout, dim=spatial_dimensions)
        shape_symbol = IndexedBase(TypedSymbol(Field.SHAPE_PREFIX + field_name, Field.SHAPE_DTYPE), shape=(1,))
        stride_symbol = IndexedBase(TypedSymbol(Field.STRIDE_PREFIX + field_name, Field.STRIDE_DTYPE), shape=(1,))
        total_dimensions = spatial_dimensions + index_dimensions
        if index_shape is None or len(index_shape) == 0:
            shape = tuple([shape_symbol[i] for i in range(total_dimensions)])
175
        else:
Martin Bauer's avatar
Martin Bauer committed
176
            shape = tuple([shape_symbol[i] for i in range(spatial_dimensions)] + list(index_shape))
177

Martin Bauer's avatar
Martin Bauer committed
178
        strides = tuple([stride_symbol[i] for i in range(total_dimensions)])
179

Martin Bauer's avatar
Martin Bauer committed
180
181
182
        np_data_type = np.dtype(dtype)
        if np_data_type.fields is not None:
            if index_dimensions != 0:
183
184
185
186
                raise ValueError("Structured arrays/fields are not allowed to have an index dimension")
            shape += (1,)
            strides += (1,)

Martin Bauer's avatar
Martin Bauer committed
187
        return Field(field_name, field_type, dtype, layout, shape, strides)
188

189
    @staticmethod
Martin Bauer's avatar
Martin Bauer committed
190
191
192
    def create_from_numpy_array(field_name: str, array: np.ndarray, index_dimensions: int = 0) -> 'Field':
        """Creates a field based on the layout, data type, and shape of a given numpy array.

193
        Kernels created for these kind of fields can only be called with arrays of the same layout, shape and type.
Martin Bauer's avatar
Martin Bauer committed
194
195
196
197
198

        Args:
            field_name: symbolic name for the field
            array: numpy array
            index_dimensions: see documentation of Field
199
        """
Martin Bauer's avatar
Martin Bauer committed
200
201
        spatial_dimensions = len(array.shape) - index_dimensions
        if spatial_dimensions < 1:
202
203
            raise ValueError("Too many index dimensions. At least one spatial dimension required")

Martin Bauer's avatar
Martin Bauer committed
204
205
206
        full_layout = get_layout_of_array(array)
        spatial_layout = tuple([i for i in full_layout if i < spatial_dimensions])
        assert len(spatial_layout) == spatial_dimensions
207

Martin Bauer's avatar
Martin Bauer committed
208
209
        strides = tuple([s // np.dtype(array.dtype).itemsize for s in array.strides])
        shape = tuple(int(s) for s in array.shape)
210

Martin Bauer's avatar
Martin Bauer committed
211
212
213
        numpy_dtype = np.dtype(array.dtype)
        if numpy_dtype.fields is not None:
            if index_dimensions != 0:
214
215
216
217
                raise ValueError("Structured arrays/fields are not allowed to have an index dimension")
            shape += (1,)
            strides += (1,)

Martin Bauer's avatar
Martin Bauer committed
218
        return Field(field_name, FieldType.GENERIC, array.dtype, spatial_layout, shape, strides)
219
220

    @staticmethod
Martin Bauer's avatar
Martin Bauer committed
221
    def create_fixed_size(field_name: str, shape: Tuple[int, ...], index_dimensions: int = 0,
222
                          dtype=np.float64, layout: str = 'numpy', strides: Optional[Sequence[int]] = None) -> 'Field':
223
        """
224
        Creates a field with fixed sizes i.e. can be called only with arrays of the same size and layout
225

Martin Bauer's avatar
Martin Bauer committed
226
227
228
229
230
231
232
        Args:
            field_name: symbolic name for the field
            shape: overall shape of the array
            index_dimensions: how many of the trailing dimensions are interpreted as index (as opposed to spatial)
            dtype: numpy data type of the array the kernel is called with later
            layout: full layout of array, not only spatial dimensions
            strides: strides in bytes or None to automatically compute them from shape (assuming no padding)
233
        """
Martin Bauer's avatar
Martin Bauer committed
234
235
        spatial_dimensions = len(shape) - index_dimensions
        assert spatial_dimensions >= 1
236

