blob: 43fa5d17fd2e79be2469c257b46ddf53d7bfd6a0 [file] [log] [blame]
Raymond Hettingerc991db22005-08-11 07:58:45 +00001
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002/* set object implementation
Raymond Hettingerae7b00e2013-09-07 15:05:00 -07003
Raymond Hettingera9d99362005-08-05 00:01:15 +00004 Written and maintained by Raymond D. Hettinger <python@rcn.com>
5 Derived from Lib/sets.py and Objects/dictobject.c.
Raymond Hettinger9f1a6792005-07-31 01:16:36 +00006
Raymond Hettingerae7b00e2013-09-07 15:05:00 -07007 The basic lookup function used by all operations.
8 This is based on Algorithm D from Knuth Vol. 3, Sec. 6.4.
9
10 The initial probe index is computed as hash mod the table size.
11 Subsequent probe indices are computed as explained in Objects/dictobject.c.
12
13 To improve cache locality, each probe inspects a series of consecutive
14 nearby entries before moving on to probes elsewhere in memory. This leaves
Raymond Hettinger64263df2017-09-04 18:54:16 -070015 us with a hybrid of linear probing and randomized probing. The linear probing
Raymond Hettingerae7b00e2013-09-07 15:05:00 -070016 reduces the cost of hash collisions because consecutive memory accesses
17 tend to be much cheaper than scattered probes. After LINEAR_PROBES steps,
Raymond Hettinger64263df2017-09-04 18:54:16 -070018 we then use more of the upper bits from the hash value and apply a simple
19 linear congruential random number genearator. This helps break-up long
20 chains of collisions.
Raymond Hettingerae7b00e2013-09-07 15:05:00 -070021
22 All arithmetic on hash should ignore overflow.
23
Raymond Hettinger4d45c102015-01-25 16:27:40 -080024 Unlike the dictionary implementation, the lookkey function can return
Raymond Hettingerae7b00e2013-09-07 15:05:00 -070025 NULL if the rich comparison returns an error.
Serhiy Storchaka13ad3b72017-09-14 09:38:36 +030026
Raymond Hettinger64263df2017-09-04 18:54:16 -070027 Use cases for sets differ considerably from dictionaries where looked-up
28 keys are more likely to be present. In contrast, sets are primarily
29 about membership testing where the presence of an element is not known in
30 advance. Accordingly, the set implementation needs to optimize for both
31 the found and not-found case.
Raymond Hettinger9f1a6792005-07-31 01:16:36 +000032*/
33
Raymond Hettingera690a992003-11-16 16:17:49 +000034#include "Python.h"
Victor Stinnerbcda8f12018-11-21 22:27:47 +010035#include "pycore_object.h"
Victor Stinner621cebe2018-11-12 16:53:38 +010036#include "pycore_pystate.h"
Raymond Hettingera9d99362005-08-05 00:01:15 +000037#include "structmember.h"
Raymond Hettinger9f1a6792005-07-31 01:16:36 +000038
Raymond Hettinger9f1a6792005-07-31 01:16:36 +000039/* Object used as dummy key to fill deleted entries */
Raymond Hettingerbfc1e1a2013-08-23 03:22:15 -050040static PyObject _dummy_struct;
41
42#define dummy (&_dummy_struct)
Raymond Hettinger9f1a6792005-07-31 01:16:36 +000043
Raymond Hettinger9f1a6792005-07-31 01:16:36 +000044
Raymond Hettinger04fd9dd2013-09-07 17:41:01 -070045/* ======================================================================== */
46/* ======= Begin logic for probing the hash table ========================= */
47
Raymond Hettinger710a67e2013-09-21 20:17:31 -070048/* Set this to zero to turn-off linear probing */
Raymond Hettinger8408dc52013-09-15 14:57:15 -070049#ifndef LINEAR_PROBES
Raymond Hettinger742d8712013-09-08 00:25:57 -070050#define LINEAR_PROBES 9
Raymond Hettinger8408dc52013-09-15 14:57:15 -070051#endif
Raymond Hettinger742d8712013-09-08 00:25:57 -070052
53/* This must be >= 1 */
54#define PERTURB_SHIFT 5
55
Raymond Hettinger9f1a6792005-07-31 01:16:36 +000056static setentry *
Antoine Pitrou9ed5f272013-08-13 20:18:52 +020057set_lookkey(PySetObject *so, PyObject *key, Py_hash_t hash)
Raymond Hettinger9f1a6792005-07-31 01:16:36 +000058{
Raymond Hettinger70559b52015-07-23 07:42:23 -040059 setentry *table;
Antoine Pitrou9ed5f272013-08-13 20:18:52 +020060 setentry *entry;
Raymond Hettinger38bb95e2015-06-24 01:22:19 -070061 size_t perturb;
Raymond Hettingerafe89092013-08-28 20:59:31 -070062 size_t mask = so->mask;
Raymond Hettinger59ecabd2015-01-31 02:45:12 -080063 size_t i = (size_t)hash & mask; /* Unsigned for defined overflow behavior */
Raymond Hettinger8408dc52013-09-15 14:57:15 -070064 size_t j;
