1"""
2This file is separate from ``_add_newdocs.py`` so that it can be mocked out by
3our sphinx ``conf.py`` during doc builds, where we want to avoid showing
4platform-dependent information.
5"""
6import sys
7import os
8from numpy.core import dtype
9from numpy.core import numerictypes as _numerictypes
10from numpy.core.function_base import add_newdoc
11
12##############################################################################
13#
14# Documentation for concrete scalar classes
15#
16##############################################################################
17
18def numeric_type_aliases(aliases):
19 def type_aliases_gen():
20 for alias, doc in aliases:
21 try:
22 alias_type = getattr(_numerictypes, alias)
23 except AttributeError:
24 # The set of aliases that actually exist varies between platforms
25 pass
26 else:
27 yield (alias_type, alias, doc)
28 return list(type_aliases_gen())
29
30
31possible_aliases = numeric_type_aliases([
32 ('int8', '8-bit signed integer (``-128`` to ``127``)'),
33 ('int16', '16-bit signed integer (``-32_768`` to ``32_767``)'),
34 ('int32', '32-bit signed integer (``-2_147_483_648`` to ``2_147_483_647``)'),
35 ('int64', '64-bit signed integer (``-9_223_372_036_854_775_808`` to ``9_223_372_036_854_775_807``)'),
36 ('intp', 'Signed integer large enough to fit pointer, compatible with C ``intptr_t``'),
37 ('uint8', '8-bit unsigned integer (``0`` to ``255``)'),
38 ('uint16', '16-bit unsigned integer (``0`` to ``65_535``)'),
39 ('uint32', '32-bit unsigned integer (``0`` to ``4_294_967_295``)'),
40 ('uint64', '64-bit unsigned integer (``0`` to ``18_446_744_073_709_551_615``)'),
41 ('uintp', 'Unsigned integer large enough to fit pointer, compatible with C ``uintptr_t``'),
42 ('float16', '16-bit-precision floating-point number type: sign bit, 5 bits exponent, 10 bits mantissa'),
43 ('float32', '32-bit-precision floating-point number type: sign bit, 8 bits exponent, 23 bits mantissa'),
44 ('float64', '64-bit precision floating-point number type: sign bit, 11 bits exponent, 52 bits mantissa'),
45 ('float96', '96-bit extended-precision floating-point number type'),
46 ('float128', '128-bit extended-precision floating-point number type'),
47 ('complex64', 'Complex number type composed of 2 32-bit-precision floating-point numbers'),
48 ('complex128', 'Complex number type composed of 2 64-bit-precision floating-point numbers'),
49 ('complex192', 'Complex number type composed of 2 96-bit extended-precision floating-point numbers'),
50 ('complex256', 'Complex number type composed of 2 128-bit extended-precision floating-point numbers'),
51 ])
52
53
54def _get_platform_and_machine():
55 try:
56 system, _, _, _, machine = os.uname()
57 except AttributeError:
58 system = sys.platform
59 if system == 'win32':
60 machine = os.environ.get('PROCESSOR_ARCHITEW6432', '') \
61 or os.environ.get('PROCESSOR_ARCHITECTURE', '')
62 else:
63 machine = 'unknown'
64 return system, machine
65
66
67_system, _machine = _get_platform_and_machine()
68_doc_alias_string = f":Alias on this platform ({_system} {_machine}):"
69
70
71def add_newdoc_for_scalar_type(obj, fixed_aliases, doc):
72 # note: `:field: value` is rST syntax which renders as field lists.
73 o = getattr(_numerictypes, obj)
74
75 character_code = dtype(o).char
76 canonical_name_doc = "" if obj == o.__name__ else \
77 f":Canonical name: `numpy.{obj}`\n "
78 if fixed_aliases:
79 alias_doc = ''.join(f":Alias: `numpy.{alias}`\n "
80 for alias in fixed_aliases)
81 else:
82 alias_doc = ''
83 alias_doc += ''.join(f"{_doc_alias_string} `numpy.{alias}`: {doc}.\n "
84 for (alias_type, alias, doc) in possible_aliases if alias_type is o)
85
86 docstring = f"""
87 {doc.strip()}
88
89 :Character code: ``'{character_code}'``
90 {canonical_name_doc}{alias_doc}
91 """
92
93 add_newdoc('numpy.core.numerictypes', obj, docstring)
94
95
96add_newdoc_for_scalar_type('bool_', ['bool8'],
97 """
98 Boolean type (True or False), stored as a byte.
99
100 .. warning::
101
102 The :class:`bool_` type is not a subclass of the :class:`int_` type
103 (the :class:`bool_` is not even a number type). This is different
104 than Python's default implementation of :class:`bool` as a
105 sub-class of :class:`int`.