237
        if isinstance(layout, str):
Martin Bauer's avatar
Martin Bauer committed
238
            layout = layout_string_to_tuple(layout, spatial_dimensions + index_dimensions)
239
240

        shape = tuple(int(s) for s in shape)
241
        if strides is None:
Martin Bauer's avatar
Martin Bauer committed
242
            strides = compute_strides(shape, layout)
243
244
245
        else:
            assert len(strides) == len(shape)
            strides = tuple([s // np.dtype(dtype).itemsize for s in strides])
246

Martin Bauer's avatar
Martin Bauer committed
247
248
249
        numpy_dtype = np.dtype(dtype)
        if numpy_dtype.fields is not None:
            if index_dimensions != 0:
250
251
252
253
                raise ValueError("Structured arrays/fields are not allowed to have an index dimension")
            shape += (1,)
            strides += (1,)

Martin Bauer's avatar
Martin Bauer committed
254
255
256
257
        spatial_layout = list(layout)
        for i in range(spatial_dimensions, len(layout)):
            spatial_layout.remove(i)
        return Field(field_name, FieldType.GENERIC, dtype, tuple(spatial_layout), shape, strides)
258

Martin Bauer's avatar
Martin Bauer committed
259
    def __init__(self, field_name, field_type, dtype, layout, shape, strides):
260
        """Do not use directly. Use static create* methods"""
261
        self._field_name = field_name
Martin Bauer's avatar
Martin Bauer committed
262
        assert isinstance(field_type, FieldType)
Martin Bauer's avatar
Martin Bauer committed
263
        self.field_type = field_type
Martin Bauer's avatar
Martin Bauer committed
264
        self._dtype = create_type(dtype)
Martin Bauer's avatar
Martin Bauer committed
265
        self._layout = normalize_layout(layout)
266
267
        self.shape = shape
        self.strides = strides
Martin Bauer's avatar
Martin Bauer committed
268
        self.latex_name: Optional[str] = None
269

Martin Bauer's avatar
Martin Bauer committed
270
    def new_field_with_different_name(self, new_name):
Martin Bauer's avatar
Martin Bauer committed
271
        return Field(new_name, self.field_type, self._dtype, self._layout, self.shape, self.strides)
272

273
    @property
Martin Bauer's avatar
Martin Bauer committed
274
    def spatial_dimensions(self) -> int:
275
276
277
        return len(self._layout)

    @property
Martin Bauer's avatar
Martin Bauer committed
278
    def index_dimensions(self) -> int:
279
        return len(self.shape) - len(self._layout)
280
281
282
283
284
285

    @property
    def layout(self):
        return self._layout

    @property
Martin Bauer's avatar
Martin Bauer committed
286
    def name(self) -> str:
287
        return self._field_name
288
289

    @property
Martin Bauer's avatar
Martin Bauer committed
290
291
    def spatial_shape(self) -> Tuple[int, ...]:
        return self.shape[:self.spatial_dimensions]
292

293
    @property
Martin Bauer's avatar
Martin Bauer committed
294
    def has_fixed_shape(self):
Martin Bauer's avatar
Martin Bauer committed
295
        return is_integer_sequence(self.shape)
296

297
    @property
Martin Bauer's avatar
Martin Bauer committed
298
299
    def index_shape(self):
        return self.shape[self.spatial_dimensions:]
300

301
    @property
Martin Bauer's avatar
Martin Bauer committed
302
303
    def has_fixed_index_shape(self):
        return is_integer_sequence(self.index_shape)
304

305
    @property
Martin Bauer's avatar
Martin Bauer committed
306
307
    def spatial_strides(self):
        return self.strides[:self.spatial_dimensions]
308
309