Raymond Hettinger4ef05282013-09-21 15:39:49 -070065 int cmp;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +000066
Raymond Hettinger70559b52015-07-23 07:42:23 -040067 entry = &so->table[i];
Raymond Hettingerafe89092013-08-28 20:59:31 -070068 if (entry->key == NULL)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +000069 return entry;
Raymond Hettinger237b34b2013-08-17 02:31:53 -070070
Raymond Hettinger38bb95e2015-06-24 01:22:19 -070071 perturb = hash;
72
Raymond Hettinger3c0a4f52013-08-19 07:36:04 -070073 while (1) {
Raymond Hettingera5ebbf62015-01-26 21:54:35 -080074 if (entry->hash == hash) {
Raymond Hettingerafe89092013-08-28 20:59:31 -070075 PyObject *startkey = entry->key;
Raymond Hettingera5ebbf62015-01-26 21:54:35 -080076 /* startkey cannot be a dummy because the dummy hash field is -1 */
77 assert(startkey != dummy);
Raymond Hettinger08e3dc02014-12-26 20:14:00 -080078 if (startkey == key)
79 return entry;
Raymond Hettinger93035c42015-01-25 16:12:49 -080080 if (PyUnicode_CheckExact(startkey)
81 && PyUnicode_CheckExact(key)
Raymond Hettingerac2ef652015-07-04 16:04:44 -070082 && _PyUnicode_EQ(startkey, key))
Raymond Hettinger93035c42015-01-25 16:12:49 -080083 return entry;
Raymond Hettinger70559b52015-07-23 07:42:23 -040084 table = so->table;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +000085 Py_INCREF(startkey);
86 cmp = PyObject_RichCompareBool(startkey, key, Py_EQ);
87 Py_DECREF(startkey);
Raymond Hettinger426d9952014-05-18 21:40:20 +010088 if (cmp < 0) /* unlikely */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +000089 return NULL;
Raymond Hettinger426d9952014-05-18 21:40:20 +010090 if (table != so->table || entry->key != startkey) /* unlikely */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +000091 return set_lookkey(so, key, hash);
Raymond Hettinger426d9952014-05-18 21:40:20 +010092 if (cmp > 0) /* likely */
Raymond Hettingerafe89092013-08-28 20:59:31 -070093 return entry;
Raymond Hettinger438f9132015-02-09 06:48:29 -060094 mask = so->mask; /* help avoid a register spill */
Raymond Hettingerafe89092013-08-28 20:59:31 -070095 }
Raymond Hettingerafe89092013-08-28 20:59:31 -070096
Raymond Hettingerc658d852015-02-02 08:35:00 -080097 if (i + LINEAR_PROBES <= mask) {
Raymond Hettinger82492822015-02-04 08:37:02 -080098 for (j = 0 ; j < LINEAR_PROBES ; j++) {
Raymond Hettingerc658d852015-02-02 08:35:00 -080099 entry++;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700100 if (entry->hash == 0 && entry->key == NULL)
101 return entry;
Raymond Hettingerc658d852015-02-02 08:35:00 -0800102 if (entry->hash == hash) {
103 PyObject *startkey = entry->key;
104 assert(startkey != dummy);
105 if (startkey == key)
106 return entry;
107 if (PyUnicode_CheckExact(startkey)
108 && PyUnicode_CheckExact(key)
Raymond Hettingerac2ef652015-07-04 16:04:44 -0700109 && _PyUnicode_EQ(startkey, key))
Raymond Hettingerc658d852015-02-02 08:35:00 -0800110 return entry;
Raymond Hettinger70559b52015-07-23 07:42:23 -0400111 table = so->table;
Raymond Hettingerc658d852015-02-02 08:35:00 -0800112 Py_INCREF(startkey);
113 cmp = PyObject_RichCompareBool(startkey, key, Py_EQ);
114 Py_DECREF(startkey);
115 if (cmp < 0)
116 return NULL;
117 if (table != so->table || entry->key != startkey)
118 return set_lookkey(so, key, hash);
119 if (cmp > 0)
120 return entry;
Raymond Hettinger438f9132015-02-09 06:48:29 -0600121 mask = so->mask;
Raymond Hettingerc658d852015-02-02 08:35:00 -0800122 }
Raymond Hettinger8651a502015-05-27 10:37:20 -0700123 }
124 }
125
126 perturb >>= PERTURB_SHIFT;
127 i = (i * 5 + 1 + perturb) & mask;
128
Raymond Hettinger70559b52015-07-23 07:42:23 -0400129 entry = &so->table[i];
Raymond Hettinger7e3592d2015-06-21 10:47:20 -0700130 if (entry->key == NULL)
Raymond Hettinger8651a502015-05-27 10:37:20 -0700131 return entry;
132 }
133}
134
Raymond Hettinger15f08692015-07-03 17:21:17 -0700135static int set_table_resize(PySetObject *, Py_ssize_t);
136
Raymond Hettinger8651a502015-05-27 10:37:20 -0700137static int
Raymond Hettingerff9e18a2015-07-20 07:34:05 -0400138set_add_entry(PySetObject *so, PyObject *key, Py_hash_t hash)
Raymond Hettinger8651a502015-05-27 10:37:20 -0700139{
Raymond Hettingerff9e18a2015-07-20 07:34:05 -0400140 setentry *table;
Raymond Hettinger38bb95e2015-06-24 01:22:19 -0700141 setentry *freeslot;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700142 setentry *entry;
