File size: 24,237 Bytes
9c6594c |
1 2 3 4 5 6 7 8 9 10 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 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 |
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied. See the License for the
# specific language governing permissions and limitations
# under the License.
from pyarrow._compute import ( # noqa
Function,
FunctionOptions,
FunctionRegistry,
HashAggregateFunction,
HashAggregateKernel,
Kernel,
ScalarAggregateFunction,
ScalarAggregateKernel,
ScalarFunction,
ScalarKernel,
VectorFunction,
VectorKernel,
# Option classes
ArraySortOptions,
AssumeTimezoneOptions,
CastOptions,
CountOptions,
CumulativeOptions,
CumulativeSumOptions,
DayOfWeekOptions,
DictionaryEncodeOptions,
RunEndEncodeOptions,
ElementWiseAggregateOptions,
ExtractRegexOptions,
ExtractRegexSpanOptions,
FilterOptions,
IndexOptions,
JoinOptions,
ListSliceOptions,
ListFlattenOptions,
MakeStructOptions,
MapLookupOptions,
MatchSubstringOptions,
ModeOptions,
NullOptions,
PadOptions,
PairwiseOptions,
PartitionNthOptions,
PivotWiderOptions,
QuantileOptions,
RandomOptions,
RankOptions,
RankQuantileOptions,
ReplaceSliceOptions,
ReplaceSubstringOptions,
RoundBinaryOptions,
RoundOptions,
RoundTemporalOptions,
RoundToMultipleOptions,
ScalarAggregateOptions,
SelectKOptions,
SetLookupOptions,
SkewOptions,
SliceOptions,
SortOptions,
SplitOptions,
SplitPatternOptions,
StrftimeOptions,
StrptimeOptions,
StructFieldOptions,
TakeOptions,
TDigestOptions,
TrimOptions,
Utf8NormalizeOptions,
VarianceOptions,
WeekOptions,
WinsorizeOptions,
# Functions
call_function,
function_registry,
get_function,
list_functions,
# Udf
call_tabular_function,
register_scalar_function,
register_tabular_function,
register_aggregate_function,
register_vector_function,
UdfContext,
# Expressions
Expression,
)
from collections import namedtuple
import inspect
from textwrap import dedent
import warnings
import pyarrow as pa
from pyarrow import _compute_docstrings
from pyarrow.vendored import docscrape
def _get_arg_names(func):
return func._doc.arg_names
_OptionsClassDoc = namedtuple('_OptionsClassDoc', ('params',))
def _scrape_options_class_doc(options_class):
if not options_class.__doc__:
return None
doc = docscrape.NumpyDocString(options_class.__doc__)
return _OptionsClassDoc(doc['Parameters'])
def _decorate_compute_function(wrapper, exposed_name, func, options_class):
# Decorate the given compute function wrapper with useful metadata
# and documentation.
cpp_doc = func._doc
wrapper.__arrow_compute_function__ = dict(
name=func.name,
arity=func.arity,
options_class=cpp_doc.options_class,
options_required=cpp_doc.options_required)
wrapper.__name__ = exposed_name
wrapper.__qualname__ = exposed_name
doc_pieces = []
# 1. One-line summary
summary = cpp_doc.summary
if not summary:
arg_str = "arguments" if func.arity > 1 else "argument"
summary = ("Call compute function {!r} with the given {}"
.format(func.name, arg_str))
doc_pieces.append(f"{summary}.\n\n")
# 2. Multi-line description
description = cpp_doc.description
if description:
doc_pieces.append(f"{description}\n\n")
doc_addition = _compute_docstrings.function_doc_additions.get(func.name)
# 3. Parameter description
doc_pieces.append(dedent("""\
Parameters
----------
"""))
# 3a. Compute function parameters
arg_names = _get_arg_names(func)
for arg_name in arg_names:
if func.kind in ('vector', 'scalar_aggregate'):
arg_type = 'Array-like'
else:
arg_type = 'Array-like or scalar-like'
doc_pieces.append(f"{arg_name} : {arg_type}\n")
doc_pieces.append(" Argument to compute function.\n")
# 3b. Compute function option values
if options_class is not None:
options_class_doc = _scrape_options_class_doc(options_class)
if options_class_doc:
for p in options_class_doc.params:
doc_pieces.append(f"{p.name} : {p.type}\n")
for s in p.desc:
doc_pieces.append(f" {s}\n")
else:
warnings.warn(f"Options class {options_class.__name__} "
f"does not have a docstring", RuntimeWarning)
options_sig = inspect.signature(options_class)
for p in options_sig.parameters.values():
doc_pieces.append(dedent("""\
{0} : optional
Parameter for {1} constructor. Either `options`
or `{0}` can be passed, but not both at the same time.
