| 1 |
"""Bigmem tests - tests for the 32-bit boundary in containers. |
| 2 |
|
| 3 |
These tests try to exercise the 32-bit boundary that is sometimes, if |
| 4 |
rarely, exceeded in practice, but almost never tested. They are really only |
| 5 |
meaningful on 64-bit builds on machines with a *lot* of memory, but the |
| 6 |
tests are always run, usually with very low memory limits to make sure the |
| 7 |
tests themselves don't suffer from bitrot. To run them for real, pass a |
| 8 |
high memory limit to regrtest, with the -M option. |
| 9 |
""" |
| 10 |
|
| 11 |
from test import support |
| 12 |
from test.support import bigmemtest, _1G, _2G, _4G, import_helper |
| 13 |
_testcapi = import_helper.import_module('_testcapi') |
| 14 |
|
| 15 |
import unittest |
| 16 |
import operator |
| 17 |
import sys |
| 18 |
|
| 19 |
# These tests all use one of the bigmemtest decorators to indicate how much |
| 20 |
# memory they use and how much memory they need to be even meaningful. The |
| 21 |
# decorators take two arguments: a 'memuse' indicator declaring |
| 22 |
# (approximate) bytes per size-unit the test will use (at peak usage), and a |
| 23 |
# 'minsize' indicator declaring a minimum *useful* size. A test that |
| 24 |
# allocates a bytestring to test various operations near the end will have a |
| 25 |
# minsize of at least 2Gb (or it wouldn't reach the 32-bit limit, so the |
| 26 |
# test wouldn't be very useful) and a memuse of 1 (one byte per size-unit, |
| 27 |
# if it allocates only one big string at a time.) |
| 28 |
# |
| 29 |
# When run with a memory limit set, both decorators skip tests that need |
| 30 |
# more memory than available to be meaningful. The precisionbigmemtest will |
| 31 |
# always pass minsize as size, even if there is much more memory available. |
| 32 |
# The bigmemtest decorator will scale size upward to fill available memory. |
| 33 |
# |
| 34 |
# Bigmem testing houserules: |
| 35 |
# |
| 36 |
# - Try not to allocate too many large objects. It's okay to rely on |
| 37 |
# refcounting semantics, and don't forget that 's = create_largestring()' |
| 38 |
# doesn't release the old 's' (if it exists) until well after its new |
| 39 |
# value has been created. Use 'del s' before the create_largestring call. |
| 40 |
# |
| 41 |
# - Do *not* compare large objects using assertEqual, assertIn or similar. |
| 42 |
# It's a lengthy operation and the errormessage will be utterly useless |
| 43 |
# due to its size. To make sure whether a result has the right contents, |
| 44 |
# better to use the strip or count methods, or compare meaningful slices. |
| 45 |
# |
| 46 |
# - Don't forget to test for large indices, offsets and results and such, |
| 47 |
# in addition to large sizes. Anything that probes the 32-bit boundary. |
| 48 |
# |
| 49 |
# - When repeating an object (say, a substring, or a small list) to create |
| 50 |
# a large object, make the subobject of a length that is not a power of |
| 51 |
# 2. That way, int-wrapping problems are more easily detected. |
| 52 |
# |
| 53 |
# - Despite the bigmemtest decorator, all tests will actually be called |
| 54 |
# with a much smaller number too, in the normal test run (5Kb currently.) |
| 55 |
# This is so the tests themselves get frequent testing. |
| 56 |
# Consequently, always make all large allocations based on the |
| 57 |
# passed-in 'size', and don't rely on the size being very large. Also, |
| 58 |
# memuse-per-size should remain sane (less than a few thousand); if your |
| 59 |
# test uses more, adjust 'size' upward, instead. |
| 60 |
|
| 61 |
# BEWARE: it seems that one failing test can yield other subsequent tests to |
| 62 |
# fail as well. I do not know whether it is due to memory fragmentation |
| 63 |
# issues, or other specifics of the platform malloc() routine. |
| 64 |
|
| 65 |
ascii_char_size = 1 |
| 66 |
ucs2_char_size = 2 |
| 67 |
ucs4_char_size = 4 |
| 68 |
pointer_size = 4 if sys.maxsize < 2**32 else 8 |
| 69 |
|
| 70 |
|
| 71 |
class BaseStrTest: |
| 72 |
|
| 73 |
def _test_capitalize(self, size): |
| 74 |
_ = self.from_latin1 |
| 75 |
SUBSTR = self.from_latin1(' abc def ghi') |
| 76 |
s = _('-') * size + SUBSTR |
| 77 |
caps = s.capitalize() |
| 78 |
self.assertEqual(caps[-len(SUBSTR):], |
| 79 |
SUBSTR.capitalize()) |
| 80 |
self.assertEqual(caps.lstrip(_('-')), SUBSTR) |
| 81 |
|
| 82 |
@bigmemtest(size=_2G + 10, memuse=1) |
| 83 |
def test_center(self, size): |
| 84 |
SUBSTR = self.from_latin1(' abc def ghi') |
| 85 |
s = SUBSTR.center(size) |
| 86 |
self.assertEqual(len(s), size) |
| 87 |
lpadsize = rpadsize = (len(s) - len(SUBSTR)) // 2 |
| 88 |
if len(s) % 2: |
| 89 |
lpadsize += 1 |
| 90 |
self.assertEqual(s[lpadsize:-rpadsize], SUBSTR) |
| 91 |
self.assertEqual(s.strip(), SUBSTR.strip()) |
| 92 |
|
| 93 |
@bigmemtest(size=_2G, memuse=2) |
| 94 |
def test_count(self, size): |
| 95 |
_ = self.from_latin1 |
| 96 |
SUBSTR = _(' abc def ghi') |
| 97 |
s = _('.') * size + SUBSTR |
| 98 |
self.assertEqual(s.count(_('.')), size) |
| 99 |
s += _('.') |
| 100 |
self.assertEqual(s.count(_('.')), size + 1) |
| 101 |
self.assertEqual(s.count(_(' ')), 3) |
| 102 |
self.assertEqual(s.count(_('i')), 1) |
| 103 |
self.assertEqual(s.count(_('j')), 0) |
| 104 |
|
| 105 |
@bigmemtest(size=_2G, memuse=2) |
| 106 |
def test_endswith(self, size): |
| 107 |
_ = self.from_latin1 |
| 108 |
SUBSTR = _(' abc def ghi') |
| 109 |
s = _('-') * size + SUBSTR |
| 110 |
self.assertTrue(s.endswith(SUBSTR)) |
| 111 |
self.assertTrue(s.endswith(s)) |
| 112 |
s2 = _('...') + s |
| 113 |
self.assertTrue(s2.endswith(s)) |
| 114 |
self.assertFalse(s.endswith(_('a') + SUBSTR)) |
| 115 |
self.assertFalse(SUBSTR.endswith(s)) |
| 116 |
|
| 117 |
@bigmemtest(size=_2G + 10, memuse=2) |
| 118 |
def test_expandtabs(self, size): |
| 119 |
_ = self.from_latin1 |
| 120 |
s = _('-') * size |
| 121 |
tabsize = 8 |
| 122 |
self.assertTrue(s.expandtabs() == s) |
| 123 |
del s |
| 124 |
slen, remainder = divmod(size, tabsize) |
| 125 |
s = _(' \t') * slen |
| 126 |
s = s.expandtabs(tabsize) |
| 127 |
self.assertEqual(len(s), size - remainder) |
| 128 |
self.assertEqual(len(s.strip(_(' '))), 0) |
| 129 |
|
| 130 |
@bigmemtest(size=_2G, memuse=2) |
| 131 |
def test_find(self, size): |
| 132 |
_ = self.from_latin1 |
| 133 |
SUBSTR = _(' abc def ghi') |
| 134 |
sublen = len(SUBSTR) |
| 135 |
s = _('').join([SUBSTR, _('-') * size, SUBSTR]) |
| 136 |
self.assertEqual(s.find(_(' ')), 0) |
| 137 |
self.assertEqual(s.find(SUBSTR), 0) |
