/usr/lib/python2.7/dist-packages/tables/atom.py is in python-tables 3.4.2-4.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 | # -*- coding: utf-8 -*-
########################################################################
#
# License: BSD
# Created: December 16, 2004
# Author: Ivan Vilata i Balaguer - ivan at selidor dot net
#
# $Id$
#
########################################################################
"""Atom classes for describing dataset contents."""
from __future__ import absolute_import
# Imports
# =======
import re
import sys
import inspect
import warnings
import numpy
from .utils import SizeType
from .misc.enum import Enum
import six.moves.cPickle
import six
import warnings
from .exceptions import FlavorWarning
# Public variables
# ================
__docformat__ = 'reStructuredText'
"""The format of documentation strings in this module."""
all_types = set() # filled as atom classes are created
"""Set of all PyTables types."""
atom_map = {} # filled as atom classes are created
"""Maps atom kinds to item sizes and atom classes.
If there is a fixed set of possible item sizes for a given kind, the
kind maps to another mapping from item size in bytes to atom class.
Otherwise, the kind maps directly to the atom class.
"""
deftype_from_kind = {} # filled as atom classes are created
"""Maps atom kinds to their default atom type (if any)."""
# Public functions
# ================
_type_re = re.compile(r'^([a-z]+)([0-9]*)$')
def split_type(type):
"""Split a PyTables type into a PyTables kind and an item size.
Returns a tuple of (kind, itemsize). If no item size is present in the type
(in the form of a precision), the returned item size is None::
>>> split_type('int32')
('int', 4)
>>> split_type('string')
('string', None)
>>> split_type('int20')
Traceback (most recent call last):
...
ValueError: precision must be a multiple of 8: 20
>>> split_type('foo bar')
Traceback (most recent call last):
...
ValueError: malformed type: 'foo bar'
"""
match = _type_re.match(type)
if not match:
raise ValueError("malformed type: %r" % type)
kind, precision = match.groups()
itemsize = None
if precision:
precision = int(precision)
itemsize, remainder = divmod(precision, 8)
if remainder: # 0 could be a valid item size
raise ValueError("precision must be a multiple of 8: %d"
% precision)
return (kind, itemsize)
# Private functions
# =================
def _invalid_itemsize_error(kind, itemsize, itemsizes):
isizes = sorted(itemsizes)
return ValueError("invalid item size for kind ``%s``: %r; "
"it must be one of ``%r``"
% (kind, itemsize, isizes))
def _abstract_atom_init(deftype, defvalue):
"""Return a constructor for an abstract `Atom` class."""
defitemsize = split_type(deftype)[1]
def __init__(self, itemsize=defitemsize, shape=(), dflt=defvalue):
assert self.kind in atom_map
try:
atomclass = atom_map[self.kind][itemsize]
except KeyError:
raise _invalid_itemsize_error(self.kind, itemsize,
atom_map[self.kind])
self.__class__ = atomclass
atomclass.__init__(self, shape, dflt)
return __init__
def _normalize_shape(shape):
"""Check that the `shape` is safe to be used and return it as a tuple."""
if isinstance(shape, (numpy.integer, six.integer_types)):
if shape < 1:
raise ValueError("shape value must be greater than 0: %d"
% shape)
shape = (shape,) # N is a shorthand for (N,)
try:
shape = tuple(shape)
except TypeError:
raise TypeError("shape must be an integer or sequence: %r"
% (shape,))
# XXX Get from HDF5 library if possible.
# HDF5 does not support ranks greater than 32
if len(shape) > 32:
raise ValueError(
"shapes with rank > 32 are not supported: %r" % (shape,))
return tuple(SizeType(s) for s in shape)
def _normalize_default(value, dtype):
"""Return `value` as a valid default of NumPy type `dtype`."""
# Create NumPy objects as defaults
# This is better in order to serialize them as attributes
if value is None:
value = 0
basedtype = dtype.base
try:
default = numpy.array(value, dtype=basedtype)
except ValueError:
array = numpy.array(value)
if array.shape != basedtype.shape:
raise
# Maybe nested dtype with "scalar" value.
default = numpy.array(value, dtype=basedtype.base)
# 0-dim arrays will be representented as NumPy scalars
# (PyTables attribute convention)
if default.shape == ():
default = default[()]
return default
def _cmp_dispatcher(other_method_name):
"""Dispatch comparisons to a method of the *other* object.
Returns a new *rich comparison* method which dispatches calls to
the method `other_method_name` of the *other* object. If there is
no such method in the object, ``False`` is returned.
This is part of the implementation of a double dispatch pattern.
"""
def dispatched_cmp(self, other):
try:
other_method = getattr(other, other_method_name)
except AttributeError:
return False
return other_method(self)
return dispatched_cmp
# Helper classes
# ==============
class MetaAtom(type):
"""Atom metaclass.