""".format(p.name, options_class.__name__)))
doc_pieces.append(dedent(f"""\
options : pyarrow.compute.{options_class.__name__}, optional
Alternative way of passing options.
"""))
doc_pieces.append(dedent("""\
memory_pool : pyarrow.MemoryPool, optional
If not passed, will allocate memory from the default memory pool.
"""))
# 4. Custom addition (e.g. examples)
if doc_addition is not None:
doc_pieces.append("\n{}\n".format(dedent(doc_addition).strip("\n")))
wrapper.__doc__ = "".join(doc_pieces)
return wrapper
def _get_options_class(func):
class_name = func._doc.options_class
if not class_name:
return None
try:
return globals()[class_name]
except KeyError:
warnings.warn("Python binding for {} not exposed"
.format(class_name), RuntimeWarning)
return None
def _handle_options(name, options_class, options, args, kwargs):
if args or kwargs:
if options is not None:
raise TypeError(
"Function {!r} called with both an 'options' argument "
"and additional arguments"
.format(name))
return options_class(*args, **kwargs)
if options is not None:
if isinstance(options, dict):
return options_class(**options)
elif isinstance(options, options_class):
return options
raise TypeError(
"Function {!r} expected a {} parameter, got {}"
.format(name, options_class, type(options)))
return None
def _make_generic_wrapper(func_name, func, options_class, arity):
if options_class is None:
def wrapper(*args, memory_pool=None):
if arity is not Ellipsis and len(args) != arity:
raise TypeError(
f"{func_name} takes {arity} positional argument(s), "
f"but {len(args)} were given"
)
if args and isinstance(args[0], Expression):
return Expression._call(func_name, list(args))
return func.call(args, None, memory_pool)
else:
def wrapper(*args, memory_pool=None, options=None, **kwargs):
if arity is not Ellipsis:
if len(args) < arity:
raise TypeError(
f"{func_name} takes {arity} positional argument(s), "
f"but {len(args)} were given"
)
option_args = args[arity:]
args = args[:arity]
else:
option_args = ()
options = _handle_options(func_name, options_class, options,
option_args, kwargs)
if args and isinstance(args[0], Expression):
return Expression._call(func_name, list(args), options)
return func.call(args, options, memory_pool)
return wrapper
def _make_signature(arg_names, var_arg_names, options_class):
from inspect import Parameter
params = []
for name in arg_names:
params.append(Parameter(name, Parameter.POSITIONAL_ONLY))
for name in var_arg_names:
params.append(Parameter(name, Parameter.VAR_POSITIONAL))
if options_class is not None:
options_sig = inspect.signature(options_class)
for p in options_sig.parameters.values():
assert p.kind in (Parameter.POSITIONAL_OR_KEYWORD,
Parameter.KEYWORD_ONLY)
if var_arg_names:
# Cannot have a positional argument after a *args
p = p.replace(kind=Parameter.KEYWORD_ONLY)
params.append(p)
params.append(Parameter("options", Parameter.KEYWORD_ONLY,
default=None))
params.append(Parameter("memory_pool", Parameter.KEYWORD_ONLY,
default=None))
return inspect.Signature(params)
def _wrap_function(name, func):
options_class = _get_options_class(func)
arg_names = _get_arg_names(func)
has_vararg = arg_names and arg_names[-1].startswith('*')
if has_vararg:
var_arg_names = [arg_names.pop().lstrip('*')]
else:
var_arg_names = []
wrapper = _make_generic_wrapper(
name, func, options_class, arity=func.arity)
wrapper.__signature__ = _make_signature(arg_names, var_arg_names,
options_class)
return _decorate_compute_function(wrapper, name, func, options_class)
def _make_global_functions():
"""
Make global functions wrapping each compute function.