| 138 |
self.assertEqual(s.find(_(' '), sublen), sublen + size) |
| 139 |
self.assertEqual(s.find(SUBSTR, len(SUBSTR)), sublen + size) |
| 140 |
self.assertEqual(s.find(_('i')), SUBSTR.find(_('i'))) |
| 141 |
self.assertEqual(s.find(_('i'), sublen), |
| 142 |
sublen + size + SUBSTR.find(_('i'))) |
| 143 |
self.assertEqual(s.find(_('i'), size), |
| 144 |
sublen + size + SUBSTR.find(_('i'))) |
| 145 |
self.assertEqual(s.find(_('j')), -1) |
| 146 |
|
| 147 |
@bigmemtest(size=_2G, memuse=2) |
| 148 |
def test_index(self, size): |
| 149 |
_ = self.from_latin1 |
| 150 |
SUBSTR = _(' abc def ghi') |
| 151 |
sublen = len(SUBSTR) |
| 152 |
s = _('').join([SUBSTR, _('-') * size, SUBSTR]) |
| 153 |
self.assertEqual(s.index(_(' ')), 0) |
| 154 |
self.assertEqual(s.index(SUBSTR), 0) |
| 155 |
self.assertEqual(s.index(_(' '), sublen), sublen + size) |
| 156 |
self.assertEqual(s.index(SUBSTR, sublen), sublen + size) |
| 157 |
self.assertEqual(s.index(_('i')), SUBSTR.index(_('i'))) |
| 158 |
self.assertEqual(s.index(_('i'), sublen), |
| 159 |
sublen + size + SUBSTR.index(_('i'))) |
| 160 |
self.assertEqual(s.index(_('i'), size), |
| 161 |
sublen + size + SUBSTR.index(_('i'))) |
| 162 |
self.assertRaises(ValueError, s.index, _('j')) |
| 163 |
|
| 164 |
@bigmemtest(size=_2G, memuse=2) |
| 165 |
def test_isalnum(self, size): |
| 166 |
_ = self.from_latin1 |
| 167 |
SUBSTR = _('123456') |
| 168 |
s = _('a') * size + SUBSTR |
| 169 |
self.assertTrue(s.isalnum()) |
| 170 |
s += _('.') |
| 171 |
self.assertFalse(s.isalnum()) |
| 172 |
|
| 173 |
@bigmemtest(size=_2G, memuse=2) |
| 174 |
def test_isalpha(self, size): |
| 175 |
_ = self.from_latin1 |
| 176 |
SUBSTR = _('zzzzzzz') |
| 177 |
s = _('a') * size + SUBSTR |
| 178 |
self.assertTrue(s.isalpha()) |
| 179 |
s += _('.') |
| 180 |
self.assertFalse(s.isalpha()) |
| 181 |
|
| 182 |
@bigmemtest(size=_2G, memuse=2) |
| 183 |
def test_isdigit(self, size): |
| 184 |
_ = self.from_latin1 |
| 185 |
SUBSTR = _('123456') |
| 186 |
s = _('9') * size + SUBSTR |
| 187 |
self.assertTrue(s.isdigit()) |
| 188 |
s += _('z') |
| 189 |
self.assertFalse(s.isdigit()) |
| 190 |
|
| 191 |
@bigmemtest(size=_2G, memuse=2) |
| 192 |
def test_islower(self, size): |
| 193 |
_ = self.from_latin1 |
| 194 |
chars = _(''.join( |
| 195 |
chr(c) for c in range(255) if not chr(c).isupper())) |
| 196 |
repeats = size // len(chars) + 2 |
| 197 |
s = chars * repeats |
| 198 |
self.assertTrue(s.islower()) |
| 199 |
s += _('A') |
| 200 |
self.assertFalse(s.islower()) |
| 201 |
|
| 202 |
@bigmemtest(size=_2G, memuse=2) |
| 203 |
def test_isspace(self, size): |
| 204 |
_ = self.from_latin1 |
| 205 |
whitespace = _(' \f\n\r\t\v') |
| 206 |
repeats = size // len(whitespace) + 2 |
| 207 |
s = whitespace * repeats |
| 208 |
self.assertTrue(s.isspace()) |
| 209 |
s += _('j') |
| 210 |
self.assertFalse(s.isspace()) |
| 211 |
|
| 212 |
@bigmemtest(size=_2G, memuse=2) |
| 213 |
def test_istitle(self, size): |
| 214 |
_ = self.from_latin1 |
| 215 |
SUBSTR = _('123456') |
| 216 |
s = _('').join([_('A'), _('a') * size, SUBSTR]) |
| 217 |
self.assertTrue(s.istitle()) |
| 218 |
s += _('A') |
| 219 |
self.assertTrue(s.istitle()) |
| 220 |
s += _('aA') |
| 221 |
self.assertFalse(s.istitle()) |
| 222 |
|
| 223 |
@bigmemtest(size=_2G, memuse=2) |
| 224 |
def test_isupper(self, size): |
| 225 |
_ = self.from_latin1 |
| 226 |
chars = _(''.join( |
| 227 |
chr(c) for c in range(255) if not chr(c).islower())) |
| 228 |
repeats = size // len(chars) + 2 |
| 229 |
s = chars * repeats |
| 230 |
self.assertTrue(s.isupper()) |
| 231 |
s += _('a') |
| 232 |
self.assertFalse(s.isupper()) |
| 233 |
|
| 234 |
@bigmemtest(size=_2G, memuse=2) |
| 235 |
def test_join(self, size): |
| 236 |
_ = self.from_latin1 |
| 237 |
s = _('A') * size |
| 238 |
x = s.join([_('aaaaa'), _('bbbbb')]) |
| 239 |
self.assertEqual(x.count(_('a')), 5) |
| 240 |
self.assertEqual(x.count(_('b')), 5) |
| 241 |
self.assertTrue(x.startswith(_('aaaaaA'))) |
| 242 |
self.assertTrue(x.endswith(_('Abbbbb'))) |
| 243 |
|
| 244 |
@bigmemtest(size=_2G + 10, memuse=1) |
| 245 |
def test_ljust(self, size): |
| 246 |
_ = self.from_latin1 |
| 247 |
SUBSTR = _(' abc def ghi') |
| 248 |
s = SUBSTR.ljust(size) |
| 249 |
self.assertTrue(s.startswith(SUBSTR + _(' '))) |
| 250 |
self.assertEqual(len(s), size) |
| 251 |
self.assertEqual(s.strip(), SUBSTR.strip()) |
| 252 |
|
| 253 |
@bigmemtest(size=_2G + 10, memuse=2) |
| 254 |
def test_lower(self, size): |
| 255 |
_ = self.from_latin1 |
| 256 |
s = _('A') * size |
| 257 |
s = s.lower() |
| 258 |
self.assertEqual(len(s), size) |
| 259 |
self.assertEqual(s.count(_('a')), size) |
| 260 |
|
| 261 |
@bigmemtest(size=_2G + 10, memuse=1) |
| 262 |
def test_lstrip(self, size): |
| 263 |
_ = self.from_latin1 |
| 264 |
SUBSTR = _('abc def ghi') |
| 265 |
s = SUBSTR.rjust(size) |
| 266 |
self.assertEqual(len(s), size) |
| 267 |
self.assertEqual(s.lstrip(), SUBSTR.lstrip()) |
| 268 |
del s |
| 269 |
s = SUBSTR.ljust(size) |
| 270 |
self.assertEqual(len(s), size) |
| 271 |
# Type-specific optimization |
| 272 |
if isinstance(s, (str, bytes)): |
| 273 |
stripped = s.lstrip() |
| 274 |
self.assertTrue(stripped is s) |
| 275 |
|
| 276 |
@bigmemtest(size=_2G + 10, memuse=2) |
| 277 |
def test_replace(self, size): |
| 278 |
_ = self.from_latin1 |
| 279 |
replacement = _('a') |
| 280 |
s = _(' ') * size |
| 281 |
s = s.replace(_(' '), replacement) |
| 282 |
self.assertEqual(len(s), size) |
| 283 |
self.assertEqual(s.count(replacement), size) |
| 284 |
s = s.replace(replacement, _(' '), size - 4) |
| 285 |
self.assertEqual(len(s), size) |
| 286 |
self.assertEqual(s.count(replacement), 4) |
| 287 |
self.assertEqual(s[-10:], _(' aaaa')) |
| 288 |
|
| 289 |
@bigmemtest(size=_2G, memuse=2) |
| 290 |
def test_rfind(self, size): |
| 291 |
_ = self.from_latin1 |
| 292 |
SUBSTR = _(' abc def ghi') |
| 293 |
sublen = len(SUBSTR) |
| 294 |
s = _('').join([SUBSTR, _('-') * size, SUBSTR]) |
| 295 |
self.assertEqual(s.rfind(_(' ')), sublen + size + SUBSTR.rfind(_(' '))) |
| 296 |
self.assertEqual(s.rfind(SUBSTR), sublen + size) |
| 297 |
self.assertEqual(s.rfind(_(' '), 0, size), SUBSTR.rfind(_(' '))) |
| 298 |
self.assertEqual(s.rfind(SUBSTR, 0, sublen + size), 0) |
| 299 |
self.assertEqual(s.rfind(_('i')), sublen + size + SUBSTR.rfind(_('i'))) |
| 300 |
self.assertEqual(s.rfind(_('i'), 0, sublen), SUBSTR.rfind(_('i'))) |
| 301 |
self.assertEqual(s.rfind(_('i'), 0, sublen + size), |
| 302 |
SUBSTR.rfind(_('i'))) |