This metaclass ensures that data about atom classes gets inserted
into the suitable registries.
"""
def __init__(class_, name, bases, dict_):
super(MetaAtom, class_).__init__(name, bases, dict_)
kind = dict_.get('kind')
itemsize = dict_.get('itemsize')
type_ = dict_.get('type')
deftype = dict_.get('_deftype')
if kind and deftype:
deftype_from_kind[kind] = deftype
if type_:
all_types.add(type_)
if kind and itemsize and not hasattr(itemsize, '__int__'):
# Atom classes with a non-fixed item size do have an
# ``itemsize``, but it's not a number (e.g. property).
atom_map[kind] = class_
return
if kind: # first definition of kind, make new entry
atom_map[kind] = {}
if itemsize and hasattr(itemsize, '__int__'): # fixed
kind = class_.kind # maybe from superclasses
atom_map[kind][int(itemsize)] = class_
# Atom classes
# ============
class Atom(six.with_metaclass(MetaAtom, object)):
"""Defines the type of atomic cells stored in a dataset.
The meaning of *atomic* is that individual elements of a cell can
not be extracted directly by indexing (i.e. __getitem__()) the
dataset; e.g. if a dataset has shape (2, 2) and its atoms have
shape (3,), to get the third element of the cell at (1, 0) one
should use dataset[1,0][2] instead of dataset[1,0,2].
The Atom class is meant to declare the different properties of the
*base element* (also known as *atom*) of CArray, EArray and
VLArray datasets, although they are also used to describe the base
elements of Array datasets. Atoms have the property that their
length is always the same. However, you can grow datasets along
the extensible dimension in the case of EArray or put a variable
number of them on a VLArray row. Moreover, they are not restricted
to scalar values, and they can be *fully multidimensional
objects*.
Parameters
----------
itemsize : int
For types with a non-fixed size, this sets the size in
bytes of individual items in the atom.
shape : tuple
Sets the shape of the atom. An integer shape of
N is equivalent to the tuple (N,).
dflt
Sets the default value for the atom.
The following are the public methods and attributes of the Atom class.
Notes
-----
A series of descendant classes are offered in order to make the
use of these element descriptions easier. You should use a
particular Atom descendant class whenever you know the exact type
you will need when writing your code. Otherwise, you may use one
of the Atom.from_*() factory Methods.
.. rubric:: Atom attributes
.. attribute:: dflt
The default value of the atom.
If the user does not supply a value for an element while
filling a dataset, this default value will be written to disk.
If the user supplies a scalar value for a multidimensional
atom, this value is automatically *broadcast* to all the items
in the atom cell. If dflt is not supplied, an appropriate zero
value (or *null* string) will be chosen by default. Please
note that default values are kept internally as NumPy objects.
.. attribute:: dtype
The NumPy dtype that most closely matches this atom.
.. attribute:: itemsize
Size in bytes of a single item in the atom.
Specially useful for atoms of the string kind.
.. attribute:: kind
The PyTables kind of the atom (a string).
.. attribute:: shape
The shape of the atom (a tuple for scalar atoms).
.. attribute:: type
The PyTables type of the atom (a string).
Atoms can be compared with atoms and other objects for
strict (in)equality without having to compare individual
attributes::
>>> atom1 = StringAtom(itemsize=10) # same as ``atom2``
>>> atom2 = Atom.from_kind('string', 10) # same as ``atom1``
>>> atom3 = IntAtom()
>>> atom1 == 'foo'
False
>>> atom1 == atom2
True
>>> atom2 != atom1
False
>>> atom1 == atom3
False
>>> atom3 != atom2
True
"""
# Class methods
# ~~~~~~~~~~~~~
@classmethod
def prefix(class_):
"""Return the atom class prefix."""
cname = class_.__name__
return cname[:cname.rfind('Atom')]
@classmethod
def from_sctype(class_, sctype, shape=(), dflt=None):
"""Create an Atom from a NumPy scalar type sctype.
Optional shape and default value may be specified as the
shape and dflt
arguments, respectively. Information in the
sctype not represented in an Atom is ignored::
>>> import numpy
>>> Atom.from_sctype(numpy.int16, shape=(2, 2))
Int16Atom(shape=(2, 2), dflt=0)
>>> Atom.from_sctype('S5', dflt='hello')
Traceback (most recent call last):
...
ValueError: unknown NumPy scalar type: 'S5'
>>> Atom.from_sctype('Float64')
Float64Atom(shape=(), dflt=0.0)
"""
if (not isinstance(sctype, type)
or not issubclass(sctype, numpy.generic)):
if sctype not in numpy.sctypeDict:
raise ValueError("unknown NumPy scalar type: %r" % (sctype,))
sctype = numpy.sctypeDict[sctype]
return class_.from_dtype(numpy.dtype((sctype, shape)), dflt)
@classmethod
def from_dtype(class_, dtype, dflt=None):
"""Create an Atom from a NumPy dtype.