    @property
Martin Bauer's avatar
Martin Bauer committed
310
311
    def index_strides(self):
        return self.strides[self.spatial_dimensions:]
312
313
314
315
316
317

    @property
    def dtype(self):
        return self._dtype

    def __repr__(self):
318
        return self._field_name
319

Martin Bauer's avatar
Martin Bauer committed
320
321
322
323
    def neighbor(self, coord_id, offset):
        offset_list = [0] * self.spatial_dimensions
        offset_list[coord_id] = offset
        return Field.Access(self, tuple(offset_list))
324

325
    def neighbors(self, stencil):
326
        return [self.__getitem__(s) for s in stencil]
327

328
    @property
Martin Bauer's avatar
Martin Bauer committed
329
330
331
    def center_vector(self):
        index_shape = self.index_shape
        if len(index_shape) == 0:
332
            return self.center
Martin Bauer's avatar
Martin Bauer committed
333
334
335
        elif len(index_shape) == 1:
            return sp.Matrix([self(i) for i in range(index_shape[0])])
        elif len(index_shape) == 2:
336
337
338
            def cb(*args):
                r = self.__call__(*args)
                return r
Martin Bauer's avatar
Martin Bauer committed
339
            return sp.Matrix(*index_shape, cb)
340

341
    @property
342
    def center(self):
Martin Bauer's avatar
Martin Bauer committed
343
        center = tuple([0] * self.spatial_dimensions)
344
345
        return Field.Access(self, center)

346
347
348
349
    def __getitem__(self, offset):
        if type(offset) is np.ndarray:
            offset = tuple(offset)
        if type(offset) is str:
Martin Bauer's avatar
Martin Bauer committed
350
            offset = tuple(direction_string_to_offset(offset, self.spatial_dimensions))
351
352
        if type(offset) is not tuple:
            offset = (offset,)
Martin Bauer's avatar
Martin Bauer committed
353
        if len(offset) != self.spatial_dimensions:
354
            raise ValueError("Wrong number of spatial indices: "
Martin Bauer's avatar
Martin Bauer committed
355
                             "Got %d, expected %d" % (len(offset), self.spatial_dimensions))
356
357
358
        return Field.Access(self, offset)

    def __call__(self, *args, **kwargs):
Martin Bauer's avatar
Martin Bauer committed
359
        center = tuple([0] * self.spatial_dimensions)
360
361
362
        return Field.Access(self, center)(*args, **kwargs)

    def __hash__(self):
363
        return hash((self._layout, self.shape, self.strides, self._dtype, self.field_type, self._field_name))
364
365

    def __eq__(self, other):
Martin Bauer's avatar
Martin Bauer committed
366
367
        self_tuple = (self.shape, self.strides, self.name, self.dtype, self.field_type)
        other_tuple = (other.shape, other.strides, other.name, other.dtype, other.field_type)
Martin Bauer's avatar
Martin Bauer committed
368
        return self_tuple == other_tuple
369

370
371
372
    PREFIX = "f"
    STRIDE_PREFIX = PREFIX + "stride_"
    SHAPE_PREFIX = PREFIX + "shape_"
Martin Bauer's avatar
Martin Bauer committed
373
374
    STRIDE_DTYPE = create_composite_type_from_string("const int *")
    SHAPE_DTYPE = create_composite_type_from_string("const int *")
375
    DATA_PREFIX = PREFIX + "d_"
376

Martin Bauer's avatar
Martin Bauer committed
377
    # noinspection PyAttributeOutsideInit,PyUnresolvedReferences
378
379
380
381
382
383
    class Access(sp.Symbol):
        def __new__(cls, name, *args, **kwargs):
            obj = Field.Access.__xnew_cached_(cls, name, *args, **kwargs)
            return obj

        def __new_stage2__(self, field, offsets=(0, 0, 0), idx=None):
Martin Bauer's avatar
Martin Bauer committed
384
385
386
            field_name = field.name
            offsets_and_index = chain(offsets, idx) if idx is not None else offsets
            constant_offsets = not any([isinstance(o, sp.Basic) and not o.is_Integer for o in offsets_and_index])
387
388

            if not idx:
Martin Bauer's avatar
Martin Bauer committed
389
                idx = tuple([0] * field.index_dimensions)
390