| 303 |
self.assertEqual(s.rfind(_('j')), -1) |
| 304 |
|
| 305 |
@bigmemtest(size=_2G, memuse=2) |
| 306 |
def test_rindex(self, size): |
| 307 |
_ = self.from_latin1 |
| 308 |
SUBSTR = _(' abc def ghi') |
| 309 |
sublen = len(SUBSTR) |
| 310 |
s = _('').join([SUBSTR, _('-') * size, SUBSTR]) |
| 311 |
self.assertEqual(s.rindex(_(' ')), |
| 312 |
sublen + size + SUBSTR.rindex(_(' '))) |
| 313 |
self.assertEqual(s.rindex(SUBSTR), sublen + size) |
| 314 |
self.assertEqual(s.rindex(_(' '), 0, sublen + size - 1), |
| 315 |
SUBSTR.rindex(_(' '))) |
| 316 |
self.assertEqual(s.rindex(SUBSTR, 0, sublen + size), 0) |
| 317 |
self.assertEqual(s.rindex(_('i')), |
| 318 |
sublen + size + SUBSTR.rindex(_('i'))) |
| 319 |
self.assertEqual(s.rindex(_('i'), 0, sublen), SUBSTR.rindex(_('i'))) |
| 320 |
self.assertEqual(s.rindex(_('i'), 0, sublen + size), |
| 321 |
SUBSTR.rindex(_('i'))) |
| 322 |
self.assertRaises(ValueError, s.rindex, _('j')) |
| 323 |
|
| 324 |
@bigmemtest(size=_2G + 10, memuse=1) |
| 325 |
def test_rjust(self, size): |
| 326 |
_ = self.from_latin1 |
| 327 |
SUBSTR = _(' abc def ghi') |
| 328 |
s = SUBSTR.ljust(size) |
| 329 |
self.assertTrue(s.startswith(SUBSTR + _(' '))) |
| 330 |
self.assertEqual(len(s), size) |
| 331 |
self.assertEqual(s.strip(), SUBSTR.strip()) |
| 332 |
|
| 333 |
@bigmemtest(size=_2G + 10, memuse=1) |
| 334 |
def test_rstrip(self, size): |
| 335 |
_ = self.from_latin1 |
| 336 |
SUBSTR = _(' abc def ghi') |
| 337 |
s = SUBSTR.ljust(size) |
| 338 |
self.assertEqual(len(s), size) |
| 339 |
self.assertEqual(s.rstrip(), SUBSTR.rstrip()) |
| 340 |
del s |
| 341 |
s = SUBSTR.rjust(size) |
| 342 |
self.assertEqual(len(s), size) |
| 343 |
# Type-specific optimization |
| 344 |
if isinstance(s, (str, bytes)): |
| 345 |
stripped = s.rstrip() |
| 346 |
self.assertTrue(stripped is s) |
| 347 |
|
| 348 |
# The test takes about size bytes to build a string, and then about |
| 349 |
# sqrt(size) substrings of sqrt(size) in size and a list to |
| 350 |
# hold sqrt(size) items. It's close but just over 2x size. |
| 351 |
@bigmemtest(size=_2G, memuse=2.1) |
| 352 |
def test_split_small(self, size): |
| 353 |
_ = self.from_latin1 |
| 354 |
# Crudely calculate an estimate so that the result of s.split won't |
| 355 |
# take up an inordinate amount of memory |
| 356 |
chunksize = int(size ** 0.5 + 2) |
| 357 |
SUBSTR = _('a') + _(' ') * chunksize |
| 358 |
s = SUBSTR * chunksize |
| 359 |
l = s.split() |
| 360 |
self.assertEqual(len(l), chunksize) |
| 361 |
expected = _('a') |
| 362 |
for item in l: |
| 363 |
self.assertEqual(item, expected) |
| 364 |
del l |
| 365 |
l = s.split(_('a')) |
| 366 |
self.assertEqual(len(l), chunksize + 1) |
| 367 |
expected = _(' ') * chunksize |
| 368 |
for item in filter(None, l): |
| 369 |
self.assertEqual(item, expected) |
| 370 |
|
| 371 |
# Allocates a string of twice size (and briefly two) and a list of |
| 372 |
# size. Because of internal affairs, the s.split() call produces a |
| 373 |
# list of size times the same one-character string, so we only |
| 374 |
# suffer for the list size. (Otherwise, it'd cost another 48 times |
| 375 |
# size in bytes!) Nevertheless, a list of size takes |
| 376 |
# 8*size bytes. |
| 377 |
@bigmemtest(size=_2G + 5, memuse=ascii_char_size * 2 + pointer_size) |
| 378 |
def test_split_large(self, size): |
| 379 |
_ = self.from_latin1 |
| 380 |
s = _(' a') * size + _(' ') |
| 381 |
l = s.split() |
| 382 |
self.assertEqual(len(l), size) |
| 383 |
self.assertEqual(set(l), set([_('a')])) |
| 384 |
del l |
| 385 |
l = s.split(_('a')) |
| 386 |
self.assertEqual(len(l), size + 1) |
| 387 |
self.assertEqual(set(l), set([_(' ')])) |
| 388 |
|
| 389 |
@bigmemtest(size=_2G, memuse=2.1) |
| 390 |
def test_splitlines(self, size): |
| 391 |
_ = self.from_latin1 |
| 392 |
# Crudely calculate an estimate so that the result of s.split won't |
| 393 |
# take up an inordinate amount of memory |
| 394 |
chunksize = int(size ** 0.5 + 2) // 2 |
| 395 |
SUBSTR = _(' ') * chunksize + _('\n') + _(' ') * chunksize + _('\r\n') |
| 396 |
s = SUBSTR * (chunksize * 2) |
| 397 |
l = s.splitlines() |
| 398 |
self.assertEqual(len(l), chunksize * 4) |
| 399 |
expected = _(' ') * chunksize |
| 400 |
for item in l: |
| 401 |
self.assertEqual(item, expected) |
| 402 |
|
| 403 |
@bigmemtest(size=_2G, memuse=2) |
| 404 |
def test_startswith(self, size): |
| 405 |
_ = self.from_latin1 |
| 406 |
SUBSTR = _(' abc def ghi') |
| 407 |
s = _('-') * size + SUBSTR |
| 408 |
self.assertTrue(s.startswith(s)) |
| 409 |
self.assertTrue(s.startswith(_('-') * size)) |
| 410 |
self.assertFalse(s.startswith(SUBSTR)) |
| 411 |
|
| 412 |
@bigmemtest(size=_2G, memuse=1) |
| 413 |
def test_strip(self, size): |
| 414 |
_ = self.from_latin1 |
| 415 |
SUBSTR = _(' abc def ghi ') |
| 416 |
s = SUBSTR.rjust(size) |
| 417 |
self.assertEqual(len(s), size) |
| 418 |
self.assertEqual(s.strip(), SUBSTR.strip()) |
| 419 |
del s |
| 420 |
s = SUBSTR.ljust(size) |
| 421 |
self.assertEqual(len(s), size) |
| 422 |
self.assertEqual(s.strip(), SUBSTR.strip()) |
| 423 |
|
| 424 |
def _test_swapcase(self, size): |
| 425 |
_ = self.from_latin1 |
| 426 |
SUBSTR = _("aBcDeFG12.'\xa9\x00") |
| 427 |
sublen = len(SUBSTR) |
| 428 |
repeats = size // sublen + 2 |
| 429 |
s = SUBSTR * repeats |
| 430 |
s = s.swapcase() |
| 431 |
self.assertEqual(len(s), sublen * repeats) |
| 432 |
self.assertEqual(s[:sublen * 3], SUBSTR.swapcase() * 3) |
| 433 |
self.assertEqual(s[-sublen * 3:], SUBSTR.swapcase() * 3) |
| 434 |
|
| 435 |
def _test_title(self, size): |
| 436 |
_ = self.from_latin1 |
| 437 |
SUBSTR = _('SpaaHAaaAaham') |
| 438 |
s = SUBSTR * (size // len(SUBSTR) + 2) |
| 439 |
s = s.title() |
| 440 |
self.assertTrue(s.startswith((SUBSTR * 3).title())) |
| 441 |
self.assertTrue(s.endswith(SUBSTR.lower() * 3)) |
| 442 |
|
| 443 |
@bigmemtest(size=_2G, memuse=2) |
| 444 |
def test_translate(self, size): |
| 445 |
_ = self.from_latin1 |
| 446 |
SUBSTR = _('aZz.z.Aaz.') |
| 447 |
trans = bytes.maketrans(b'.aZ', b'-!$') |
| 448 |
sublen = len(SUBSTR) |
| 449 |
repeats = size // sublen + 2 |
| 450 |
s = SUBSTR * repeats |
| 451 |
s = s.translate(trans) |
| 452 |
self.assertEqual(len(s), repeats * sublen) |
| 453 |
self.assertEqual(s[:sublen], SUBSTR.translate(trans)) |
| 454 |
self.assertEqual(s[-sublen:], SUBSTR.translate(trans)) |
| 455 |
self.assertEqual(s.count(_('.')), 0) |
| 456 |
self.assertEqual(s.count(_('!')), repeats * 2) |
| 457 |
self.assertEqual(s.count(_('z')), repeats * 3) |
