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