Raymond Hettinger38bb95e2015-06-24 01:22:19 -0700143 size_t perturb;
Raymond Hettingerff9e18a2015-07-20 07:34:05 -0400144 size_t mask;
145 size_t i; /* Unsigned for defined overflow behavior */
Raymond Hettinger8651a502015-05-27 10:37:20 -0700146 size_t j;
147 int cmp;
148
Raymond Hettingerff9e18a2015-07-20 07:34:05 -0400149 /* Pre-increment is necessary to prevent arbitrary code in the rich
150 comparison from deallocating the key just before the insertion. */
151 Py_INCREF(key);
152
153 restart:
154
Raymond Hettingerff9e18a2015-07-20 07:34:05 -0400155 mask = so->mask;
156 i = (size_t)hash & mask;
157
Raymond Hettinger70559b52015-07-23 07:42:23 -0400158 entry = &so->table[i];
Raymond Hettinger8651a502015-05-27 10:37:20 -0700159 if (entry->key == NULL)
Raymond Hettinger15f08692015-07-03 17:21:17 -0700160 goto found_unused;
Raymond Hettinger6ee588f2015-06-20 21:39:51 -0700161
162 freeslot = NULL;
163 perturb = hash;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700164
165 while (1) {
166 if (entry->hash == hash) {
167 PyObject *startkey = entry->key;
168 /* startkey cannot be a dummy because the dummy hash field is -1 */
169 assert(startkey != dummy);
170 if (startkey == key)
171 goto found_active;
172 if (PyUnicode_CheckExact(startkey)
173 && PyUnicode_CheckExact(key)
Raymond Hettingerac2ef652015-07-04 16:04:44 -0700174 && _PyUnicode_EQ(startkey, key))
Raymond Hettinger8651a502015-05-27 10:37:20 -0700175 goto found_active;
Raymond Hettinger70559b52015-07-23 07:42:23 -0400176 table = so->table;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700177 Py_INCREF(startkey);
178 cmp = PyObject_RichCompareBool(startkey, key, Py_EQ);
179 Py_DECREF(startkey);
Raymond Hettinger8651a502015-05-27 10:37:20 -0700180 if (cmp > 0) /* likely */
181 goto found_active;
Raymond Hettingerdaffc912015-07-31 07:58:56 -0700182 if (cmp < 0)
183 goto comparison_error;
184 /* Continuing the search from the current entry only makes
185 sense if the table and entry are unchanged; otherwise,
186 we have to restart from the beginning */
187 if (table != so->table || entry->key != startkey)
188 goto restart;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700189 mask = so->mask; /* help avoid a register spill */
190 }
Raymond Hettinger33299922018-01-14 10:20:13 -0800191 else if (entry->hash == -1)
Raymond Hettinger8651a502015-05-27 10:37:20 -0700192 freeslot = entry;
193
194 if (i + LINEAR_PROBES <= mask) {
195 for (j = 0 ; j < LINEAR_PROBES ; j++) {
196 entry++;
197 if (entry->hash == 0 && entry->key == NULL)
Raymond Hettinger15f08692015-07-03 17:21:17 -0700198 goto found_unused_or_dummy;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700199 if (entry->hash == hash) {
200 PyObject *startkey = entry->key;
201 assert(startkey != dummy);
202 if (startkey == key)
203 goto found_active;
204 if (PyUnicode_CheckExact(startkey)
205 && PyUnicode_CheckExact(key)
Raymond Hettingerac2ef652015-07-04 16:04:44 -0700206 && _PyUnicode_EQ(startkey, key))
Raymond Hettinger8651a502015-05-27 10:37:20 -0700207 goto found_active;
Raymond Hettinger70559b52015-07-23 07:42:23 -0400208 table = so->table;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700209 Py_INCREF(startkey);
210 cmp = PyObject_RichCompareBool(startkey, key, Py_EQ);
211 Py_DECREF(startkey);
Raymond Hettingerdaffc912015-07-31 07:58:56 -0700212 if (cmp > 0)
213 goto found_active;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700214 if (cmp < 0)
Raymond Hettingerff9e18a2015-07-20 07:34:05 -0400215 goto comparison_error;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700216 if (table != so->table || entry->key != startkey)
Raymond Hettingerff9e18a2015-07-20 07:34:05 -0400217 goto restart;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700218 mask = so->mask;
219 }
Raymond Hettinger33299922018-01-14 10:20:13 -0800220 else if (entry->hash == -1)
Raymond Hettingerc658d852015-02-02 08:35:00 -0800221 freeslot = entry;
Raymond Hettingera35adf52013-09-02 03:23:21 -0700222 }
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000223 }
Raymond Hettinger3c0a4f52013-08-19 07:36:04 -0700224
Raymond Hettinger3c0a4f52013-08-19 07:36:04 -0700225 perturb >>= PERTURB_SHIFT;
Raymond Hettinger59ecabd2015-01-31 02:45:12 -0800226 i = (i * 5 + 1 + perturb) & mask;