An optional default value may be specified as the dflt
argument. Information in the dtype not represented in an Atom is
ignored::
>>> import numpy
>>> Atom.from_dtype(numpy.dtype((numpy.int16, (2, 2))))
Int16Atom(shape=(2, 2), dflt=0)
>>> Atom.from_dtype(numpy.dtype('Float64'))
Float64Atom(shape=(), dflt=0.0)
Note: for easier use in Python 3, where all strings lead to the
Unicode dtype, this dtype will also generate a StringAtom. Since
this is only viable for strings that are castable as ascii, a
warning is issued.
>>> Atom.from_dtype(numpy.dtype('U20')) # doctest: +SKIP
Atom.py:392: FlavorWarning: support for unicode type is very limited,
and only works for strings that can be cast as ascii
StringAtom(itemsize=20, shape=(), dflt=b'')
"""
basedtype = dtype.base
if basedtype.names:
raise ValueError("compound data types are not supported: %r"
% dtype)
if basedtype.shape != ():
raise ValueError("nested data types are not supported: %r"
% dtype)
if basedtype.kind == 'S': # can not reuse something like 'string80'
itemsize = basedtype.itemsize
return class_.from_kind('string', itemsize, dtype.shape, dflt)
elif basedtype.kind == 'U':
# workaround for unicode type (standard string type in Python 3)
warnings.warn("support for unicode type is very limited, "
"and only works for strings that can be cast as ascii", FlavorWarning)
itemsize = basedtype.itemsize // 4
assert str(itemsize) in basedtype.str, "something went wrong in handling unicode."
return class_.from_kind('string', itemsize, dtype.shape, dflt)
# Most NumPy types have direct correspondence with PyTables types.
return class_.from_type(basedtype.name, dtype.shape, dflt)
@classmethod
def from_type(class_, type, shape=(), dflt=None):
"""Create an Atom from a PyTables type.
Optional shape and default value may be specified as the
shape and dflt arguments, respectively::
>>> Atom.from_type('bool')
BoolAtom(shape=(), dflt=False)
>>> Atom.from_type('int16', shape=(2, 2))
Int16Atom(shape=(2, 2), dflt=0)
>>> Atom.from_type('string40', dflt='hello')
Traceback (most recent call last):
...
ValueError: unknown type: 'string40'
>>> Atom.from_type('Float64')
Traceback (most recent call last):
...
ValueError: unknown type: 'Float64'
"""
if type not in all_types:
raise ValueError("unknown type: %r" % (type,))
kind, itemsize = split_type(type)
return class_.from_kind(kind, itemsize, shape, dflt)
@classmethod
def from_kind(class_, kind, itemsize=None, shape=(), dflt=None):
"""Create an Atom from a PyTables kind.
Optional item size, shape and default value may be
specified as the itemsize, shape and dflt
arguments, respectively. Bear in mind that not all atoms support
a default item size::
>>> Atom.from_kind('int', itemsize=2, shape=(2, 2))
Int16Atom(shape=(2, 2), dflt=0)
>>> Atom.from_kind('int', shape=(2, 2))
Int32Atom(shape=(2, 2), dflt=0)
>>> Atom.from_kind('int', shape=1)
Int32Atom(shape=(1,), dflt=0)
>>> Atom.from_kind('string', dflt=b'hello')
Traceback (most recent call last):
...
ValueError: no default item size for kind ``string``
>>> Atom.from_kind('Float')
Traceback (most recent call last):
...
ValueError: unknown kind: 'Float'
Moreover, some kinds with atypical constructor signatures
are not supported; you need to use the proper
constructor::
>>> Atom.from_kind('enum') #doctest: +ELLIPSIS
Traceback (most recent call last):
...
ValueError: the ``enum`` kind is not supported...
"""
kwargs = {'shape': shape}
if kind not in atom_map:
raise ValueError("unknown kind: %r" % (kind,))
# This incompatibility detection may get out-of-date and is
# too hard-wired, but I couldn't come up with something
# smarter. -- Ivan (2007-02-08)
if kind in ['enum']:
raise ValueError("the ``%s`` kind is not supported; "
"please use the appropriate constructor"
% kind)
# If no `itemsize` is given, try to get the default type of the
# kind (which has a fixed item size).
if itemsize is None:
if kind not in deftype_from_kind:
raise ValueError("no default item size for kind ``%s``"
% kind)
type_ = deftype_from_kind[kind]
kind, itemsize = split_type(type_)
kdata = atom_map[kind]
# Look up the class and set a possible item size.
if hasattr(kdata, 'kind'): # atom class: non-fixed item size
atomclass = kdata
kwargs['itemsize'] = itemsize
else: # dictionary: fixed item size
if itemsize not in kdata:
raise _invalid_itemsize_error(kind, itemsize, kdata)
atomclass = kdata[itemsize]
# Only set a `dflt` argument if given (`None` may not be understood).
if dflt is not None:
kwargs['dflt'] = dflt
return atomclass(**kwargs)
# Properties
# ~~~~~~~~~~
@property
def size(self):
"""Total size in bytes of the atom."""
return self.dtype.itemsize
@property
def recarrtype(self):
"""String type to be used in numpy.rec.array()."""
return str(self.dtype.shape) + self.dtype.base.str[1:]
@property
def ndim(self):
"""The number of dimensions of the atom.