Martin Bauer's avatar
Martin Bauer committed
391
392
393
            if constant_offsets:
                offset_name = offset_to_direction_string(offsets)
                if field.index_dimensions == 0:
394
                    superscript = None
Martin Bauer's avatar
Martin Bauer committed
395
                elif field.index_dimensions == 1:
396
                    superscript = str(idx[0])
397
                else:
Martin Bauer's avatar
Martin Bauer committed
398
399
400
401
                    idx_str = ",".join([str(e) for e in idx])
                    superscript = idx_str
                if field.has_fixed_index_shape and not isinstance(field.dtype, StructType):
                    for i, bound in zip(idx, field.index_shape):
402
403
                        if i >= bound:
                            raise ValueError("Field index out of bounds")
404
            else:
Martin Bauer's avatar
Martin Bauer committed
405
                offset_name = "%0.10X" % (abs(hash(tuple(offsets_and_index))))
406
                superscript = None
407

Martin Bauer's avatar
Martin Bauer committed
408
            symbol_name = "%s_%s" % (field_name, offset_name)
409
            if superscript is not None:
Martin Bauer's avatar
Martin Bauer committed
410
                symbol_name += "^" + superscript
411

Martin Bauer's avatar
Martin Bauer committed
412
            obj = super(Field.Access, self).__xnew__(self, symbol_name)
413
414
415
416
417
418
419
            obj._field = field
            obj._offsets = []
            for o in offsets:
                if isinstance(o, sp.Basic):
                    obj._offsets.append(o)
                else:
                    obj._offsets.append(int(o))
Martin Bauer's avatar
Martin Bauer committed
420
            obj._offsetName = offset_name
421
            obj._superscript = superscript
422
423
424
425
            obj._index = idx

            return obj

426
        def __getnewargs__(self):
427
            return self.field, self.offsets, self.index
428

Martin Bauer's avatar
Martin Bauer committed
429
        # noinspection SpellCheckingInspection
430
        __xnew__ = staticmethod(__new_stage2__)
Martin Bauer's avatar
Martin Bauer committed
431
        # noinspection SpellCheckingInspection
432
433
434
        __xnew_cached_ = staticmethod(cacheit(__new_stage2__))

        def __call__(self, *idx):
Martin Bauer's avatar
Martin Bauer committed
435
            if self._index != tuple([0] * self.field.index_dimensions):
436
437
438
                raise ValueError("Indexing an already indexed Field.Access")

            idx = tuple(idx)
439

Martin Bauer's avatar
Martin Bauer committed
440
            if self.field.index_dimensions == 0 and idx == (0,):
441
442
                idx = ()

Martin Bauer's avatar
Martin Bauer committed
443
            if len(idx) != self.field.index_dimensions:
444
                raise ValueError("Wrong number of indices: "
Martin Bauer's avatar
Martin Bauer committed
445
                                 "Got %d, expected %d" % (len(idx), self.field.index_dimensions))
446
447
            return Field.Access(self.field, self._offsets, idx)

Martin Bauer's avatar
Martin Bauer committed
448
449
450
        def __getitem__(self, *idx):
            return self.__call__(*idx)

Martin Bauer's avatar
Martin Bauer committed
451
452
453
454
455
        def __iter__(self):
            """This is necessary to work with parts of sympy that test if an object is iterable (e.g. simplify).
            The __getitem__ would make it iterable"""
            raise TypeError("Field access is not iterable")

456
457
458
459
460
461
        @property
        def field(self):
            return self._field

        @property
        def offsets(self):
462
            return tuple(self._offsets)
463