Raymond Hettinger3c0a4f52013-08-19 07:36:04 -0700227
Raymond Hettinger70559b52015-07-23 07:42:23 -0400228 entry = &so->table[i];
Raymond Hettinger7e3592d2015-06-21 10:47:20 -0700229 if (entry->key == NULL)
Raymond Hettinger15f08692015-07-03 17:21:17 -0700230 goto found_unused_or_dummy;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000231 }
Raymond Hettinger8651a502015-05-27 10:37:20 -0700232
Raymond Hettinger15f08692015-07-03 17:21:17 -0700233 found_unused_or_dummy:
234 if (freeslot == NULL)
235 goto found_unused;
Raymond Hettinger15f08692015-07-03 17:21:17 -0700236 so->used++;
237 freeslot->key = key;
238 freeslot->hash = hash;
239 return 0;
240
241 found_unused:
Raymond Hettinger6ee588f2015-06-20 21:39:51 -0700242 so->fill++;
243 so->used++;
244 entry->key = key;
245 entry->hash = hash;
Raymond Hettinger5cd87a82017-02-04 02:43:42 -0800246 if ((size_t)so->fill*5 < mask*3)
Raymond Hettinger15f08692015-07-03 17:21:17 -0700247 return 0;
INADA Naokie82cf862017-04-01 17:20:25 +0900248 return set_table_resize(so, so->used>50000 ? so->used*2 : so->used*4);
Raymond Hettinger8651a502015-05-27 10:37:20 -0700249
250 found_active:
Raymond Hettinger482c05c2015-07-20 01:23:32 -0400251 Py_DECREF(key);
Raymond Hettinger8651a502015-05-27 10:37:20 -0700252 return 0;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000253
Raymond Hettingerff9e18a2015-07-20 07:34:05 -0400254 comparison_error:
Raymond Hettinger061091a2015-07-15 23:54:02 -0700255 Py_DECREF(key);
256 return -1;
257}
258
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000259/*
Thomas Wouters89f507f2006-12-13 04:49:30 +0000260Internal routine used by set_table_resize() to insert an item which is
261known to be absent from the set. This routine also assumes that
262the set contains no deleted entries. Besides the performance benefit,
Raymond Hettinger6019c8c2015-11-17 08:28:07 -0800263there is also safety benefit since using set_add_entry() risks making
264a callback in the middle of a set_table_resize(), see issue 1456209.
265The caller is responsible for updating the key's reference count and
266the setobject's fill and used fields.
Thomas Wouters89f507f2006-12-13 04:49:30 +0000267*/
268static void
Raymond Hettinger5088f602015-12-13 18:45:01 -0800269set_insert_clean(setentry *table, size_t mask, PyObject *key, Py_hash_t hash)
Thomas Wouters89f507f2006-12-13 04:49:30 +0000270{
Antoine Pitrou9ed5f272013-08-13 20:18:52 +0200271 setentry *entry;
Raymond Hettinger3c0a4f52013-08-19 07:36:04 -0700272 size_t perturb = hash;
Raymond Hettinger3037e842015-01-26 21:33:48 -0800273 size_t i = (size_t)hash & mask;
Raymond Hettingera35adf52013-09-02 03:23:21 -0700274 size_t j;
Thomas Wouters89f507f2006-12-13 04:49:30 +0000275
Raymond Hettinger3c0a4f52013-08-19 07:36:04 -0700276 while (1) {
Raymond Hettinger3037e842015-01-26 21:33:48 -0800277 entry = &table[i];
278 if (entry->key == NULL)
279 goto found_null;
280 if (i + LINEAR_PROBES <= mask) {
Raymond Hettinger82492822015-02-04 08:37:02 -0800281 for (j = 0; j < LINEAR_PROBES; j++) {
Raymond Hettinger9edd7532015-01-30 20:09:23 -0800282 entry++;
Raymond Hettinger3037e842015-01-26 21:33:48 -0800283 if (entry->key == NULL)
284 goto found_null;
285 }
Raymond Hettingera35adf52013-09-02 03:23:21 -0700286 }
Raymond Hettinger3c0a4f52013-08-19 07:36:04 -0700287 perturb >>= PERTURB_SHIFT;
Raymond Hettinger3037e842015-01-26 21:33:48 -0800288 i = (i * 5 + 1 + perturb) & mask;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000289 }
Raymond Hettingera35adf52013-09-02 03:23:21 -0700290 found_null:
Raymond Hettinger5088f602015-12-13 18:45:01 -0800291 entry->key = key;
Raymond Hettinger86d322f2015-12-13 19:27:17 -0800292 entry->hash = hash;
Thomas Wouters89f507f2006-12-13 04:49:30 +0000293}
294
Raymond Hettinger04fd9dd2013-09-07 17:41:01 -0700295/* ======== End logic for probing the hash table ========================== */
296/* ======================================================================== */
297
Thomas Wouters89f507f2006-12-13 04:49:30 +0000298/*
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000299Restructure the table by allocating a new table and reinserting all
Raymond Hettinger06d8cf82005-07-31 15:36:06 +0000300keys again. When entries have been deleted, the new table may
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000301actually be smaller than the old one.