.. versionadded:: 2.4"""
return len(self.shape)
# Special methods
# ~~~~~~~~~~~~~~~
def __init__(self, nptype, shape, dflt):
if not hasattr(self, 'type'):
raise NotImplementedError("``%s`` is an abstract class; "
"please use one of its subclasses"
% self.__class__.__name__)
self.shape = shape = _normalize_shape(shape)
"""The shape of the atom (a tuple for scalar atoms)."""
# Curiously enough, NumPy isn't generally able to accept NumPy
# integers in a shape. ;(
npshape = tuple(int(s) for s in shape)
self.dtype = dtype = numpy.dtype((nptype, npshape))
"""The NumPy dtype that most closely matches this atom."""
self.dflt = _normalize_default(dflt, dtype)
"""The default value of the atom.
If the user does not supply a value for an element while
filling a dataset, this default value will be written to
disk. If the user supplies a scalar value for a
multidimensional atom, this value is automatically *broadcast*
to all the items in the atom cell. If dflt is not supplied, an
appropriate zero value (or *null* string) will be chosen by
default. Please note that default values are kept internally
as NumPy objects."""
def __repr__(self):
args = 'shape=%s, dflt=%r' % (self.shape, self.dflt)
if not hasattr(self.__class__.itemsize, '__int__'): # non-fixed
args = 'itemsize=%s, %s' % (self.itemsize, args)
return '%s(%s)' % (self.__class__.__name__, args)
__eq__ = _cmp_dispatcher('_is_equal_to_atom')
def __ne__(self, other):
return not self.__eq__(other)
# XXX: API incompatible change for PyTables 3 line
# Overriding __eq__ blocks inheritance of __hash__ in 3.x
# def __hash__(self):
# return hash((self.__class__, self.type, self.shape, self.itemsize,
# self.dflt))
# Public methods
# ~~~~~~~~~~~~~~
def copy(self, **override):
"""Get a copy of the atom, possibly overriding some arguments.
Constructor arguments to be overridden must be passed as
keyword arguments::
>>> atom1 = Int32Atom(shape=12)
>>> atom2 = atom1.copy()
>>> print(atom1)
Int32Atom(shape=(12,), dflt=0)
>>> print(atom2)
Int32Atom(shape=(12,), dflt=0)
>>> atom1 is atom2
False
>>> atom3 = atom1.copy(shape=(2, 2))
>>> print(atom3)
Int32Atom(shape=(2, 2), dflt=0)
>>> atom1.copy(foobar=42)
Traceback (most recent call last):
...
TypeError: __init__() got an unexpected keyword argument 'foobar'
"""
newargs = self._get_init_args()
newargs.update(override)
return self.__class__(**newargs)
# Private methods
# ~~~~~~~~~~~~~~~
def _get_init_args(self):
"""Get a dictionary of instance constructor arguments.
This implementation works on classes which use the same names
for both constructor arguments and instance attributes.
"""
# @COMPATIBILITY: inspect.getargspec has been deprecated since
# Python 3.5
try:
# inspect.signature is new in Python 3.5
signature = inspect.signature(self.__init__)
except AttributeError:
args = inspect.getargspec(self.__init__)[0]
else:
parameters = signature.parameters
args = [arg for arg, p in parameters.items()
if p.kind is p.POSITIONAL_OR_KEYWORD]
return dict((arg, getattr(self, arg)) for arg in args if arg != 'self')
def _is_equal_to_atom(self, atom):
"""Is this object equal to the given `atom`?"""
return (self.type == atom.type and self.shape == atom.shape
and self.itemsize == atom.itemsize
and numpy.all(self.dflt == atom.dflt))
class StringAtom(Atom):
"""Defines an atom of type string.
The item size is the *maximum* length in characters of strings.