Note that some of the automatically-generated wrappers may be overridden
by custom versions below.
"""
g = globals()
reg = function_registry()
# Avoid clashes with Python keywords
rewrites = {'and': 'and_',
'or': 'or_'}
for cpp_name in reg.list_functions():
name = rewrites.get(cpp_name, cpp_name)
func = reg.get_function(cpp_name)
if func.kind == "hash_aggregate":
# Hash aggregate functions are not callable,
# so let's not expose them at module level.
continue
if func.kind == "scalar_aggregate" and func.arity == 0:
# Nullary scalar aggregate functions are not callable
# directly so let's not expose them at module level.
continue
assert name not in g, name
g[cpp_name] = g[name] = _wrap_function(name, func)
_make_global_functions()
def cast(arr, target_type=None, safe=None, options=None, memory_pool=None):
"""
Cast array values to another data type. Can also be invoked as an array
instance method.
Parameters
----------
arr : Array-like
target_type : DataType or str
Type to cast to
safe : bool, default True
Check for overflows or other unsafe conversions
options : CastOptions, default None
Additional checks pass by CastOptions
memory_pool : MemoryPool, optional
memory pool to use for allocations during function execution.
Examples
--------
>>> from datetime import datetime
>>> import pyarrow as pa
>>> arr = pa.array([datetime(2010, 1, 1), datetime(2015, 1, 1)])
>>> arr.type
TimestampType(timestamp[us])
You can use ``pyarrow.DataType`` objects to specify the target type:
>>> cast(arr, pa.timestamp('ms'))
<pyarrow.lib.TimestampArray object at ...>
[
2010-01-01 00:00:00.000,
2015-01-01 00:00:00.000
]
>>> cast(arr, pa.timestamp('ms')).type
TimestampType(timestamp[ms])
Alternatively, it is also supported to use the string aliases for these
types:
>>> arr.cast('timestamp[ms]')
<pyarrow.lib.TimestampArray object at ...>
[
2010-01-01 00:00:00.000,
2015-01-01 00:00:00.000
]
>>> arr.cast('timestamp[ms]').type
TimestampType(timestamp[ms])
Returns
-------
casted : Array
The cast result as a new Array
"""
safe_vars_passed = (safe is not None) or (target_type is not None)
if safe_vars_passed and (options is not None):
raise ValueError("Must either pass values for 'target_type' and 'safe'"
" or pass a value for 'options'")
if options is None:
target_type = pa.types.lib.ensure_type(target_type)
if safe is False:
options = CastOptions.unsafe(target_type)
else:
options = CastOptions.safe(target_type)
return call_function("cast", [arr], options, memory_pool)
def index(data, value, start=None, end=None, *, memory_pool=None):
"""
Find the index of the first occurrence of a given value.
Parameters
----------
data : Array-like
value : Scalar-like object
The value to search for.
start : int, optional
end : int, optional
memory_pool : MemoryPool, optional
If not passed, will allocate memory from the default memory pool.
Returns
-------
index : int
the index, or -1 if not found
Examples
--------
>>> import pyarrow as pa
>>> import pyarrow.compute as pc
>>> arr = pa.array(["Lorem", "ipsum", "dolor", "sit", "Lorem", "ipsum"])
>>> pc.index(arr, "ipsum")
<pyarrow.Int64Scalar: 1>
>>> pc.index(arr, "ipsum", start=2)
<pyarrow.Int64Scalar: 5>
>>> pc.index(arr, "amet")
<pyarrow.Int64Scalar: -1>
"""
if start is not None:
if end is not None:
data = data.slice(start, end - start)
else:
data = data.slice(start)
elif end is not None:
data = data.slice(0, end)
if not isinstance(value, pa.Scalar):
value = pa.scalar(value, type=data.type)
elif data.type != value.type:
value = pa.scalar(value.as_py(), type=data.type)
options = IndexOptions(value=value)
result = call_function('index', [data], options, memory_pool)
if start is not None and result.as_py() >= 0:
result = pa.scalar(result.as_py() + start, type=pa.int64())
return result
def take(data, indices, *, boundscheck=True, memory_pool=None):
"""
Select values (or records) from array- or table-like data given integer
selection indices.