302*/
303static int
Thomas Wouters0e3f5912006-08-11 14:57:12 +0000304set_table_resize(PySetObject *so, Py_ssize_t minused)
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000305{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000306 setentry *oldtable, *newtable, *entry;
Raymond Hettinger5088f602015-12-13 18:45:01 -0800307 Py_ssize_t oldmask = so->mask;
308 size_t newmask;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000309 int is_oldtable_malloced;
310 setentry small_copy[PySet_MINSIZE];
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000311
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000312 assert(minused >= 0);
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000313
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000314 /* Find the smallest table size > minused. */
Raymond Hettinger426d9952014-05-18 21:40:20 +0100315 /* XXX speed-up with intrinsics */
Sergey Fedoseev6c7d67c2018-09-12 04:18:01 +0500316 size_t newsize = PySet_MINSIZE;
317 while (newsize <= (size_t)minused) {
318 newsize <<= 1; // The largest possible value is PY_SSIZE_T_MAX + 1.
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000319 }
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000320
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000321 /* Get space for a new table. */
322 oldtable = so->table;
323 assert(oldtable != NULL);
324 is_oldtable_malloced = oldtable != so->smalltable;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000325
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000326 if (newsize == PySet_MINSIZE) {
327 /* A large table is shrinking, or we can't get any smaller. */
328 newtable = so->smalltable;
329 if (newtable == oldtable) {
330 if (so->fill == so->used) {
331 /* No dummies, so no point doing anything. */
332 return 0;
333 }
334 /* We're not going to resize it, but rebuild the
335 table anyway to purge old dummy entries.
336 Subtle: This is *necessary* if fill==size,
337 as set_lookkey needs at least one virgin slot to
338 terminate failing searches. If fill < size, it's
339 merely desirable, as dummies slow searches. */
340 assert(so->fill > so->used);
341 memcpy(small_copy, oldtable, sizeof(small_copy));
342 oldtable = small_copy;
343 }
344 }
345 else {
346 newtable = PyMem_NEW(setentry, newsize);
347 if (newtable == NULL) {
348 PyErr_NoMemory();
349 return -1;
350 }
351 }
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000352
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000353 /* Make the set empty, using the new table. */
354 assert(newtable != oldtable);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000355 memset(newtable, 0, sizeof(setentry) * newsize);
Raymond Hettinger08e3dc02014-12-26 20:14:00 -0800356 so->mask = newsize - 1;
357 so->table = newtable;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000358
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000359 /* Copy the data over; this is refcount-neutral for active entries;
360 dummy entries aren't copied over, of course */
Raymond Hettinger5088f602015-12-13 18:45:01 -0800361 newmask = (size_t)so->mask;
Raymond Hettingere1af6962017-02-02 08:24:48 -0800362 if (so->fill == so->used) {
Raymond Hettinger5088f602015-12-13 18:45:01 -0800363 for (entry = oldtable; entry <= oldtable + oldmask; entry++) {
Raymond Hettinger08e3dc02014-12-26 20:14:00 -0800364 if (entry->key != NULL) {
Raymond Hettinger5088f602015-12-13 18:45:01 -0800365 set_insert_clean(newtable, newmask, entry->key, entry->hash);
Raymond Hettinger08e3dc02014-12-26 20:14:00 -0800366 }
367 }
368 } else {
Raymond Hettingere1af6962017-02-02 08:24:48 -0800369 so->fill = so->used;
Raymond Hettinger5088f602015-12-13 18:45:01 -0800370 for (entry = oldtable; entry <= oldtable + oldmask; entry++) {
Raymond Hettinger08e3dc02014-12-26 20:14:00 -0800371 if (entry->key != NULL && entry->key != dummy) {
Raymond Hettinger5088f602015-12-13 18:45:01 -0800372 set_insert_clean(newtable, newmask, entry->key, entry->hash);
Raymond Hettinger08e3dc02014-12-26 20:14:00 -0800373 }
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000374 }
375 }
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000376
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000377 if (is_oldtable_malloced)
378 PyMem_DEL(oldtable);
379 return 0;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000380}
381
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000382static int
Raymond Hettingerb48d6a62015-07-05 16:27:44 -0700383set_contains_entry(PySetObject *so, PyObject *key, Py_hash_t hash)
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000384{
Raymond Hettingerb48d6a62015-07-05 16:27:44 -0700385 setentry *entry;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000386
Raymond Hettingerb48d6a62015-07-05 16:27:44 -0700387 entry = set_lookkey(so, key, hash);
388 if (entry != NULL)
389 return entry->key != NULL;
390 return -1;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000391}
392
393#define DISCARD_NOTFOUND 0
394#define DISCARD_FOUND 1
395
396static int
Raymond Hettinger73799b12015-07-05 16:06:10 -0700397set_discard_entry(PySetObject *so, PyObject *key, Py_hash_t hash)
Raymond Hettinger8edf27c2014-12-26 23:08:58 -0800398{
399 setentry *entry;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000400 PyObject *old_key;
Raymond Hettingerc991db22005-08-11 07:58:45 +0000401
Raymond Hettingerdc28d5a2015-07-05 10:03:20 -0700402 entry = set_lookkey(so, key, hash);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000403 if (entry == NULL)