"""
kind = 'string'
type = 'string'
_defvalue = b''
@property
def itemsize(self):
"Size in bytes of a sigle item in the atom."
return self.dtype.base.itemsize
def __init__(self, itemsize, shape=(), dflt=_defvalue):
if not hasattr(itemsize, '__int__') or int(itemsize) < 0:
raise ValueError("invalid item size for kind ``%s``: %r; "
"it must be a positive integer"
% ('string', itemsize))
Atom.__init__(self, 'S%d' % itemsize, shape, dflt)
class BoolAtom(Atom):
"""Defines an atom of type bool."""
kind = 'bool'
itemsize = 1
type = 'bool'
_deftype = 'bool8'
_defvalue = False
def __init__(self, shape=(), dflt=_defvalue):
Atom.__init__(self, self.type, shape, dflt)
class IntAtom(Atom):
"""Defines an atom of a signed integral type (int kind)."""
kind = 'int'
signed = True
_deftype = 'int32'
_defvalue = 0
__init__ = _abstract_atom_init(_deftype, _defvalue)
class UIntAtom(Atom):
"""Defines an atom of an unsigned integral type (uint kind)."""
kind = 'uint'
signed = False
_deftype = 'uint32'
_defvalue = 0
__init__ = _abstract_atom_init(_deftype, _defvalue)
class FloatAtom(Atom):
"""Defines an atom of a floating point type (float kind)."""
kind = 'float'
_deftype = 'float64'
_defvalue = 0.0
__init__ = _abstract_atom_init(_deftype, _defvalue)
def _create_numeric_class(baseclass, itemsize):
"""Create a numeric atom class with the given `baseclass` and an
`itemsize`."""
prefix = '%s%d' % (baseclass.prefix(), itemsize * 8)
type_ = prefix.lower()
classdict = {'itemsize': itemsize, 'type': type_,
'__doc__': "Defines an atom of type ``%s``." % type_}
def __init__(self, shape=(), dflt=baseclass._defvalue):
Atom.__init__(self, self.type, shape, dflt)
classdict['__init__'] = __init__
return type('%sAtom' % prefix, (baseclass,), classdict)
Int8Atom = _create_numeric_class(IntAtom, 1)
Int16Atom = _create_numeric_class(IntAtom, 2)
Int32Atom = _create_numeric_class(IntAtom, 4)
Int64Atom = _create_numeric_class(IntAtom, 8)
UInt8Atom = _create_numeric_class(UIntAtom, 1)
UInt16Atom = _create_numeric_class(UIntAtom, 2)
UInt32Atom = _create_numeric_class(UIntAtom, 4)
UInt64Atom = _create_numeric_class(UIntAtom, 8)
if hasattr(numpy, 'float16'):
Float16Atom = _create_numeric_class(FloatAtom, 2)
Float32Atom = _create_numeric_class(FloatAtom, 4)
Float64Atom = _create_numeric_class(FloatAtom, 8)
if hasattr(numpy, 'float96'):
Float96Atom = _create_numeric_class(FloatAtom, 12)
if hasattr(numpy, 'float128'):
Float128Atom = _create_numeric_class(FloatAtom, 16)
class ComplexAtom(Atom):
"""Defines an atom of kind complex.
Allowed item sizes are 8 (single precision) and 16 (double precision). This
class must be used instead of more concrete ones to avoid confusions with
numarray-like precision specifications used in PyTables 1.X.
"""
# This definition is a little more complex (no pun intended)
# because, although the complex kind is a normal numerical one,
# the usage of bottom-level classes is artificially forbidden.
# Everything will be back to normality when people has stopped
# using the old bottom-level complex classes.
kind = 'complex'
_deftype = 'complex128'
_defvalue = 0j
_isizes = [8, 16]
@property
def itemsize(self):
"Size in bytes of a sigle item in the atom."
return self.dtype.base.itemsize
# Only instances have a `type` attribute, so complex types must be
# registered by hand.
all_types.add('complex64')
all_types.add('complex128')
if hasattr(numpy, 'complex192'):
all_types.add('complex192')
_isizes.append(24)
if hasattr(numpy, 'complex256'):
all_types.add('complex256')
_isizes.append(32)
def __init__(self, itemsize, shape=(), dflt=_defvalue):
if itemsize not in self._isizes:
raise _invalid_itemsize_error('complex', itemsize, self._isizes)
self.type = '%s%d' % (self.kind, itemsize * 8)
Atom.__init__(self, self.type, shape, dflt)
class _ComplexErrorAtom(six.with_metaclass(type, ComplexAtom)):
"""Reminds the user to stop using the old complex atom names."""
def __init__(self, shape=(), dflt=ComplexAtom._defvalue):
raise TypeError(
"to avoid confusions with PyTables 1.X complex atom names, "
"please use ``ComplexAtom(itemsize=N)``, "
"where N=8 for single precision complex atoms, "
"and N=16 for double precision complex atoms")
Complex32Atom = Complex64Atom = Complex128Atom = _ComplexErrorAtom
if hasattr(numpy, 'complex192'):
Complex192Atom = _ComplexErrorAtom
if hasattr(numpy, 'complex256'):
Complex256Atom = _ComplexErrorAtom
class TimeAtom(Atom):
"""Defines an atom of time type (time kind).