| 458 |
|
| 459 |
@bigmemtest(size=_2G + 5, memuse=2) |
| 460 |
def test_upper(self, size): |
| 461 |
_ = self.from_latin1 |
| 462 |
s = _('a') * size |
| 463 |
s = s.upper() |
| 464 |
self.assertEqual(len(s), size) |
| 465 |
self.assertEqual(s.count(_('A')), size) |
| 466 |
|
| 467 |
@bigmemtest(size=_2G + 20, memuse=1) |
| 468 |
def test_zfill(self, size): |
| 469 |
_ = self.from_latin1 |
| 470 |
SUBSTR = _('-568324723598234') |
| 471 |
s = SUBSTR.zfill(size) |
| 472 |
self.assertTrue(s.endswith(_('0') + SUBSTR[1:])) |
| 473 |
self.assertTrue(s.startswith(_('-0'))) |
| 474 |
self.assertEqual(len(s), size) |
| 475 |
self.assertEqual(s.count(_('0')), size - len(SUBSTR)) |
| 476 |
|
| 477 |
# This test is meaningful even with size < 2G, as long as the |
| 478 |
# doubled string is > 2G (but it tests more if both are > 2G :) |
| 479 |
@bigmemtest(size=_1G + 2, memuse=3) |
| 480 |
def test_concat(self, size): |
| 481 |
_ = self.from_latin1 |
| 482 |
s = _('.') * size |
| 483 |
self.assertEqual(len(s), size) |
| 484 |
s = s + s |
| 485 |
self.assertEqual(len(s), size * 2) |
| 486 |
self.assertEqual(s.count(_('.')), size * 2) |
| 487 |
|
| 488 |
# This test is meaningful even with size < 2G, as long as the |
| 489 |
# repeated string is > 2G (but it tests more if both are > 2G :) |
| 490 |
@bigmemtest(size=_1G + 2, memuse=3) |
| 491 |
def test_repeat(self, size): |
| 492 |
_ = self.from_latin1 |
| 493 |
s = _('.') * size |
| 494 |
self.assertEqual(len(s), size) |
| 495 |
s = s * 2 |
| 496 |
self.assertEqual(len(s), size * 2) |
| 497 |
self.assertEqual(s.count(_('.')), size * 2) |
| 498 |
|
| 499 |
@bigmemtest(size=_2G + 20, memuse=2) |
| 500 |
def test_slice_and_getitem(self, size): |
| 501 |
_ = self.from_latin1 |
| 502 |
SUBSTR = _('0123456789') |
| 503 |
sublen = len(SUBSTR) |
| 504 |
s = SUBSTR * (size // sublen) |
| 505 |
stepsize = len(s) // 100 |
| 506 |
stepsize = stepsize - (stepsize % sublen) |
| 507 |
for i in range(0, len(s) - stepsize, stepsize): |
| 508 |
self.assertEqual(s[i], SUBSTR[0]) |
| 509 |
self.assertEqual(s[i:i + sublen], SUBSTR) |
| 510 |
self.assertEqual(s[i:i + sublen:2], SUBSTR[::2]) |
| 511 |
if i > 0: |
| 512 |
self.assertEqual(s[i + sublen - 1:i - 1:-3], |
| 513 |
SUBSTR[sublen::-3]) |
| 514 |
# Make sure we do some slicing and indexing near the end of the |
| 515 |
# string, too. |
| 516 |
self.assertEqual(s[len(s) - 1], SUBSTR[-1]) |
| 517 |
self.assertEqual(s[-1], SUBSTR[-1]) |
| 518 |
self.assertEqual(s[len(s) - 10], SUBSTR[0]) |
| 519 |
self.assertEqual(s[-sublen], SUBSTR[0]) |
| 520 |
self.assertEqual(s[len(s):], _('')) |
| 521 |
self.assertEqual(s[len(s) - 1:], SUBSTR[-1:]) |
| 522 |
self.assertEqual(s[-1:], SUBSTR[-1:]) |
| 523 |
self.assertEqual(s[len(s) - sublen:], SUBSTR) |
| 524 |
self.assertEqual(s[-sublen:], SUBSTR) |
| 525 |
self.assertEqual(len(s[:]), len(s)) |
| 526 |
self.assertEqual(len(s[:len(s) - 5]), len(s) - 5) |
| 527 |
self.assertEqual(len(s[5:-5]), len(s) - 10) |
| 528 |
|
| 529 |
self.assertRaises(IndexError, operator.getitem, s, len(s)) |
| 530 |
self.assertRaises(IndexError, operator.getitem, s, len(s) + 1) |
| 531 |
self.assertRaises(IndexError, operator.getitem, s, len(s) + 1<<31) |
| 532 |
|
| 533 |
@bigmemtest(size=_2G, memuse=2) |
| 534 |
def test_contains(self, size): |
| 535 |
_ = self.from_latin1 |
| 536 |
SUBSTR = _('0123456789') |
| 537 |
edge = _('-') * (size // 2) |
| 538 |
s = _('').join([edge, SUBSTR, edge]) |
| 539 |
del edge |
| 540 |
self.assertTrue(SUBSTR in s) |
| 541 |
self.assertFalse(SUBSTR * 2 in s) |
| 542 |
self.assertTrue(_('-') in s) |
| 543 |
self.assertFalse(_('a') in s) |
| 544 |
s += _('a') |
| 545 |
self.assertTrue(_('a') in s) |
| 546 |
|
| 547 |
@bigmemtest(size=_2G + 10, memuse=2) |
| 548 |
def test_compare(self, size): |
| 549 |
_ = self.from_latin1 |
| 550 |
s1 = _('-') * size |
| 551 |
s2 = _('-') * size |
| 552 |
self.assertTrue(s1 == s2) |
| 553 |
del s2 |
| 554 |
s2 = s1 + _('a') |
| 555 |
self.assertFalse(s1 == s2) |
| 556 |
del s2 |
| 557 |
s2 = _('.') * size |
| 558 |
self.assertFalse(s1 == s2) |
| 559 |
|
| 560 |
@bigmemtest(size=_2G + 10, memuse=1) |
| 561 |
def test_hash(self, size): |
| 562 |
# Not sure if we can do any meaningful tests here... Even if we |
| 563 |
# start relying on the exact algorithm used, the result will be |
| 564 |
# different depending on the size of the C 'long int'. Even this |
| 565 |
# test is dodgy (there's no *guarantee* that the two things should |
| 566 |
# have a different hash, even if they, in the current |
| 567 |
# implementation, almost always do.) |
| 568 |
_ = self.from_latin1 |
| 569 |
s = _('\x00') * size |
| 570 |
h1 = hash(s) |
| 571 |
del s |
| 572 |
s = _('\x00') * (size + 1) |
| 573 |
self.assertNotEqual(h1, hash(s)) |
| 574 |
|
| 575 |
|
| 576 |
class StrTest(unittest.TestCase, BaseStrTest): |
| 577 |
|
| 578 |
def from_latin1(self, s): |
| 579 |
return s |
| 580 |
|
| 581 |
def basic_encode_test(self, size, enc, c='.', expectedsize=None): |
| 582 |
if expectedsize is None: |
| 583 |
expectedsize = size |
| 584 |
try: |
| 585 |
s = c * size |
| 586 |
self.assertEqual(len(s.encode(enc)), expectedsize) |
| 587 |
finally: |
| 588 |
s = None |
| 589 |
|
| 590 |
def setUp(self): |
| 591 |
# HACK: adjust memory use of tests inherited from BaseStrTest |
| 592 |
# according to character size. |
| 593 |
self._adjusted = {} |
| 594 |
for name in dir(BaseStrTest): |
| 595 |
if not name.startswith('test_'): |
| 596 |
continue |
| 597 |
meth = getattr(type(self), name) |
| 598 |
try: |
| 599 |
memuse = meth.memuse |
| 600 |
except AttributeError: |
| 601 |
continue |
| 602 |
meth.memuse = ascii_char_size * memuse |
| 603 |
self._adjusted[name] = memuse |
| 604 |
|
| 605 |
def tearDown(self): |
| 606 |
for name, memuse in self._adjusted.items(): |
| 607 |
getattr(type(self), name).memuse = memuse |
| 608 |
|
| 609 |
@bigmemtest(size=_2G, memuse=ucs4_char_size * 3 + ascii_char_size * 2) |
| 610 |
def test_capitalize(self, size): |
| 611 |
self._test_capitalize(size) |
| 612 |
|
| 613 |
@bigmemtest(size=_2G, memuse=ucs4_char_size * 3 + ascii_char_size * 2) |
| 614 |
def test_title(self, size): |
| 615 |
self._test_title(size) |
| 616 |
|
| 617 |
@bigmemtest(size=_2G, memuse=ucs4_char_size * 3 + ascii_char_size * 2) |
| 618 |
def test_swapcase(self, size): |
| 619 |
self._test_swapcase(size) |
| 620 |
|
| 621 |
# Many codecs convert to the legacy representation first, explaining |
| 622 |