The result will be of the same type(s) as the input, with elements taken
from the input array (or record batch / table fields) at the given
indices. If an index is null then the corresponding value in the output
will be null.
Parameters
----------
data : Array, ChunkedArray, RecordBatch, or Table
indices : Array, ChunkedArray
Must be of integer type
boundscheck : boolean, default True
Whether to boundscheck the indices. If False and there is an out of
bounds index, will likely cause the process to crash.
memory_pool : MemoryPool, optional
If not passed, will allocate memory from the default memory pool.
Returns
-------
result : depends on inputs
Selected values for the given indices
Examples
--------
>>> import pyarrow as pa
>>> arr = pa.array(["a", "b", "c", None, "e", "f"])
>>> indices = pa.array([0, None, 4, 3])
>>> arr.take(indices)
<pyarrow.lib.StringArray object at ...>
[
"a",
null,
"e",
null
]
"""
options = TakeOptions(boundscheck=boundscheck)
return call_function('take', [data, indices], options, memory_pool)
def fill_null(values, fill_value):
"""Replace each null element in values with a corresponding
element from fill_value.
If fill_value is scalar-like, then every null element in values
will be replaced with fill_value. If fill_value is array-like,
then the i-th element in values will be replaced with the i-th
element in fill_value.
The fill_value's type must be the same as that of values, or it
must be able to be implicitly casted to the array's type.
This is an alias for :func:`coalesce`.
Parameters
----------
values : Array, ChunkedArray, or Scalar-like object
Each null element is replaced with the corresponding value
from fill_value.
fill_value : Array, ChunkedArray, or Scalar-like object
If not same type as values, will attempt to cast.
Returns
-------
result : depends on inputs
Values with all null elements replaced
Examples
--------
>>> import pyarrow as pa
>>> arr = pa.array([1, 2, None, 3], type=pa.int8())
>>> fill_value = pa.scalar(5, type=pa.int8())
>>> arr.fill_null(fill_value)
<pyarrow.lib.Int8Array object at ...>
[
1,
2,
5,
3
]
>>> arr = pa.array([1, 2, None, 4, None])
>>> arr.fill_null(pa.array([10, 20, 30, 40, 50]))
<pyarrow.lib.Int64Array object at ...>
[
1,
2,
30,
4,
50
]
"""
if not isinstance(fill_value, (pa.Array, pa.ChunkedArray, pa.Scalar)):
fill_value = pa.scalar(fill_value, type=values.type)
elif values.type != fill_value.type:
fill_value = pa.scalar(fill_value.as_py(), type=values.type)
return call_function("coalesce", [values, fill_value])
def top_k_unstable(values, k, sort_keys=None, *, memory_pool=None):
"""
Select the indices of the top-k ordered elements from array- or table-like
data.
This is a specialization for :func:`select_k_unstable`. Output is not
guaranteed to be stable.
Parameters
----------
values : Array, ChunkedArray, RecordBatch, or Table
Data to sort and get top indices from.
k : int
The number of `k` elements to keep.
sort_keys : List-like
Column key names to order by when input is table-like data.
memory_pool : MemoryPool, optional
If not passed, will allocate memory from the default memory pool.
Returns
-------
result : Array
Indices of the top-k ordered elements
Examples
--------
>>> import pyarrow as pa
>>> import pyarrow.compute as pc
>>> arr = pa.array(["a", "b", "c", None, "e", "f"])
>>> pc.top_k_unstable(arr, k=3)
<pyarrow.lib.UInt64Array object at ...>
[
5,
4,
2
]
"""
if sort_keys is None:
sort_keys = []
if isinstance(values, (pa.Array, pa.ChunkedArray)):
sort_keys.append(("dummy", "descending"))
else:
sort_keys = map(lambda key_name: (key_name, "descending"), sort_keys)
options = SelectKOptions(k, sort_keys)
return call_function("select_k_unstable", [values], options, memory_pool)
def bottom_k_unstable(values, k, sort_keys=None, *, memory_pool=None):
"""
Select the indices of the bottom-k ordered elements from
array- or table-like data.