404 return -1;
Raymond Hettinger8651a502015-05-27 10:37:20 -0700405 if (entry->key == NULL)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000406 return DISCARD_NOTFOUND;
407 old_key = entry->key;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000408 entry->key = dummy;
Raymond Hettingerb335dfe2015-01-25 16:38:52 -0800409 entry->hash = -1;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000410 so->used--;
411 Py_DECREF(old_key);
412 return DISCARD_FOUND;
Raymond Hettingerc991db22005-08-11 07:58:45 +0000413}
414
415static int
Raymond Hettingerb48d6a62015-07-05 16:27:44 -0700416set_add_key(PySetObject *so, PyObject *key)
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000417{
Antoine Pitrou9ed5f272013-08-13 20:18:52 +0200418 Py_hash_t hash;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000419
Raymond Hettingerb48d6a62015-07-05 16:27:44 -0700420 if (!PyUnicode_CheckExact(key) ||
421 (hash = ((PyASCIIObject *) key)->hash) == -1) {
422 hash = PyObject_Hash(key);
423 if (hash == -1)
424 return -1;
425 }
426 return set_add_entry(so, key, hash);
427}
428
429static int
430set_contains_key(PySetObject *so, PyObject *key)
431{
432 Py_hash_t hash;
433
434 if (!PyUnicode_CheckExact(key) ||
435 (hash = ((PyASCIIObject *) key)->hash) == -1) {
436 hash = PyObject_Hash(key);
437 if (hash == -1)
438 return -1;
439 }
440 return set_contains_entry(so, key, hash);
441}
442
443static int
444set_discard_key(PySetObject *so, PyObject *key)
445{
446 Py_hash_t hash;
Christian Heimes0ded5b52007-12-10 15:50:56 +0000447
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000448 if (!PyUnicode_CheckExact(key) ||
Martin v. Löwisd63a3b82011-09-28 07:41:54 +0200449 (hash = ((PyASCIIObject *) key)->hash) == -1) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000450 hash = PyObject_Hash(key);
451 if (hash == -1)
452 return -1;
453 }
Raymond Hettinger73799b12015-07-05 16:06:10 -0700454 return set_discard_entry(so, key, hash);
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000455}
456
Raymond Hettinger4ea90802013-09-07 21:01:29 -0700457static void
458set_empty_to_minsize(PySetObject *so)
459{
460 memset(so->smalltable, 0, sizeof(so->smalltable));
461 so->fill = 0;
462 so->used = 0;
463 so->mask = PySet_MINSIZE - 1;
464 so->table = so->smalltable;
465 so->hash = -1;
466}
467
Raymond Hettingerfe889f32005-08-06 05:43:39 +0000468static int
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000469set_clear_internal(PySetObject *so)
470{
Raymond Hettinger08e3dc02014-12-26 20:14:00 -0800471 setentry *entry;
472 setentry *table = so->table;
473 Py_ssize_t fill = so->fill;
474 Py_ssize_t used = so->used;
475 int table_is_malloced = table != so->smalltable;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000476 setentry small_copy[PySet_MINSIZE];
Raymond Hettingera580c472005-08-05 17:19:54 +0000477
Raymond Hettinger583cd032013-09-07 22:06:35 -0700478 assert (PyAnySet_Check(so));
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000479 assert(table != NULL);
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000480
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000481 /* This is delicate. During the process of clearing the set,
482 * decrefs can cause the set to mutate. To avoid fatal confusion
483 * (voice of experience), we have to make the set empty before
484 * clearing the slots, and never refer to anything via so->ref while
485 * clearing.
486 */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000487 if (table_is_malloced)
Raymond Hettinger4ea90802013-09-07 21:01:29 -0700488 set_empty_to_minsize(so);
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000489
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000490 else if (fill > 0) {
491 /* It's a small table with something that needs to be cleared.
492 * Afraid the only safe way is to copy the set entries into
493 * another small table first.
494 */
495 memcpy(small_copy, table, sizeof(small_copy));
496 table = small_copy;
Raymond Hettinger4ea90802013-09-07 21:01:29 -0700497 set_empty_to_minsize(so);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000498 }
499 /* else it's a small table that's already empty */
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000500
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000501 /* Now we can finally clear things. If C had refcounts, we could
502 * assert that the refcount on table is 1 now, i.e. that this function
503 * has unique access to it, so decref side-effects can't alter it.
504 */
Raymond Hettinger08e3dc02014-12-26 20:14:00 -0800505 for (entry = table; used > 0; entry++) {
506 if (entry->key && entry->key != dummy) {
507 used--;
508 Py_DECREF(entry->key);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000509 }
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000510 }
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000511
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000512 if (table_is_malloced)
513 PyMem_DEL(table);
514 return 0;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000515}
516
517/*
518 * Iterate over a set table. Use like so:
519 *
Thomas Wouters0e3f5912006-08-11 14:57:12 +0000520 * Py_ssize_t pos;
Raymond Hettingerc991db22005-08-11 07:58:45 +0000521 * setentry *entry;
Raymond Hettingerd7946662005-08-01 21:39:29 +0000522 * pos = 0; # important! pos should not otherwise be changed by you
Raymond Hettingerc991db22005-08-11 07:58:45 +0000523 * while (set_next(yourset, &pos, &entry)) {
524 * Refer to borrowed reference in entry->key.