# why we add 'ucs4_char_size' to the 'memuse' below. |
| 623 |
|
| 624 |
@bigmemtest(size=_2G + 2, memuse=ascii_char_size + 1) |
| 625 |
def test_encode(self, size): |
| 626 |
return self.basic_encode_test(size, 'utf-8') |
| 627 |
|
| 628 |
@bigmemtest(size=_4G // 6 + 2, memuse=ascii_char_size + ucs4_char_size + 1) |
| 629 |
def test_encode_raw_unicode_escape(self, size): |
| 630 |
try: |
| 631 |
return self.basic_encode_test(size, 'raw_unicode_escape') |
| 632 |
except MemoryError: |
| 633 |
pass # acceptable on 32-bit |
| 634 |
|
| 635 |
@bigmemtest(size=_4G // 5 + 70, memuse=ascii_char_size + 8 + 1) |
| 636 |
def test_encode_utf7(self, size): |
| 637 |
try: |
| 638 |
return self.basic_encode_test(size, 'utf7') |
| 639 |
except MemoryError: |
| 640 |
pass # acceptable on 32-bit |
| 641 |
|
| 642 |
@bigmemtest(size=_4G // 4 + 5, memuse=ascii_char_size + ucs4_char_size + 4) |
| 643 |
def test_encode_utf32(self, size): |
| 644 |
try: |
| 645 |
return self.basic_encode_test(size, 'utf32', expectedsize=4 * size + 4) |
| 646 |
except MemoryError: |
| 647 |
pass # acceptable on 32-bit |
| 648 |
|
| 649 |
@bigmemtest(size=_2G - 1, memuse=ascii_char_size + 1) |
| 650 |
def test_encode_ascii(self, size): |
| 651 |
return self.basic_encode_test(size, 'ascii', c='A') |
| 652 |
|
| 653 |
# str % (...) uses a Py_UCS4 intermediate representation |
| 654 |
|
| 655 |
@bigmemtest(size=_2G + 10, memuse=ascii_char_size * 2 + ucs4_char_size) |
| 656 |
def test_format(self, size): |
| 657 |
s = '-' * size |
| 658 |
sf = '%s' % (s,) |
| 659 |
self.assertTrue(s == sf) |
| 660 |
del sf |
| 661 |
sf = '..%s..' % (s,) |
| 662 |
self.assertEqual(len(sf), len(s) + 4) |
| 663 |
self.assertTrue(sf.startswith('..-')) |
| 664 |
self.assertTrue(sf.endswith('-..')) |
| 665 |
del s, sf |
| 666 |
|
| 667 |
size //= 2 |
| 668 |
edge = '-' * size |
| 669 |
s = ''.join([edge, '%s', edge]) |
| 670 |
del edge |
| 671 |
s = s % '...' |
| 672 |
self.assertEqual(len(s), size * 2 + 3) |
| 673 |
self.assertEqual(s.count('.'), 3) |
| 674 |
self.assertEqual(s.count('-'), size * 2) |
| 675 |
|
| 676 |
@bigmemtest(size=_2G + 10, memuse=ascii_char_size * 2) |
| 677 |
def test_repr_small(self, size): |
| 678 |
s = '-' * size |
| 679 |
s = repr(s) |
| 680 |
self.assertEqual(len(s), size + 2) |
| 681 |
self.assertEqual(s[0], "'") |
| 682 |
self.assertEqual(s[-1], "'") |
| 683 |
self.assertEqual(s.count('-'), size) |
| 684 |
del s |
| 685 |
# repr() will create a string four times as large as this 'binary |
| 686 |
# string', but we don't want to allocate much more than twice |
| 687 |
# size in total. (We do extra testing in test_repr_large()) |
| 688 |
size = size // 5 * 2 |
| 689 |
s = '\x00' * size |
| 690 |
s = repr(s) |
| 691 |
self.assertEqual(len(s), size * 4 + 2) |
| 692 |
self.assertEqual(s[0], "'") |
| 693 |
self.assertEqual(s[-1], "'") |
| 694 |
self.assertEqual(s.count('\\'), size) |
| 695 |
self.assertEqual(s.count('0'), size * 2) |
| 696 |
|
| 697 |
@bigmemtest(size=_2G + 10, memuse=ascii_char_size * 5) |
| 698 |
def test_repr_large(self, size): |
| 699 |
s = '\x00' * size |
| 700 |
s = repr(s) |
| 701 |
self.assertEqual(len(s), size * 4 + 2) |
| 702 |
self.assertEqual(s[0], "'") |
| 703 |
self.assertEqual(s[-1], "'") |
| 704 |
self.assertEqual(s.count('\\'), size) |
| 705 |
self.assertEqual(s.count('0'), size * 2) |
| 706 |
|
| 707 |
# ascii() calls encode('ascii', 'backslashreplace'), which itself |
| 708 |
# creates a temporary Py_UNICODE representation in addition to the |
| 709 |
# original (Py_UCS2) one |
| 710 |
# There's also some overallocation when resizing the ascii() result |
| 711 |
# that isn't taken into account here. |
| 712 |
@bigmemtest(size=_2G // 5 + 1, memuse=ucs2_char_size + |
| 713 |
ucs4_char_size + ascii_char_size * 6) |
| 714 |
def test_unicode_repr(self, size): |
| 715 |
# Use an assigned, but not printable code point. |
| 716 |
# It is in the range of the low surrogates \uDC00-\uDFFF. |
| 717 |
char = "\uDCBA" |
| 718 |
s = char * size |
| 719 |
try: |
| 720 |
for f in (repr, ascii): |
| 721 |
r = f(s) |
| 722 |
self.assertEqual(len(r), 2 + (len(f(char)) - 2) * size) |
| 723 |
self.assertTrue(r.endswith(r"\udcba'"), r[-10:]) |
| 724 |
r = None |
| 725 |
finally: |
| 726 |
r = s = None |
| 727 |
|
| 728 |
@bigmemtest(size=_2G // 5 + 1, memuse=ucs4_char_size * 2 + ascii_char_size * 10) |
| 729 |
def test_unicode_repr_wide(self, size): |
| 730 |
char = "\U0001DCBA" |
| 731 |
s = char * size |
| 732 |
try: |
| 733 |
for f in (repr, ascii): |
| 734 |
r = f(s) |
| 735 |
self.assertEqual(len(r), 2 + (len(f(char)) - 2) * size) |
| 736 |
self.assertTrue(r.endswith(r"\U0001dcba'"), r[-12:]) |
| 737 |
r = None |
| 738 |
finally: |
| 739 |
r = s = None |
| 740 |
|
| 741 |
# The original test_translate is overridden here, so as to get the |
| 742 |
# correct size estimate: str.translate() uses an intermediate Py_UCS4 |
| 743 |
# representation. |
| 744 |
|
| 745 |
@bigmemtest(size=_2G, memuse=ascii_char_size * 2 + ucs4_char_size) |
| 746 |
def test_translate(self, size): |
| 747 |
_ = self.from_latin1 |
| 748 |
SUBSTR = _('aZz.z.Aaz.') |
| 749 |
trans = { |
| 750 |
ord(_('.')): _('-'), |
| 751 |
ord(_('a')): _('!'), |
| 752 |
ord(_('Z')): _('$'), |
| 753 |
} |
| 754 |
sublen = len(SUBSTR) |
| 755 |
repeats = size // sublen + 2 |
| 756 |
s = SUBSTR * repeats |
| 757 |
s = s.translate(trans) |
| 758 |
self.assertEqual(len(s), repeats * sublen) |
| 759 |
self.assertEqual(s[:sublen], SUBSTR.translate(trans)) |
| 760 |
self.assertEqual(s[-sublen:], SUBSTR.translate(trans)) |
| 761 |
self.assertEqual(s.count(_('.')), 0) |
| 762 |
self.assertEqual(s.count(_('!')), repeats * 2) |
| 763 |
self.assertEqual(s.count(_('z')), repeats * 3) |
| 764 |
|
| 765 |
|
| 766 |
class BytesTest(unittest.TestCase, BaseStrTest): |
| 767 |
|
| 768 |
def from_latin1(self, s): |
| 769 |
return s.encode("latin-1") |
| 770 |
|
| 771 |
@bigmemtest(size=_2G + 2, memuse=1 + ascii_char_size) |
| 772 |
def test_decode(self, size): |
| 773 |
s = self.from_latin1('.') * size |
| 774 |
self.assertEqual(len(s.decode('utf-8')), size) |
| 775 |
|
| 776 |
@bigmemtest(size=_2G, memuse=2) |
| 777 |
def test_capitalize(self, size): |
| 778 |
self._test_capitalize(size) |
| 779 |
|
| 780 |
@bigmemtest(size=_2G, memuse=2) |
| 781 |
def test_title(self, size): |
| 782 |
self._test_title(size) |
| 783 |
|
| 784 |
@bigmemtest(size=_2G, memuse=2) |
| 785 |
def test_swapcase(self, size): |
| 786 |
self._test_swapcase(size) |
| 787 |
|
| 788 |
|
| 789 |