464
        @property
Martin Bauer's avatar
Martin Bauer committed
465
        def required_ghost_layers(self):
466
467
468
            return int(np.max(np.abs(self._offsets)))

        @property
Martin Bauer's avatar
Martin Bauer committed
469
        def nr_of_coordinates(self):
470
471
472
            return len(self._offsets)

        @property
Martin Bauer's avatar
Martin Bauer committed
473
        def offset_name(self) -> str:
474
475
476
477
478
479
            return self._offsetName

        @property
        def index(self):
            return self._index

Martin Bauer's avatar
Martin Bauer committed
480
481
482
483
        def neighbor(self, coord_id: int, offset: Sequence[int]) -> 'Field.Access':
            offset_list = list(self.offsets)
            offset_list[coord_id] += offset
            return Field.Access(self.field, tuple(offset_list), self.index)
484

Martin Bauer's avatar
Martin Bauer committed
485
        def get_shifted(self, *shift)-> 'Field.Access':
486
487
            return Field.Access(self.field, tuple(a + b for a, b in zip(shift, self.offsets)), self.index)

488
489
490
        def at_index(self, *idx_tuple):
            return Field.Access(self.field, self.offsets, idx_tuple)

491
        def _hashable_content(self):
Martin Bauer's avatar
Martin Bauer committed
492
493
            super_class_contents = list(super(Field.Access, self)._hashable_content())
            t = tuple(super_class_contents + [hash(self._field), self._index] + self._offsets)
494
            return t
Martin Bauer's avatar
Martin Bauer committed
495

Martin Bauer's avatar
Martin Bauer committed
496
        def _latex(self, _):
497
            n = self._field.latex_name if self._field.latex_name else self._field.name
Martin Bauer's avatar
Martin Bauer committed
498
499
500
501
            if self._superscript:
                return "{{%s}_{%s}^{%s}}" % (n, self._offsetName, self._superscript)
            else:
                return "{{%s}_{%s}}" % (n, self._offsetName)
Martin Bauer's avatar
Martin Bauer committed
502
503


Martin Bauer's avatar
Martin Bauer committed
504
505
def get_layout_from_strides(strides: Sequence[int], index_dimension_ids: Optional[List[int]] = None):
    index_dimension_ids = [] if index_dimension_ids is None else index_dimension_ids
506
    coordinates = list(range(len(strides)))
Martin Bauer's avatar
Martin Bauer committed
507
    relevant_strides = [stride for i, stride in enumerate(strides) if i not in index_dimension_ids]
508
    result = [x for (y, x) in sorted(zip(relevant_strides, coordinates), key=lambda pair: pair[0], reverse=True)]
Martin Bauer's avatar
Martin Bauer committed
509
    return normalize_layout(result)
510
511


Martin Bauer's avatar
Martin Bauer committed
512
513
514
515
516
517
def get_layout_of_array(arr: np.ndarray, index_dimension_ids: Optional[List[int]] = None):
    """ Returns a list indicating the memory layout (linearization order) of the numpy array.

    Examples:
        >>> get_layout_of_array(np.zeros([3,3,3]))
        (0, 1, 2)
Martin Bauer's avatar
Martin Bauer committed
518
519
520
521
522

    In this example the loop over the zeroth coordinate should be the outermost loop,
    followed by the first and second. Elements arr[x,y,0] and arr[x,y,1] are adjacent in memory.
    Normally constructed numpy arrays have this order, however by stride tricks or other frameworks, arrays
    with different memory layout can be created.
Martin Bauer's avatar
Martin Bauer committed
523

Martin Bauer's avatar
Martin Bauer committed
524
    The index_dimension_ids parameter leaves specifies which coordinates should not be
Martin Bauer's avatar
Martin Bauer committed
525
    """
Martin Bauer's avatar
Martin Bauer committed
526
527
    index_dimension_ids = [] if index_dimension_ids is None else index_dimension_ids
    return get_layout_from_strides(arr.strides, index_dimension_ids)
528
529


Martin Bauer's avatar
Martin Bauer committed
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
def create_numpy_array_with_layout(shape, layout, alignment=False, byte_offset=0, **kwargs):
    """Creates numpy array with given memory layout.