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000525 * }
526 *
Raymond Hettingerc991db22005-08-11 07:58:45 +0000527 * CAUTION: In general, it isn't safe to use set_next in a loop that
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000528 * mutates the table.
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000529 */
530static int
Martin v. Löwis18e16552006-02-15 17:27:45 +0000531set_next(PySetObject *so, Py_ssize_t *pos_ptr, setentry **entry_ptr)
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000532{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000533 Py_ssize_t i;
534 Py_ssize_t mask;
Raymond Hettingeref6bd7d2015-07-06 08:43:37 -0700535 setentry *entry;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000536
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000537 assert (PyAnySet_Check(so));
538 i = *pos_ptr;
539 assert(i >= 0);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000540 mask = so->mask;
Raymond Hettingeref6bd7d2015-07-06 08:43:37 -0700541 entry = &so->table[i];
542 while (i <= mask && (entry->key == NULL || entry->key == dummy)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000543 i++;
Raymond Hettingeref6bd7d2015-07-06 08:43:37 -0700544 entry++;
545 }
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000546 *pos_ptr = i+1;
547 if (i > mask)
548 return 0;
Raymond Hettingeref6bd7d2015-07-06 08:43:37 -0700549 assert(entry != NULL);
550 *entry_ptr = entry;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000551 return 1;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000552}
553
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000554static void
555set_dealloc(PySetObject *so)
556{
Antoine Pitrou9ed5f272013-08-13 20:18:52 +0200557 setentry *entry;
Raymond Hettinger08e3dc02014-12-26 20:14:00 -0800558 Py_ssize_t used = so->used;
559
INADA Naokia6296d32017-08-24 14:55:17 +0900560 /* bpo-31095: UnTrack is needed before calling any callbacks */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000561 PyObject_GC_UnTrack(so);
Jeroen Demeyer351c6742019-05-10 19:21:11 +0200562 Py_TRASHCAN_BEGIN(so, set_dealloc)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000563 if (so->weakreflist != NULL)
564 PyObject_ClearWeakRefs((PyObject *) so);
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000565
Raymond Hettinger08e3dc02014-12-26 20:14:00 -0800566 for (entry = so->table; used > 0; entry++) {
567 if (entry->key && entry->key != dummy) {
568 used--;
Raymond Hettingerfcf3b502013-08-22 08:20:31 -0700569 Py_DECREF(entry->key);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000570 }
571 }
572 if (so->table != so->smalltable)
573 PyMem_DEL(so->table);
Raymond Hettinger8f8839e2013-09-07 20:26:50 -0700574 Py_TYPE(so)->tp_free(so);
Jeroen Demeyer351c6742019-05-10 19:21:11 +0200575 Py_TRASHCAN_END
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000576}
577
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000578static PyObject *
579set_repr(PySetObject *so)
580{
Martin v. Löwisd63a3b82011-09-28 07:41:54 +0200581 PyObject *result=NULL, *keys, *listrepr, *tmp;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000582 int status = Py_ReprEnter((PyObject*)so);
Thomas Wouters902d6eb2007-01-09 23:18:33 +0000583
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000584 if (status != 0) {
585 if (status < 0)
586 return NULL;
587 return PyUnicode_FromFormat("%s(...)", Py_TYPE(so)->tp_name);
588 }
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000589
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000590 /* shortcut for the empty set */
591 if (!so->used) {
592 Py_ReprLeave((PyObject*)so);
593 return PyUnicode_FromFormat("%s()", Py_TYPE(so)->tp_name);
594 }
Georg Brandlc4996ba2006-08-28 19:37:11 +0000595
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000596 keys = PySequence_List((PyObject *)so);
597 if (keys == NULL)
598 goto done;
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000599
Martin v. Löwisd63a3b82011-09-28 07:41:54 +0200600 /* repr(keys)[1:-1] */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000601 listrepr = PyObject_Repr(keys);
602 Py_DECREF(keys);
603 if (listrepr == NULL)
604 goto done;
Martin v. Löwisd63a3b82011-09-28 07:41:54 +0200605 tmp = PyUnicode_Substring(listrepr, 1, PyUnicode_GET_LENGTH(listrepr)-1);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000606 Py_DECREF(listrepr);
Martin v. Löwisd63a3b82011-09-28 07:41:54 +0200607 if (tmp == NULL)
608 goto done;
609 listrepr = tmp;
Victor Stinnera1a807b2011-05-26 14:24:30 +0200610
Andy Lester55728702020-03-06 16:53:17 -0600611 if (!Py_IS_TYPE(so, &PySet_Type))
Martin v. Löwisd63a3b82011-09-28 07:41:54 +0200612 result = PyUnicode_FromFormat("%s({%U})",
613 Py_TYPE(so)->tp_name,
614 listrepr);
615 else
616 result = PyUnicode_FromFormat("{%U}", listrepr);
617 Py_DECREF(listrepr);
Thomas Wouters902d6eb2007-01-09 23:18:33 +0000618done:
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000619 Py_ReprLeave((PyObject*)so);
620 return result;
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000621}
622
Martin v. Löwis18e16552006-02-15 17:27:45 +0000623static Py_ssize_t
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000624set_len(PyObject *so)
625{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000626 return ((PySetObject *)so)->used;
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000627}
628
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000629static int
Raymond Hettingerc991db22005-08-11 07:58:45 +0000630set_merge(PySetObject *so, PyObject *otherset)
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000631{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000632 PySetObject *other;
Raymond Hettingerfaf7b7f2010-09-03 10:00:50 +0000633 PyObject *key;
Antoine Pitrou9ed5f272013-08-13 20:18:52 +0200634 Py_ssize_t i;
Raymond Hettinger1bd8d752015-05-13 01:26:14 -0700635 setentry *so_entry;
636 setentry *other_entry;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000637
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000638 assert (PyAnySet_Check(so));
639 assert (PyAnySet_Check(otherset));
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000640
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000641 other = (PySetObject*)otherset;
642 if (other == so || other->used == 0)
Raymond Hettinger1bd8d752015-05-13 01:26:14 -0700643 /* a.update(a) or a.update(set()); nothing to do */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000644 return 0;
645 /* Do one big resize at the start, rather than
646 * incrementally resizing as we insert new keys. Expect
647 * that there will be no (or few) overlapping keys.