This is a specialization for :func:`select_k_unstable`. Output is not
guaranteed to be stable.
Parameters
----------
values : Array, ChunkedArray, RecordBatch, or Table
Data to sort and get bottom indices from.
k : int
The number of `k` elements to keep.
sort_keys : List-like
Column key names to order by when input is table-like data.
memory_pool : MemoryPool, optional
If not passed, will allocate memory from the default memory pool.
Returns
-------
result : Array of indices
Indices of the bottom-k ordered elements
Examples
--------
>>> import pyarrow as pa
>>> import pyarrow.compute as pc
>>> arr = pa.array(["a", "b", "c", None, "e", "f"])
>>> pc.bottom_k_unstable(arr, k=3)
<pyarrow.lib.UInt64Array object at ...>
[
0,
1,
2
]
"""
if sort_keys is None:
sort_keys = []
if isinstance(values, (pa.Array, pa.ChunkedArray)):
sort_keys.append(("dummy", "ascending"))
else:
sort_keys = map(lambda key_name: (key_name, "ascending"), sort_keys)
options = SelectKOptions(k, sort_keys)
return call_function("select_k_unstable", [values], options, memory_pool)
def random(n, *, initializer='system', options=None, memory_pool=None):
"""
Generate numbers in the range [0, 1).
Generated values are uniformly-distributed, double-precision
in range [0, 1). Algorithm and seed can be changed via RandomOptions.
Parameters
----------
n : int
Number of values to generate, must be greater than or equal to 0
initializer : int or str
How to initialize the underlying random generator.
If an integer is given, it is used as a seed.
If "system" is given, the random generator is initialized with
a system-specific source of (hopefully true) randomness.
Other values are invalid.
options : pyarrow.compute.RandomOptions, optional
Alternative way of passing options.
memory_pool : pyarrow.MemoryPool, optional
If not passed, will allocate memory from the default memory pool.
"""
options = RandomOptions(initializer=initializer)
return call_function("random", [], options, memory_pool, length=n)
def field(*name_or_index):
"""Reference a column of the dataset.
Stores only the field's name. Type and other information is known only when
the expression is bound to a dataset having an explicit scheme.
Nested references are allowed by passing multiple names or a tuple of
names. For example ``('foo', 'bar')`` references the field named "bar"
inside the field named "foo".
Parameters
----------
*name_or_index : string, multiple strings, tuple or int
The name or index of the (possibly nested) field the expression
references to.
Returns
-------
field_expr : Expression
Reference to the given field
Examples
--------
>>> import pyarrow.compute as pc
>>> pc.field("a")
<pyarrow.compute.Expression a>
>>> pc.field(1)
<pyarrow.compute.Expression FieldPath(1)>
>>> pc.field(("a", "b"))
<pyarrow.compute.Expression FieldRef.Nested(FieldRef.Name(a) ...
>>> pc.field("a", "b")
<pyarrow.compute.Expression FieldRef.Nested(FieldRef.Name(a) ...
"""
n = len(name_or_index)
if n == 1:
if isinstance(name_or_index[0], (str, int)):
return Expression._field(name_or_index[0])
elif isinstance(name_or_index[0], tuple):
return Expression._nested_field(name_or_index[0])
else:
raise TypeError(
"field reference should be str, multiple str, tuple or "
f"integer, got {type(name_or_index[0])}"
)
# In case of multiple strings not supplied in a tuple
else:
return Expression._nested_field(name_or_index)
def scalar(value):
"""Expression representing a scalar value.
Creates an Expression object representing a scalar value that can be used
in compute expressions and predicates.
Parameters
----------
value : bool, int, float or string
Python value of the scalar. This function accepts any value that can be
converted to a ``pyarrow.Scalar`` using ``pa.scalar()``.
Notes
-----
This function differs from ``pyarrow.scalar()`` in the following way:
* ``pyarrow.scalar()`` creates a ``pyarrow.Scalar`` object that represents
a single value in Arrow's memory model.
* ``pyarrow.compute.scalar()`` creates an ``Expression`` object representing
a scalar value that can be used in compute expressions, predicates, and
dataset filtering operations.
Returns
-------
scalar_expr : Expression
An Expression representing the scalar value
"""
return Expression._scalar(value)
|