There are two distinct supported types of time: a 32 bit integer value and
a 64 bit floating point value. Both of them reflect the number of seconds
since the Unix epoch. This atom has the property of being stored using the
HDF5 time datatypes.
"""
kind = 'time'
_deftype = 'time32'
_defvalue = 0
__init__ = _abstract_atom_init(_deftype, _defvalue)
class Time32Atom(TimeAtom):
"""Defines an atom of type time32."""
itemsize = 4
type = 'time32'
_defvalue = 0
def __init__(self, shape=(), dflt=_defvalue):
Atom.__init__(self, 'int32', shape, dflt)
class Time64Atom(TimeAtom):
"""Defines an atom of type time64."""
itemsize = 8
type = 'time64'
_defvalue = 0.0
def __init__(self, shape=(), dflt=_defvalue):
Atom.__init__(self, 'float64', shape, dflt)
class EnumAtom(Atom):
"""Description of an atom of an enumerated type.
Instances of this class describe the atom type used to store enumerated
values. Those values belong to an enumerated type, defined by the first
argument (enum) in the constructor of the atom, which accepts the same
kinds of arguments as the Enum class (see :ref:`EnumClassDescr`). The
enumerated type is stored in the enum attribute of the atom.
A default value must be specified as the second argument (dflt) in the
constructor; it must be the *name* (a string) of one of the enumerated
values in the enumerated type. When the atom is created, the corresponding
concrete value is broadcast and stored in the dflt attribute (setting
different default values for items in a multidimensional atom is not
supported yet). If the name does not match any value in the enumerated
type, a KeyError is raised.
Another atom must be specified as the base argument in order to determine
the base type used for storing the values of enumerated values in memory
and disk. This *storage atom* is kept in the base attribute of the created
atom. As a shorthand, you may specify a PyTables type instead of the
storage atom, implying that this has a scalar shape.
The storage atom should be able to represent each and every concrete value
in the enumeration. If it is not, a TypeError is raised. The default value
of the storage atom is ignored.
The type attribute of enumerated atoms is always enum.
Enumerated atoms also support comparisons with other objects::
>>> enum = ['T0', 'T1', 'T2']
>>> atom1 = EnumAtom(enum, 'T0', 'int8') # same as ``atom2``
>>> atom2 = EnumAtom(enum, 'T0', Int8Atom()) # same as ``atom1``
>>> atom3 = EnumAtom(enum, 'T0', 'int16')
>>> atom4 = Int8Atom()
>>> atom1 == enum
False
>>> atom1 == atom2
True
>>> atom2 != atom1
False
>>> atom1 == atom3
False
>>> atom1 == atom4
False
>>> atom4 != atom1
True
Examples
--------
The next C enum construction::
enum myEnum {
T0,
T1,
T2
};
would correspond to the following PyTables
declaration::
>>> my_enum_atom = EnumAtom(['T0', 'T1', 'T2'], 'T0', 'int32')
Please note the dflt argument with a value of 'T0'. Since the concrete
value matching T0 is unknown right now (we have not used explicit concrete
values), using the name is the only option left for defining a default
value for the atom.
The chosen representation of values for this enumerated atom uses unsigned
32-bit integers, which surely wastes quite a lot of memory. Another size
could be selected by using the base argument (this time with a full-blown
storage atom)::
>>> my_enum_atom = EnumAtom(['T0', 'T1', 'T2'], 'T0', UInt8Atom())
You can also define multidimensional arrays for data elements::
>>> my_enum_atom = EnumAtom(
... ['T0', 'T1', 'T2'], 'T0', base='uint32', shape=(3,2))
for 3x2 arrays of uint32.
"""
# Registering this class in the class map may be a little wrong,
# since the ``Atom.from_kind()`` method fails miserably with
# enumerations, as they don't support an ``itemsize`` argument.
# However, resetting ``__metaclass__`` to ``type`` doesn't seem to
# work and I don't feel like creating a subclass of ``MetaAtom``.
kind = 'enum'
type = 'enum'
# Properties
# ~~~~~~~~~~
@property
def itemsize(self):
"""Size in bytes of a single item in the atom."""
return self.dtype.base.itemsize
# Private methods
# ~~~~~~~~~~~~~~~
def _checkbase(self, base):
"""Check the `base` storage atom."""