    Args:
        shape: shape of the resulting array
        layout: layout as tuple, where the coordinates are ordered from slow to fast
        alignment: number of bytes to align the beginning and the innermost coordinate to, or False for no alignment
        byte_offset: only used when alignment is specified, align not beginning but address at this offset
                     mostly used to align first inner cell, not ghost cells

    Example:
        >>> res = create_numpy_array_with_layout(shape=(2, 3, 4, 5), layout=(3, 2, 0, 1))
        >>> res.shape
        (2, 3, 4, 5)
        >>> get_layout_of_array(res)
        (3, 2, 0, 1)
546
547
    """
    assert set(layout) == set(range(len(shape))), "Wrong layout descriptor"
Martin Bauer's avatar
Martin Bauer committed
548
    cur_layout = list(range(len(shape)))
549
550
    swaps = []
    for i in range(len(layout)):
Martin Bauer's avatar
Martin Bauer committed
551
552
553
554
555
        if cur_layout[i] != layout[i]:
            index_to_swap_with = cur_layout.index(layout[i])
            swaps.append((i, index_to_swap_with))
            cur_layout[i], cur_layout[index_to_swap_with] = cur_layout[index_to_swap_with], cur_layout[i]
    assert tuple(cur_layout) == tuple(layout)
556
557
558
559
560

    shape = list(shape)
    for a, b in swaps:
        shape[a], shape[b] = shape[b], shape[a]

561
562
563
564
565
    if not alignment:
        res = np.empty(shape, order='c', **kwargs)
    else:
        if alignment is True:
            alignment = 8 * 4
Martin Bauer's avatar
Martin Bauer committed
566
        res = aligned_empty(shape, alignment, byte_offset=byte_offset, **kwargs)
567

568
569
570
571
572
    for a, b in reversed(swaps):
        res = res.swapaxes(a, b)
    return res


Martin Bauer's avatar
Martin Bauer committed
573
574
def spatial_layout_string_to_tuple(layout_str: str, dim: int) -> Tuple[int, ...]:
    if layout_str in ('fzyx', 'zyxf'):
575
576
        assert dim <= 3
        return tuple(reversed(range(dim)))
577

Martin Bauer's avatar
Martin Bauer committed
578
    if layout_str in ('fzyx', 'f', 'reverse_numpy', 'SoA'):
579
        return tuple(reversed(range(dim)))
Martin Bauer's avatar
Martin Bauer committed
580
    elif layout_str in ('c', 'numpy', 'AoS'):
581
        return tuple(range(dim))
Martin Bauer's avatar
Martin Bauer committed
582
    raise ValueError("Unknown layout descriptor " + layout_str)
583
584


Martin Bauer's avatar
Martin Bauer committed
585
586
587
def layout_string_to_tuple(layout_str, dim):
    layout_str = layout_str.lower()
    if layout_str == 'fzyx' or layout_str == 'soa':
588
589
        assert dim <= 4
        return tuple(reversed(range(dim)))
Martin Bauer's avatar
Martin Bauer committed
590
    elif layout_str == 'zyxf' or layout_str == 'aos':
591
        assert dim <= 4
Martin Bauer's avatar
Martin Bauer committed
592
        return tuple(reversed(range(dim - 1))) + (dim - 1,)
Martin Bauer's avatar
Martin Bauer committed
593
    elif layout_str == 'f' or layout_str == 'reverse_numpy':
594
        return tuple(reversed(range(dim)))
Martin Bauer's avatar
Martin Bauer committed
595
    elif layout_str == 'c' or layout_str == 'numpy':
596
        return tuple(range(dim))
Martin Bauer's avatar
Martin Bauer committed
597
    raise ValueError("Unknown layout descriptor " + layout_str)
598
599