class BytearrayTest(unittest.TestCase, BaseStrTest): |
| 790 |
|
| 791 |
def from_latin1(self, s): |
| 792 |
return bytearray(s.encode("latin-1")) |
| 793 |
|
| 794 |
@bigmemtest(size=_2G + 2, memuse=1 + ascii_char_size) |
| 795 |
def test_decode(self, size): |
| 796 |
s = self.from_latin1('.') * size |
| 797 |
self.assertEqual(len(s.decode('utf-8')), size) |
| 798 |
|
| 799 |
@bigmemtest(size=_2G, memuse=2) |
| 800 |
def test_capitalize(self, size): |
| 801 |
self._test_capitalize(size) |
| 802 |
|
| 803 |
@bigmemtest(size=_2G, memuse=2) |
| 804 |
def test_title(self, size): |
| 805 |
self._test_title(size) |
| 806 |
|
| 807 |
@bigmemtest(size=_2G, memuse=2) |
| 808 |
def test_swapcase(self, size): |
| 809 |
self._test_swapcase(size) |
| 810 |
|
| 811 |
test_hash = None |
| 812 |
test_split_large = None |
| 813 |
|
| 814 |
class TupleTest(unittest.TestCase): |
| 815 |
|
| 816 |
# Tuples have a small, fixed-sized head and an array of pointers to |
| 817 |
# data. Since we're testing 64-bit addressing, we can assume that the |
| 818 |
# pointers are 8 bytes, and that thus that the tuples take up 8 bytes |
| 819 |
# per size. |
| 820 |
|
| 821 |
# As a side-effect of testing long tuples, these tests happen to test |
| 822 |
# having more than 2<<31 references to any given object. Hence the |
| 823 |
# use of different types of objects as contents in different tests. |
| 824 |
|
| 825 |
@bigmemtest(size=_2G + 2, memuse=pointer_size * 2) |
| 826 |
def test_compare(self, size): |
| 827 |
t1 = ('',) * size |
| 828 |
t2 = ('',) * size |
| 829 |
self.assertTrue(t1 == t2) |
| 830 |
del t2 |
| 831 |
t2 = ('',) * (size + 1) |
| 832 |
self.assertFalse(t1 == t2) |
| 833 |
del t2 |
| 834 |
t2 = (1,) * size |
| 835 |
self.assertFalse(t1 == t2) |
| 836 |
|
| 837 |
# Test concatenating into a single tuple of more than 2G in length, |
| 838 |
# and concatenating a tuple of more than 2G in length separately, so |
| 839 |
# the smaller test still gets run even if there isn't memory for the |
| 840 |
# larger test (but we still let the tester know the larger test is |
| 841 |
# skipped, in verbose mode.) |
| 842 |
def basic_concat_test(self, size): |
| 843 |
t = ((),) * size |
| 844 |
self.assertEqual(len(t), size) |
| 845 |
t = t + t |
| 846 |
self.assertEqual(len(t), size * 2) |
| 847 |
|
| 848 |
@bigmemtest(size=_2G // 2 + 2, memuse=pointer_size * 3) |
| 849 |
def test_concat_small(self, size): |
| 850 |
return self.basic_concat_test(size) |
| 851 |
|
| 852 |
@bigmemtest(size=_2G + 2, memuse=pointer_size * 3) |
| 853 |
def test_concat_large(self, size): |
| 854 |
return self.basic_concat_test(size) |
| 855 |
|
| 856 |
@bigmemtest(size=_2G // 5 + 10, memuse=pointer_size * 5) |
| 857 |
def test_contains(self, size): |
| 858 |
t = (1, 2, 3, 4, 5) * size |
| 859 |
self.assertEqual(len(t), size * 5) |
| 860 |
self.assertTrue(5 in t) |
| 861 |
self.assertFalse((1, 2, 3, 4, 5) in t) |
| 862 |
self.assertFalse(0 in t) |
| 863 |
|
| 864 |
@bigmemtest(size=_2G + 10, memuse=pointer_size) |
| 865 |
def test_hash(self, size): |
| 866 |
t1 = (0,) * size |
| 867 |
h1 = hash(t1) |
| 868 |
del t1 |
| 869 |
t2 = (0,) * (size + 1) |
| 870 |
self.assertFalse(h1 == hash(t2)) |
| 871 |
|
| 872 |
@bigmemtest(size=_2G + 10, memuse=pointer_size) |
| 873 |
def test_index_and_slice(self, size): |
| 874 |
t = (None,) * size |
| 875 |
self.assertEqual(len(t), size) |
| 876 |
self.assertEqual(t[-1], None) |
| 877 |
self.assertEqual(t[5], None) |
| 878 |
self.assertEqual(t[size - 1], None) |
| 879 |
self.assertRaises(IndexError, operator.getitem, t, size) |
| 880 |
self.assertEqual(t[:5], (None,) * 5) |
| 881 |
self.assertEqual(t[-5:], (None,) * 5) |
| 882 |
self.assertEqual(t[20:25], (None,) * 5) |
| 883 |
self.assertEqual(t[-25:-20], (None,) * 5) |
| 884 |
self.assertEqual(t[size - 5:], (None,) * 5) |
| 885 |
self.assertEqual(t[size - 5:size], (None,) * 5) |
| 886 |
self.assertEqual(t[size - 6:size - 2], (None,) * 4) |
| 887 |
self.assertEqual(t[size:size], ()) |
| 888 |
self.assertEqual(t[size:size+5], ()) |
| 889 |
|
| 890 |
# Like test_concat, split in two. |
| 891 |
def basic_test_repeat(self, size): |
| 892 |
t = ('',) * size |
| 893 |
self.assertEqual(len(t), size) |
| 894 |
t = t * 2 |
| 895 |
self.assertEqual(len(t), size * 2) |
| 896 |
|
| 897 |
@bigmemtest(size=_2G // 2 + 2, memuse=pointer_size * 3) |
| 898 |
def test_repeat_small(self, size): |
| 899 |
return self.basic_test_repeat(size) |
| 900 |
|
| 901 |
@bigmemtest(size=_2G + 2, memuse=pointer_size * 3) |
| 902 |
def test_repeat_large(self, size): |
| 903 |
return self.basic_test_repeat(size) |
| 904 |
|
| 905 |
@bigmemtest(size=_1G - 1, memuse=12) |
| 906 |
def test_repeat_large_2(self, size): |
| 907 |
return self.basic_test_repeat(size) |
| 908 |
|
| 909 |
@bigmemtest(size=_1G - 1, memuse=pointer_size * 2) |
| 910 |
def test_from_2G_generator(self, size): |
| 911 |
try: |
| 912 |
t = tuple(iter([42]*size)) |
| 913 |
except MemoryError: |
| 914 |
pass # acceptable on 32-bit |
| 915 |
else: |
| 916 |
self.assertEqual(len(t), size) |
| 917 |
self.assertEqual(t[:10], (42,) * 10) |
| 918 |
self.assertEqual(t[-10:], (42,) * 10) |
| 919 |
|
| 920 |
@bigmemtest(size=_1G - 25, memuse=pointer_size * 2) |
| 921 |
def test_from_almost_2G_generator(self, size): |
| 922 |
try: |
| 923 |
t = tuple(iter([42]*size)) |
| 924 |
except MemoryError: |
| 925 |
pass # acceptable on 32-bit |
| 926 |
else: |
| 927 |
self.assertEqual(len(t), size) |
| 928 |
self.assertEqual(t[:10], (42,) * 10) |
| 929 |
self.assertEqual(t[-10:], (42,) * 10) |
| 930 |
|
| 931 |
# Like test_concat, split in two. |
| 932 |
def basic_test_repr(self, size): |
| 933 |
t = (False,) * size |
| 934 |
s = repr(t) |
| 935 |
# The repr of a tuple of Falses is exactly 7 times the tuple length. |
| 936 |
self.assertEqual(len(s), size * 7) |
| 937 |
self.assertEqual(s[:10], '(False, Fa') |
| 938 |
self.assertEqual(s[-10:], 'se, False)') |
| 939 |
|
| 940 |
@bigmemtest(size=_2G // 7 + 2, memuse=pointer_size + ascii_char_size * 7) |
| 941 |
def test_repr_small(self, size): |
| 942 |
return self.basic_test_repr(size) |
| 943 |
|
| 944 |
@bigmemtest(size=_2G + 2, memuse=pointer_size + ascii_char_size * 7) |
| 945 |
def test_repr_large(self, size): |
| 946 |
return self.basic_test_repr(size) |
| 947 |
|
| 948 |
class ListTest(unittest.TestCase): |
| 949 |
|
| 950 |
# Like tuples, lists have a small, fixed-sized head and an array of |
| 951 |
# pointers to data, so 8 bytes per size. Also like tuples, we make the |
| 952 |