106 """)
107
108add_newdoc_for_scalar_type('byte', [],
109 """
110 Signed integer type, compatible with C ``char``.
111 """)
112
113add_newdoc_for_scalar_type('short', [],
114 """
115 Signed integer type, compatible with C ``short``.
116 """)
117
118add_newdoc_for_scalar_type('intc', [],
119 """
120 Signed integer type, compatible with C ``int``.
121 """)
122
123add_newdoc_for_scalar_type('int_', [],
124 """
125 Signed integer type, compatible with Python `int` and C ``long``.
126 """)
127
128add_newdoc_for_scalar_type('longlong', [],
129 """
130 Signed integer type, compatible with C ``long long``.
131 """)
132
133add_newdoc_for_scalar_type('ubyte', [],
134 """
135 Unsigned integer type, compatible with C ``unsigned char``.
136 """)
137
138add_newdoc_for_scalar_type('ushort', [],
139 """
140 Unsigned integer type, compatible with C ``unsigned short``.
141 """)
142
143add_newdoc_for_scalar_type('uintc', [],
144 """
145 Unsigned integer type, compatible with C ``unsigned int``.
146 """)
147
148add_newdoc_for_scalar_type('uint', [],
149 """
150 Unsigned integer type, compatible with C ``unsigned long``.
151 """)
152
153add_newdoc_for_scalar_type('ulonglong', [],
154 """
155 Signed integer type, compatible with C ``unsigned long long``.
156 """)
157
158add_newdoc_for_scalar_type('half', [],
159 """
160 Half-precision floating-point number type.
161 """)
162
163add_newdoc_for_scalar_type('single', [],
164 """
165 Single-precision floating-point number type, compatible with C ``float``.
166 """)
167
168add_newdoc_for_scalar_type('double', ['float_'],
169 """
170 Double-precision floating-point number type, compatible with Python `float`
171 and C ``double``.
172 """)
173
174add_newdoc_for_scalar_type('longdouble', ['longfloat'],
175 """
176 Extended-precision floating-point number type, compatible with C
177 ``long double`` but not necessarily with IEEE 754 quadruple-precision.
178 """)
179
180add_newdoc_for_scalar_type('csingle', ['singlecomplex'],
181 """
182 Complex number type composed of two single-precision floating-point
183 numbers.
184 """)
185
186add_newdoc_for_scalar_type('cdouble', ['cfloat', 'complex_'],
187 """
188 Complex number type composed of two double-precision floating-point
189 numbers, compatible with Python `complex`.
190 """)
191
192add_newdoc_for_scalar_type('clongdouble', ['clongfloat', 'longcomplex'],
193 """
194 Complex number type composed of two extended-precision floating-point
195 numbers.
196 """)
197
198add_newdoc_for_scalar_type('object_', [],
199 """
200 Any Python object.
201 """)
202
203add_newdoc_for_scalar_type('str_', ['unicode_'],
204 r"""
205 A unicode string.
206
207 When used in arrays, this type strips trailing null codepoints.
208
209 Unlike the builtin `str`, this supports the :ref:`python:bufferobjects`, exposing its
210 contents as UCS4:
211
212 >>> m = memoryview(np.str_("abc"))
213 >>> m.format
214 '3w'
215 >>> m.tobytes()
216 b'a\x00\x00\x00b\x00\x00\x00c\x00\x00\x00'
217 """)
218
219add_newdoc_for_scalar_type('bytes_', ['string_'],
220 r"""
221 A byte string.
222
223 When used in arrays, this type strips trailing null bytes.
224 """)
225
226add_newdoc_for_scalar_type('void', [],
227 r"""
228 np.void(length_or_data, /, dtype=None)
229
230 Create a new structured or unstructured void scalar.
231
232 Parameters
233 ----------
234 length_or_data : int, array-like, bytes-like, object
235 One of multiple meanings (see notes). The length or
236 bytes data of an unstructured void. Or alternatively,
237 the data to be stored in the new scalar when `dtype`
238 is provided.
239 This can be an array-like, in which case an array may
240 be returned.
241 dtype : dtype, optional
242 If provided the dtype of the new scalar. This dtype must
243 be "void" dtype (i.e. a structured or unstructured void,
244 see also :ref:`defining-structured-types`).