Martin Bauer's avatar
Martin Bauer committed
600
def normalize_layout(layout):
601
    """Takes a layout tuple and subtracts the minimum from all entries"""
Martin Bauer's avatar
Martin Bauer committed
602
603
    min_entry = min(layout)
    return tuple(i - min_entry for i in layout)
604
605


Martin Bauer's avatar
Martin Bauer committed
606
def compute_strides(shape, layout):
607
608
    """
    Computes strides assuming no padding exists
Martin Bauer's avatar
Martin Bauer committed
609
610
611
612
613
614
615

    Args:
        shape: shape (size) of array
        layout: layout specification as tuple

    Returns:
        strides in elements, not in bytes
616
    """
Martin Bauer's avatar
Martin Bauer committed
617
618
619
620
    dim = len(shape)
    assert len(layout) == dim
    assert len(set(layout)) == dim
    strides = [0] * dim
621
    product = 1
622
    for j in reversed(layout):
623
624
625
        strides[j] = product
        product *= shape[j]
    return tuple(strides)
Martin Bauer's avatar
Martin Bauer committed
626
627


Martin Bauer's avatar
Martin Bauer committed
628
629
630
def offset_component_to_direction_string(coordinate_id: int, value: int) -> str:
    """Translates numerical offset to string notation.

Martin Bauer's avatar
Martin Bauer committed
631
632
633
634
635
    x offsets are labeled with east 'E' and 'W',
    y offsets with north 'N' and 'S' and
    z offsets with top 'T' and bottom 'B'
    If the absolute value of the offset is bigger than 1, this number is prefixed.

Martin Bauer's avatar
Martin Bauer committed
636
637
638
639
640
641
642
643
644
    Args:
        coordinate_id: integer 0, 1 or 2 standing for x,y and z
        value: integer offset

    Examples:
        >>> offset_component_to_direction_string(0, 1)
        'E'
        >>> offset_component_to_direction_string(1, 2)
        '2N'
Martin Bauer's avatar
Martin Bauer committed
645
    """
646
    assert 0 <= coordinate_id < 3, "Works only for at most 3D arrays"
Martin Bauer's avatar
Martin Bauer committed
647
648
    name_components = (('W', 'E'),  # west, east
                       ('S', 'N'),  # south, north
649
                       ('B', 'T'))  # bottom, top
Martin Bauer's avatar
Martin Bauer committed
650
651
652
    if value == 0:
        result = ""
    elif value < 0:
Martin Bauer's avatar
Martin Bauer committed
653
        result = name_components[coordinate_id][0]
Martin Bauer's avatar
Martin Bauer committed
654
    else:
Martin Bauer's avatar
Martin Bauer committed
655
        result = name_components[coordinate_id][1]
Martin Bauer's avatar
Martin Bauer committed
656
657
658
659
660
    if abs(value) > 1:
        result = "%d%s" % (abs(value), result)
    return result


Martin Bauer's avatar
Martin Bauer committed
661
def offset_to_direction_string(offsets: Sequence[int]) -> str:
Martin Bauer's avatar
Martin Bauer committed
662
663
    """
    Translates numerical offset to string notation.
Martin Bauer's avatar
Martin Bauer committed
664
665
666
667
668
669
670
671
672
    For details see :func:`offset_component_to_direction_string`
    Args:
        offsets: 3-tuple with x,y,z offset

    Examples:
        >>> offset_to_direction_string([1, -1, 0])
        'SE'
        >>> offset_to_direction_string(([-3, 0, -2]))
        '2B3W'
Martin Bauer's avatar
Martin Bauer committed
673
    """
674
675
    if len(offsets) > 3:
        return str(offsets)
Martin Bauer's avatar
Martin Bauer committed
676
    names = ["", "", ""]
Martin Bauer's avatar
Martin Bauer committed
677
678
    for i in range(len(offsets)):
        names[i] = offset_component_to_direction_string(i, offsets[i])
Martin Bauer's avatar
Martin Bauer committed
679
680
681
682
683
684
    name = "".join(reversed(names))
    if name == "":
        name = "C"
    return name