# lists hold references to various objects to test their refcount |
| 953 |
# limits. |
| 954 |
|
| 955 |
@bigmemtest(size=_2G + 2, memuse=pointer_size * 2) |
| 956 |
def test_compare(self, size): |
| 957 |
l1 = [''] * size |
| 958 |
l2 = [''] * size |
| 959 |
self.assertTrue(l1 == l2) |
| 960 |
del l2 |
| 961 |
l2 = [''] * (size + 1) |
| 962 |
self.assertFalse(l1 == l2) |
| 963 |
del l2 |
| 964 |
l2 = [2] * size |
| 965 |
self.assertFalse(l1 == l2) |
| 966 |
|
| 967 |
# Test concatenating into a single list of more than 2G in length, |
| 968 |
# and concatenating a list of more than 2G in length separately, so |
| 969 |
# the smaller test still gets run even if there isn't memory for the |
| 970 |
# larger test (but we still let the tester know the larger test is |
| 971 |
# skipped, in verbose mode.) |
| 972 |
def basic_test_concat(self, size): |
| 973 |
l = [[]] * size |
| 974 |
self.assertEqual(len(l), size) |
| 975 |
l = l + l |
| 976 |
self.assertEqual(len(l), size * 2) |
| 977 |
|
| 978 |
@bigmemtest(size=_2G // 2 + 2, memuse=pointer_size * 3) |
| 979 |
def test_concat_small(self, size): |
| 980 |
return self.basic_test_concat(size) |
| 981 |
|
| 982 |
@bigmemtest(size=_2G + 2, memuse=pointer_size * 3) |
| 983 |
def test_concat_large(self, size): |
| 984 |
return self.basic_test_concat(size) |
| 985 |
|
| 986 |
# XXX This tests suffers from overallocation, just like test_append. |
| 987 |
# This should be fixed in future. |
| 988 |
def basic_test_inplace_concat(self, size): |
| 989 |
l = [sys.stdout] * size |
| 990 |
l += l |
| 991 |
self.assertEqual(len(l), size * 2) |
| 992 |
self.assertTrue(l[0] is l[-1]) |
| 993 |
self.assertTrue(l[size - 1] is l[size + 1]) |
| 994 |
|
| 995 |
@bigmemtest(size=_2G // 2 + 2, memuse=pointer_size * 2 * 9/8) |
| 996 |
def test_inplace_concat_small(self, size): |
| 997 |
return self.basic_test_inplace_concat(size) |
| 998 |
|
| 999 |
@bigmemtest(size=_2G + 2, memuse=pointer_size * 2 * 9/8) |
| 1000 |
def test_inplace_concat_large(self, size): |
| 1001 |
return self.basic_test_inplace_concat(size) |
| 1002 |
|
| 1003 |
@bigmemtest(size=_2G // 5 + 10, memuse=pointer_size * 5) |
| 1004 |
def test_contains(self, size): |
| 1005 |
l = [1, 2, 3, 4, 5] * size |
| 1006 |
self.assertEqual(len(l), size * 5) |
| 1007 |
self.assertTrue(5 in l) |
| 1008 |
self.assertFalse([1, 2, 3, 4, 5] in l) |
| 1009 |
self.assertFalse(0 in l) |
| 1010 |
|
| 1011 |
@bigmemtest(size=_2G + 10, memuse=pointer_size) |
| 1012 |
def test_hash(self, size): |
| 1013 |
l = [0] * size |
| 1014 |
self.assertRaises(TypeError, hash, l) |
| 1015 |
|
| 1016 |
@bigmemtest(size=_2G + 10, memuse=pointer_size) |
| 1017 |
def test_index_and_slice(self, size): |
| 1018 |
l = [None] * size |
| 1019 |
self.assertEqual(len(l), size) |
| 1020 |
self.assertEqual(l[-1], None) |
| 1021 |
self.assertEqual(l[5], None) |
| 1022 |
self.assertEqual(l[size - 1], None) |
| 1023 |
self.assertRaises(IndexError, operator.getitem, l, size) |
| 1024 |
self.assertEqual(l[:5], [None] * 5) |
| 1025 |
self.assertEqual(l[-5:], [None] * 5) |
| 1026 |
self.assertEqual(l[20:25], [None] * 5) |
| 1027 |
self.assertEqual(l[-25:-20], [None] * 5) |
| 1028 |
self.assertEqual(l[size - 5:], [None] * 5) |
| 1029 |
self.assertEqual(l[size - 5:size], [None] * 5) |
| 1030 |
self.assertEqual(l[size - 6:size - 2], [None] * 4) |
| 1031 |
self.assertEqual(l[size:size], []) |
| 1032 |
self.assertEqual(l[size:size+5], []) |
| 1033 |
|
| 1034 |
l[size - 2] = 5 |
| 1035 |
self.assertEqual(len(l), size) |
| 1036 |
self.assertEqual(l[-3:], [None, 5, None]) |
| 1037 |
self.assertEqual(l.count(5), 1) |
| 1038 |
self.assertRaises(IndexError, operator.setitem, l, size, 6) |
| 1039 |
self.assertEqual(len(l), size) |
| 1040 |
|
| 1041 |
l[size - 7:] = [1, 2, 3, 4, 5] |
| 1042 |
size -= 2 |
| 1043 |
self.assertEqual(len(l), size) |
| 1044 |
self.assertEqual(l[-7:], [None, None, 1, 2, 3, 4, 5]) |
| 1045 |
|
| 1046 |
l[:7] = [1, 2, 3, 4, 5] |
| 1047 |
size -= 2 |
| 1048 |
self.assertEqual(len(l), size) |
| 1049 |
self.assertEqual(l[:7], [1, 2, 3, 4, 5, None, None]) |
| 1050 |
|
| 1051 |
del l[size - 1] |
| 1052 |
size -= 1 |
| 1053 |
self.assertEqual(len(l), size) |
| 1054 |
self.assertEqual(l[-1], 4) |
| 1055 |
|
| 1056 |
del l[-2:] |
| 1057 |
size -= 2 |
| 1058 |
self.assertEqual(len(l), size) |
| 1059 |
self.assertEqual(l[-1], 2) |
| 1060 |
|
| 1061 |
del l[0] |
| 1062 |
size -= 1 |
| 1063 |
self.assertEqual(len(l), size) |
| 1064 |
self.assertEqual(l[0], 2) |
| 1065 |
|
| 1066 |
del l[:2] |
| 1067 |
size -= 2 |
| 1068 |
self.assertEqual(len(l), size) |
| 1069 |
self.assertEqual(l[0], 4) |
| 1070 |
|
| 1071 |
# Like test_concat, split in two. |
| 1072 |
def basic_test_repeat(self, size): |
| 1073 |
l = [] * size |
| 1074 |
self.assertFalse(l) |
| 1075 |
l = [''] * size |
| 1076 |
self.assertEqual(len(l), size) |
| 1077 |
l = l * 2 |
| 1078 |
self.assertEqual(len(l), size * 2) |
| 1079 |
|
| 1080 |
@bigmemtest(size=_2G // 2 + 2, memuse=pointer_size * 3) |
| 1081 |
def test_repeat_small(self, size): |
| 1082 |
return self.basic_test_repeat(size) |
| 1083 |
|
| 1084 |
@bigmemtest(size=_2G + 2, memuse=pointer_size * 3) |
| 1085 |
def test_repeat_large(self, size): |
| 1086 |
return self.basic_test_repeat(size) |
| 1087 |
|
| 1088 |
# XXX This tests suffers from overallocation, just like test_append. |
| 1089 |
# This should be fixed in future. |
| 1090 |
def basic_test_inplace_repeat(self, size): |
| 1091 |
l = [''] |
| 1092 |
l *= size |
| 1093 |
self.assertEqual(len(l), size) |
| 1094 |
self.assertTrue(l[0] is l[-1]) |
| 1095 |
del l |
| 1096 |
|
| 1097 |
l = [''] * size |
| 1098 |
l *= 2 |
| 1099 |
self.assertEqual(len(l), size * 2) |
| 1100 |
self.assertTrue(l[size - 1] is l[-1]) |
| 1101 |
|
| 1102 |
@bigmemtest(size=_2G // 2 + 2, memuse=pointer_size * 2 * 9/8) |
| 1103 |
def test_inplace_repeat_small(self, size): |
| 1104 |
return self.basic_test_inplace_repeat(size) |
| 1105 |
|
| 1106 |
@bigmemtest(size=_2G + 2, memuse=pointer_size * 2 * 9/8) |
| 1107 |
def test_inplace_repeat_large(self, size): |
| 1108 |
return self.basic_test_inplace_repeat(size) |
| 1109 |
|
| 1110 |
def basic_test_repr(self, size): |
| 1111 |
l = [False] * size |
| 1112 |
s = repr(l) |
| 1113 |
# The repr of a list of Falses is exactly 7 times the list length. |
| 1114 |
self.assertEqual(len(s), size * 7) |
| 1115 |
self.assertEqual(s[:10], '[False, Fa') |
| 1116 |
self.assertEqual(s[-10:], 'se, False]') |
| 1117 |