245
246 ..versionadded:: 1.24
247
248 Notes
249 -----
250 For historical reasons and because void scalars can represent both
251 arbitrary byte data and structured dtypes, the void constructor
252 has three calling conventions:
253
254 1. ``np.void(5)`` creates a ``dtype="V5"`` scalar filled with five
255 ``\0`` bytes. The 5 can be a Python or NumPy integer.
256 2. ``np.void(b"bytes-like")`` creates a void scalar from the byte string.
257 The dtype itemsize will match the byte string length, here ``"V10"``.
258 3. When a ``dtype=`` is passed the call is rougly the same as an
259 array creation. However, a void scalar rather than array is returned.
260
261 Please see the examples which show all three different conventions.
262
263 Examples
264 --------
265 >>> np.void(5)
266 void(b'\x00\x00\x00\x00\x00')
267 >>> np.void(b'abcd')
268 void(b'\x61\x62\x63\x64')
269 >>> np.void((5, 3.2, "eggs"), dtype="i,d,S5")
270 (5, 3.2, b'eggs') # looks like a tuple, but is `np.void`
271 >>> np.void(3, dtype=[('x', np.int8), ('y', np.int8)])
272 (3, 3) # looks like a tuple, but is `np.void`
273
274 """)
275
276add_newdoc_for_scalar_type('datetime64', [],
277 """
278 If created from a 64-bit integer, it represents an offset from
279 ``1970-01-01T00:00:00``.
280 If created from string, the string can be in ISO 8601 date
281 or datetime format.
282
283 >>> np.datetime64(10, 'Y')
284 numpy.datetime64('1980')
285 >>> np.datetime64('1980', 'Y')
286 numpy.datetime64('1980')
287 >>> np.datetime64(10, 'D')
288 numpy.datetime64('1970-01-11')
289
290 See :ref:`arrays.datetime` for more information.
291 """)
292
293add_newdoc_for_scalar_type('timedelta64', [],
294 """
295 A timedelta stored as a 64-bit integer.
296
297 See :ref:`arrays.datetime` for more information.
298 """)
299
300add_newdoc('numpy.core.numerictypes', "integer", ('is_integer',
301 """
302 integer.is_integer() -> bool
303
304 Return ``True`` if the number is finite with integral value.
305
306 .. versionadded:: 1.22
307
308 Examples
309 --------
310 >>> np.int64(-2).is_integer()
311 True
312 >>> np.uint32(5).is_integer()
313 True
314 """))
315
316# TODO: work out how to put this on the base class, np.floating
317for float_name in ('half', 'single', 'double', 'longdouble'):
318 add_newdoc('numpy.core.numerictypes', float_name, ('as_integer_ratio',
319 """
320 {ftype}.as_integer_ratio() -> (int, int)
321
322 Return a pair of integers, whose ratio is exactly equal to the original
323 floating point number, and with a positive denominator.
324 Raise `OverflowError` on infinities and a `ValueError` on NaNs.
325
326 >>> np.{ftype}(10.0).as_integer_ratio()
327 (10, 1)
328 >>> np.{ftype}(0.0).as_integer_ratio()
329 (0, 1)
330 >>> np.{ftype}(-.25).as_integer_ratio()
331 (-1, 4)
332 """.format(ftype=float_name)))
333
334 add_newdoc('numpy.core.numerictypes', float_name, ('is_integer',
335 f"""
336 {float_name}.is_integer() -> bool
337
338 Return ``True`` if the floating point number is finite with integral
339 value, and ``False`` otherwise.
340
341 .. versionadded:: 1.22
342
343 Examples
344 --------
345 >>> np.{float_name}(-2.0).is_integer()
346 True
347 >>> np.{float_name}(3.2).is_integer()
348 False
349 """))
350
351for int_name in ('int8', 'uint8', 'int16', 'uint16', 'int32', 'uint32',
352 'int64', 'uint64', 'int64', 'uint64', 'int64', 'uint64'):
353 # Add negative examples for signed cases by checking typecode
354 add_newdoc('numpy.core.numerictypes', int_name, ('bit_count',
355 f"""
356 {int_name}.bit_count() -> int
357
358 Computes the number of 1-bits in the absolute value of the input.
359 Analogous to the builtin `int.bit_count` or ``popcount`` in C++.
360
361 Examples
362 --------
363 >>> np.{int_name}(127).bit_count()
364 7""" +
365 (f"""
366 >>> np.{int_name}(-127).bit_count()
367 7
368 """ if dtype(int_name).char.islower() else "")))