if base.kind == 'enum':
raise TypeError("can not use an enumerated atom "
"as a storage atom: %r" % base)
# Check whether the storage atom can represent concrete values
# in the enumeration...
basedtype = base.dtype
pyvalues = [value for (name, value) in self.enum]
try:
npgenvalues = numpy.array(pyvalues)
except ValueError:
raise TypeError("concrete values are not uniformly-shaped")
try:
npvalues = numpy.array(npgenvalues, dtype=basedtype.base)
except ValueError:
raise TypeError("storage atom type is incompatible with "
"concrete values in the enumeration")
if npvalues.shape[1:] != basedtype.shape:
raise TypeError("storage atom shape does not match that of "
"concrete values in the enumeration")
if npvalues.tolist() != npgenvalues.tolist():
raise TypeError("storage atom type lacks precision for "
"concrete values in the enumeration")
# ...with some implementation limitations.
if not npvalues.dtype.kind in ['i', 'u']:
raise NotImplementedError("only integer concrete values "
"are supported for the moment, sorry")
if len(npvalues.shape) > 1:
raise NotImplementedError("only scalar concrete values "
"are supported for the moment, sorry")
def _get_init_args(self):
"""Get a dictionary of instance constructor arguments."""
return dict(enum=self.enum, dflt=self._defname,
base=self.base, shape=self.shape)
def _is_equal_to_atom(self, atom):
"""Is this object equal to the given `atom`?"""
return False
def _is_equal_to_enumatom(self, enumatom):
"""Is this object equal to the given `enumatom`?"""
return (self.enum == enumatom.enum and self.shape == enumatom.shape
and numpy.all(self.dflt == enumatom.dflt)
and self.base == enumatom.base)
# Special methods
# ~~~~~~~~~~~~~~~
def __init__(self, enum, dflt, base, shape=()):
if not isinstance(enum, Enum):
enum = Enum(enum)
self.enum = enum
if isinstance(base, str):
base = Atom.from_type(base)
self._checkbase(base)
self.base = base
default = enum[dflt] # check default value
self._defname = dflt # kept for representation purposes
# These are kept to ease dumping this particular
# representation of the enumeration to storage.
names, values = [], []
for (name, value) in enum:
names.append(name)
values.append(value)
basedtype = self.base.dtype
self._names = names
self._values = numpy.array(values, dtype=basedtype.base)
Atom.__init__(self, basedtype, shape, default)
def __repr__(self):
return ('EnumAtom(enum=%r, dflt=%r, base=%r, shape=%r)'
% (self.enum, self._defname, self.base, self.shape))
__eq__ = _cmp_dispatcher('_is_equal_to_enumatom')
# XXX: API incompatible change for PyTables 3 line
# Overriding __eq__ blocks inheritance of __hash__ in 3.x
# def __hash__(self):
# return hash((self.__class__, self.enum, self.shape, self.dflt,
# self.base))
class ReferenceAtom(Atom):
"""Defines an atom of type object to read references.
This atom is read-only.
"""
kind = 'reference'
type = 'object'
_deftype = 'NoneType'
_defvalue = None
# Properties
# ~~~~~~~~~~
@property
def itemsize(self):
"""Size in bytes of a single item in the atom."""
return self.dtype.base.itemsize
def __init__(self, shape=()):
Atom.__init__(self, self.type, shape, self._defvalue)
def __repr__(self):
return 'ReferenceAtom(shape=%s)' % (self.shape,)
# Pseudo-atom classes
# ===================
#
# Now, there come three special classes, `ObjectAtom`, `VLStringAtom`
# and `VLUnicodeAtom`, that actually do not descend from `Atom`, but
# which goal is so similar that they should be described here.
# Pseudo-atoms can only be used with `VLArray` datasets, and they do
# not support multidimensional values, nor multiple values per row.
#
# They can be recognised because they also have ``kind``, ``type`` and
# ``shape`` attributes, but no ``size``, ``itemsize`` or ``dflt``
# ones. Instead, they have a ``base`` atom which defines the elements
# used for storage.
#
# See ``examples/vlarray1.py`` and ``examples/vlarray2.py`` for
# further examples on `VLArray` datasets, including object
# serialization and string management.
class PseudoAtom(object):
"""Pseudo-atoms can only be used in ``VLArray`` nodes.
They can be recognised because they also have `kind`, `type` and
`shape` attributes, but no `size`, `itemsize` or `dflt` ones.
Instead, they have a `base` atom which defines the elements used
for storage.
"""
def __repr__(self):
return '%s()' % self.__class__.__name__
def toarray(self, object_):
"""Convert an `object_` into an array of base atoms."""
raise NotImplementedError
def fromarray(self, array):
"""Convert an `array` of base atoms into an object."""
raise NotImplementedError
class _BufferedAtom(PseudoAtom):
"""Pseudo-atom which stores data as a buffer (flat array of uints)."""
shape = ()
def toarray(self, object_):
buffer_ = self._tobuffer(object_)
array = numpy.ndarray(buffer=buffer_, dtype=self.base.dtype,
shape=len(buffer_))
return array
def _tobuffer(self, object_):
"""Convert an `object_` into a buffer."""
raise NotImplementedError
class VLStringAtom(_BufferedAtom):
"""Defines an atom of type ``vlstring``.