self.assertEqual(s.count('F'), size) |
| 1118 |
|
| 1119 |
@bigmemtest(size=_2G // 7 + 2, memuse=pointer_size + ascii_char_size * 7) |
| 1120 |
def test_repr_small(self, size): |
| 1121 |
return self.basic_test_repr(size) |
| 1122 |
|
| 1123 |
@bigmemtest(size=_2G + 2, memuse=pointer_size + ascii_char_size * 7) |
| 1124 |
def test_repr_large(self, size): |
| 1125 |
return self.basic_test_repr(size) |
| 1126 |
|
| 1127 |
# list overallocates ~1/8th of the total size (on first expansion) so |
| 1128 |
# the single list.append call puts memuse at 9 bytes per size. |
| 1129 |
@bigmemtest(size=_2G, memuse=pointer_size * 9/8) |
| 1130 |
def test_append(self, size): |
| 1131 |
l = [object()] * size |
| 1132 |
l.append(object()) |
| 1133 |
self.assertEqual(len(l), size+1) |
| 1134 |
self.assertTrue(l[-3] is l[-2]) |
| 1135 |
self.assertFalse(l[-2] is l[-1]) |
| 1136 |
|
| 1137 |
@bigmemtest(size=_2G // 5 + 2, memuse=pointer_size * 5) |
| 1138 |
def test_count(self, size): |
| 1139 |
l = [1, 2, 3, 4, 5] * size |
| 1140 |
self.assertEqual(l.count(1), size) |
| 1141 |
self.assertEqual(l.count("1"), 0) |
| 1142 |
|
| 1143 |
# XXX This tests suffers from overallocation, just like test_append. |
| 1144 |
# This should be fixed in future. |
| 1145 |
def basic_test_extend(self, size): |
| 1146 |
l = [object] * size |
| 1147 |
l.extend(l) |
| 1148 |
self.assertEqual(len(l), size * 2) |
| 1149 |
self.assertTrue(l[0] is l[-1]) |
| 1150 |
self.assertTrue(l[size - 1] is l[size + 1]) |
| 1151 |
|
| 1152 |
@bigmemtest(size=_2G // 2 + 2, memuse=pointer_size * 2 * 9/8) |
| 1153 |
def test_extend_small(self, size): |
| 1154 |
return self.basic_test_extend(size) |
| 1155 |
|
| 1156 |
@bigmemtest(size=_2G + 2, memuse=pointer_size * 2 * 9/8) |
| 1157 |
def test_extend_large(self, size): |
| 1158 |
return self.basic_test_extend(size) |
| 1159 |
|
| 1160 |
@bigmemtest(size=_2G // 5 + 2, memuse=pointer_size * 5) |
| 1161 |
def test_index(self, size): |
| 1162 |
l = [1, 2, 3, 4, 5] * size |
| 1163 |
size *= 5 |
| 1164 |
self.assertEqual(l.index(1), 0) |
| 1165 |
self.assertEqual(l.index(5, size - 5), size - 1) |
| 1166 |
self.assertEqual(l.index(5, size - 5, size), size - 1) |
| 1167 |
self.assertRaises(ValueError, l.index, 1, size - 4, size) |
| 1168 |
self.assertRaises(ValueError, l.index, 6) |
| 1169 |
|
| 1170 |
# This tests suffers from overallocation, just like test_append. |
| 1171 |
@bigmemtest(size=_2G + 10, memuse=pointer_size * 9/8) |
| 1172 |
def test_insert(self, size): |
| 1173 |
l = [1.0] * size |
| 1174 |
l.insert(size - 1, "A") |
| 1175 |
size += 1 |
| 1176 |
self.assertEqual(len(l), size) |
| 1177 |
self.assertEqual(l[-3:], [1.0, "A", 1.0]) |
| 1178 |
|
| 1179 |
l.insert(size + 1, "B") |
| 1180 |
size += 1 |
| 1181 |
self.assertEqual(len(l), size) |
| 1182 |
self.assertEqual(l[-3:], ["A", 1.0, "B"]) |
| 1183 |
|
| 1184 |
l.insert(1, "C") |
| 1185 |
size += 1 |
| 1186 |
self.assertEqual(len(l), size) |
| 1187 |
self.assertEqual(l[:3], [1.0, "C", 1.0]) |
| 1188 |
self.assertEqual(l[size - 3:], ["A", 1.0, "B"]) |
| 1189 |
|
| 1190 |
@bigmemtest(size=_2G // 5 + 4, memuse=pointer_size * 5) |
| 1191 |
def test_pop(self, size): |
| 1192 |
l = ["a", "b", "c", "d", "e"] * size |
| 1193 |
size *= 5 |
| 1194 |
self.assertEqual(len(l), size) |
| 1195 |
|
| 1196 |
item = l.pop() |
| 1197 |
size -= 1 |
| 1198 |
self.assertEqual(len(l), size) |
| 1199 |
self.assertEqual(item, "e") |
| 1200 |
self.assertEqual(l[-2:], ["c", "d"]) |
| 1201 |
|
| 1202 |
item = l.pop(0) |
| 1203 |
size -= 1 |
| 1204 |
self.assertEqual(len(l), size) |
| 1205 |
self.assertEqual(item, "a") |
| 1206 |
self.assertEqual(l[:2], ["b", "c"]) |
| 1207 |
|
| 1208 |
item = l.pop(size - 2) |
| 1209 |
size -= 1 |
| 1210 |
self.assertEqual(len(l), size) |
| 1211 |
self.assertEqual(item, "c") |
| 1212 |
self.assertEqual(l[-2:], ["b", "d"]) |
| 1213 |
|
| 1214 |
@bigmemtest(size=_2G + 10, memuse=pointer_size) |
| 1215 |
def test_remove(self, size): |
| 1216 |
l = [10] * size |
| 1217 |
self.assertEqual(len(l), size) |
| 1218 |
|
| 1219 |
l.remove(10) |
| 1220 |
size -= 1 |
| 1221 |
self.assertEqual(len(l), size) |
| 1222 |
|
| 1223 |
# Because of the earlier l.remove(), this append doesn't trigger |
| 1224 |
# a resize. |
| 1225 |
l.append(5) |
| 1226 |
size += 1 |
| 1227 |
self.assertEqual(len(l), size) |
| 1228 |
self.assertEqual(l[-2:], [10, 5]) |
| 1229 |
l.remove(5) |
| 1230 |
size -= 1 |
| 1231 |
self.assertEqual(len(l), size) |
| 1232 |
self.assertEqual(l[-2:], [10, 10]) |
| 1233 |
|
| 1234 |
@bigmemtest(size=_2G // 5 + 2, memuse=pointer_size * 5) |
| 1235 |
def test_reverse(self, size): |
| 1236 |
l = [1, 2, 3, 4, 5] * size |
| 1237 |
l.reverse() |
| 1238 |
self.assertEqual(len(l), size * 5) |
| 1239 |
self.assertEqual(l[-5:], [5, 4, 3, 2, 1]) |
| 1240 |
self.assertEqual(l[:5], [5, 4, 3, 2, 1]) |
| 1241 |
|
| 1242 |
@bigmemtest(size=_2G // 5 + 2, memuse=pointer_size * 5 * 1.5) |
| 1243 |
def test_sort(self, size): |
| 1244 |
l = [1, 2, 3, 4, 5] * size |
| 1245 |
l.sort() |
| 1246 |
self.assertEqual(len(l), size * 5) |
| 1247 |
self.assertEqual(l.count(1), size) |
| 1248 |
self.assertEqual(l[:10], [1] * 10) |
| 1249 |
self.assertEqual(l[-10:], [5] * 10) |
| 1250 |
|
| 1251 |
|
| 1252 |
class DictTest(unittest.TestCase): |
| 1253 |
|
| 1254 |
@bigmemtest(size=357913941, memuse=160) |
| 1255 |
def test_dict(self, size): |
| 1256 |
# https://github.com/python/cpython/issues/102701 |
| 1257 |
d = dict.fromkeys(range(size)) |
| 1258 |
d[size] = 1 |
| 1259 |
|
| 1260 |
|
| 1261 |
class ImmortalityTest(unittest.TestCase): |
| 1262 |
|
| 1263 |
@bigmemtest(size=_2G, memuse=pointer_size * 9/8) |
| 1264 |
def test_stickiness(self, size): |
| 1265 |
"""Check that immortality is "sticky", so that |
| 1266 |
once an object is immortal it remains so.""" |
| 1267 |
if size < _2G: |
| 1268 |
# Not enough memory to cause immortality on overflow |
| 1269 |
return |
| 1270 |
o1 = o2 = o3 = o4 = o5 = o6 = o7 = o8 = object() |
| 1271 |
l = [o1] * (size-20) |
| 1272 |
self.assertFalse(_testcapi.is_immortal(o1)) |
| 1273 |
for _ in range(30): |
| 1274 |
l.append(l[0]) |
| 1275 |
self.assertTrue(_testcapi.is_immortal(o1)) |
| 1276 |
del o2, o3, o4, o5, o6, o7, o8 |
| 1277 |
self.assertTrue(_testcapi.is_immortal(o1)) |
| 1278 |
del l |
| 1279 |
self.assertTrue(_testcapi.is_immortal(o1)) |
| 1280 |
|
| 1281 |
|
| 1282 |
if __name__ == '__main__': |
| 1283 |
if len(sys.argv) > 1: |
| 1284 |
support.set_memlimit(sys.argv[1]) |
| 1285 |
unittest.main() |