Martin Bauer's avatar
Martin Bauer committed
685
def direction_string_to_offset(direction: str, dim: int = 3):
Martin Bauer's avatar
Martin Bauer committed
686
    """
Martin Bauer's avatar
Martin Bauer committed
687
    Reverse mapping of :func:`offset_to_direction_string`
Martin Bauer's avatar
Martin Bauer committed
688
689
690
691
692
693
694
695
696
697
698
699

    Args:
        direction: string representation of offset
        dim: dimension of offset, i.e the length of the returned list

    Examples:
        >>> direction_string_to_offset('NW', dim=3)
        array([-1,  1,  0])
        >>> direction_string_to_offset('NW', dim=2)
        array([-1,  1])
        >>> direction_string_to_offset(offset_to_direction_string((3,-2,1)))
        array([ 3, -2,  1])
Martin Bauer's avatar
Martin Bauer committed
700
    """
Martin Bauer's avatar
Martin Bauer committed
701
    offset_dict = {
Martin Bauer's avatar
Martin Bauer committed
702
703
704
705
706
707
708
709
710
711
712
713
714
        'C': np.array([0, 0, 0]),

        'W': np.array([-1, 0, 0]),
        'E': np.array([1, 0, 0]),

        'S': np.array([0, -1, 0]),
        'N': np.array([0, 1, 0]),

        'B': np.array([0, 0, -1]),
        'T': np.array([0, 0, 1]),
    }
    offset = np.array([0, 0, 0])

Martin Bauer's avatar
Martin Bauer committed
715
    while len(direction) > 0:
Martin Bauer's avatar
Martin Bauer committed
716
        factor = 1
Martin Bauer's avatar
Martin Bauer committed
717
718
719
720
721
722
723
724
725
        first_non_digit = 0
        while direction[first_non_digit].isdigit():
            first_non_digit += 1
        if first_non_digit > 0:
            factor = int(direction[:first_non_digit])
            direction = direction[first_non_digit:]
        cur_offset = offset_dict[direction[0]]
        offset += factor * cur_offset
        direction = direction[1:]
Martin Bauer's avatar
Martin Bauer committed
726
    return offset[:dim]
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778

def _parse_type_description(type_description):
    if not type_description:
        return np.float64, None
    elif '[' in type_description:
        assert type_description[-1] == ']'
        splitted = type_description[:-1].split("[", )
        type_part, size_part = type_description[:-1].split("[", )
        if not type_part:
            type_part = "float64"
        if size_part.lower()[-1] == 'd':
            size_part = int(size_part[:-1])
        else:
            size_part = tuple(int(i) for i in size_part.split(','))
    else:
        type_part, size_part = type_description, None

    dtype = np.dtype(type_part).type
    return dtype, size_part


def _parse_field_description(description):
    if '(' not in description:
        return description, ()
    assert description[-1] == ')'
    name, index_shape = description[:-1].split('(')
    index_shape = tuple(int(i) for i in index_shape.split(','))
    return name, index_shape


def _parse_description(description):
    description = description.replace(' ', '')
    if ':' in description:
        name_descr, type_descr = description.split(':')
    else:
        name_descr, type_descr = description, ''

    # correct ',' splits inside brackets
    field_names = name_descr.split(',')
    cleaned_field_names = []
    prefix = ''
    for field_name in field_names:
        full_field_name = prefix + field_name
        if '(' in full_field_name and ')' not in full_field_name:
            prefix += field_name + ','
        else:
            prefix = ''
            cleaned_field_names.append(full_field_name)

    dtype, size = _parse_type_description(type_descr)
    fields_info = tuple(_parse_field_description(fd) for fd in cleaned_field_names)
    return fields_info, dtype, size