This class describes a *row* of the VLArray class, rather than an atom. It
differs from the StringAtom class in that you can only add *one instance of
it to one specific row*, i.e. the :meth:`VLArray.append` method only
accepts one object when the base atom is of this type.
This class stores bytestrings. It does not make assumptions on the
encoding of the string, and raw bytes are stored as is. To store a string
you will need to *explicitly* convert it to a bytestring before you can
save them::
>>> s = 'A unicode string: hbar = \u210f'
>>> bytestring = s.encode('utf-8')
>>> VLArray.append(bytestring) # doctest: +SKIP
For full Unicode support, using VLUnicodeAtom (see :ref:`VLUnicodeAtom`) is
recommended.
Variable-length string atoms do not accept parameters and they cause the
reads of rows to always return Python bytestrings. You can regard vlstring
atoms as an easy way to save generic variable length strings.
"""
kind = 'vlstring'
type = 'vlstring'
base = UInt8Atom()
def _tobuffer(self, object_):
if isinstance(object_, six.text_type):
warnings.warn("Storing non bytestrings in VLStringAtom is "
"deprecated.", DeprecationWarning)
elif not isinstance(object_, bytes):
raise TypeError("object is not a string: %r" % (object_,))
return numpy.string_(object_)
def fromarray(self, array):
return array.tostring()
class VLUnicodeAtom(_BufferedAtom):
"""Defines an atom of type vlunicode.
This class describes a *row* of the VLArray class, rather than an atom. It
is very similar to VLStringAtom (see :ref:`VLStringAtom`), but it stores
Unicode strings (using 32-bit characters a la UCS-4, so all strings of the
same length also take up the same space).
This class does not make assumptions on the encoding of plain input
strings. Plain strings are supported as long as no character is out of the
ASCII set; otherwise, you will need to *explicitly* convert them to Unicode
before you can save them.
Variable-length Unicode atoms do not accept parameters and they cause the
reads of rows to always return Python Unicode strings. You can regard
vlunicode atoms as an easy way to save variable length Unicode strings.
"""
kind = 'vlunicode'
type = 'vlunicode'
base = UInt32Atom()
if sys.version_info[0] > 2 or sys.maxunicode <= 0xffff:
# numpy.unicode_ no more implements the buffer interface in Python 3
#
# When the Python build is UCS-2, we need to promote the
# Unicode string to UCS-4. We *must* use a 0-d array since
# NumPy scalars inherit the UCS-2 encoding from Python (see
# NumPy ticket #525). Since ``_tobuffer()`` can't return an
# array, we must override ``toarray()`` itself.
def toarray(self, object_):
if isinstance(object_, bytes):
warnings.warn("Storing bytestrings in VLUnicodeAtom is "
"deprecated.", DeprecationWarning)
elif not isinstance(object_, six.text_type):
raise TypeError("object is not a string: %r" % (object_,))
ustr = six.text_type(object_)
uarr = numpy.array(ustr, dtype='U')
return numpy.ndarray(
buffer=uarr, dtype=self.base.dtype, shape=len(ustr))
def _tobuffer(self, object_):
# This works (and is used) only with UCS-4 builds of Python,
# where the width of the internal representation of a
# character matches that of the base atoms.
if isinstance(object_, bytes):
warnings.warn("Storing bytestrings in VLUnicodeAtom is "
"deprecated.", DeprecationWarning)
elif not isinstance(object_, six.text_type):
raise TypeError("object is not a string: %r" % (object_,))
return numpy.unicode_(object_)
def fromarray(self, array):
length = len(array)
if length == 0:
return u'' # ``array.view('U0')`` raises a `TypeError`
return array.view('U%d' % length).item()
class ObjectAtom(_BufferedAtom):
"""Defines an atom of type object.
This class is meant to fit *any* kind of Python object in a row of a
VLArray dataset by using pickle behind the scenes. Due to the fact that
you can not foresee how long will be the output of the pickle
serialization (i.e. the atom already has a *variable* length), you can only
fit *one object per row*. However, you can still group several objects in a
single tuple or list and pass it to the :meth:`VLArray.append` method.
Object atoms do not accept parameters and they cause the reads of rows to
always return Python objects. You can regard object atoms as an easy way to
save an arbitrary number of generic Python objects in a VLArray dataset.
"""
kind = 'object'
type = 'object'
base = UInt8Atom()
def _tobuffer(self, object_):
return six.moves.cPickle.dumps(object_, six.moves.cPickle.HIGHEST_PROTOCOL)
def fromarray(self, array):
# We have to check for an empty array because of a possible
# bug in HDF5 which makes it claim that a dataset has one
# record when in fact it is empty.
if array.size == 0:
return None
return six.moves.cPickle.loads(array.tostring())
|