648 */
Raymond Hettinger5cd87a82017-02-04 02:43:42 -0800649 if ((so->fill + other->used)*5 >= so->mask*3) {
INADA Naokie82cf862017-04-01 17:20:25 +0900650 if (set_table_resize(so, (so->used + other->used)*2) != 0)
651 return -1;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000652 }
Raymond Hettinger1bd8d752015-05-13 01:26:14 -0700653 so_entry = so->table;
654 other_entry = other->table;
655
656 /* If our table is empty, and both tables have the same size, and
657 there are no dummies to eliminate, then just copy the pointers. */
658 if (so->fill == 0 && so->mask == other->mask && other->fill == other->used) {
659 for (i = 0; i <= other->mask; i++, so_entry++, other_entry++) {
660 key = other_entry->key;
661 if (key != NULL) {
662 assert(so_entry->key == NULL);
663 Py_INCREF(key);
664 so_entry->key = key;
665 so_entry->hash = other_entry->hash;
666 }
667 }
668 so->fill = other->fill;
669 so->used = other->used;
670 return 0;
671 }
672
673 /* If our table is empty, we can use set_insert_clean() */
674 if (so->fill == 0) {
Raymond Hettinger5088f602015-12-13 18:45:01 -0800675 setentry *newtable = so->table;
676 size_t newmask = (size_t)so->mask;
Raymond Hettinger6019c8c2015-11-17 08:28:07 -0800677 so->fill = other->used;
678 so->used = other->used;
Raymond Hettingere4495872015-12-15 00:42:30 -0800679 for (i = other->mask + 1; i > 0 ; i--, other_entry++) {
Raymond Hettinger1bd8d752015-05-13 01:26:14 -0700680 key = other_entry->key;
681 if (key != NULL && key != dummy) {
682 Py_INCREF(key);
Raymond Hettinger5088f602015-12-13 18:45:01 -0800683 set_insert_clean(newtable, newmask, key, other_entry->hash);
Raymond Hettinger1bd8d752015-05-13 01:26:14 -0700684 }
685 }
686 return 0;
687 }
688
689 /* We can't assure there are no duplicates, so do normal insertions */
Raymond Hettingera3626bc2015-07-15 23:50:14 -0700690 for (i = 0; i <= other->mask; i++) {
691 other_entry = &other->table[i];
Raymond Hettinger1bd8d752015-05-13 01:26:14 -0700692 key = other_entry->key;
693 if (key != NULL && key != dummy) {
Raymond Hettinger73799b12015-07-05 16:06:10 -0700694 if (set_add_entry(so, key, other_entry->hash))
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000695 return -1;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000696 }
697 }
698 return 0;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000699}
700
Raymond Hettingerce8185e2005-08-13 09:28:48 +0000701static PyObject *
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +0530702set_pop(PySetObject *so, PyObject *Py_UNUSED(ignored))
Raymond Hettingerce8185e2005-08-13 09:28:48 +0000703{
Raymond Hettinger9cd6a782015-01-18 16:06:18 -0800704 /* Make sure the search finger is in bounds */
Raymond Hettingerf9ec1b92018-11-11 14:35:47 -0800705 setentry *entry = so->table + (so->finger & so->mask);
706 setentry *limit = so->table + so->mask;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000707 PyObject *key;
Raymond Hettingerce8185e2005-08-13 09:28:48 +0000708
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000709 if (so->used == 0) {
710 PyErr_SetString(PyExc_KeyError, "pop from an empty set");
711 return NULL;
712 }
Raymond Hettingercf5863f2018-11-09 02:31:56 -0800713 while (entry->key == NULL || entry->key==dummy) {
714 entry++;
715 if (entry > limit)
716 entry = so->table;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000717 }
718 key = entry->key;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000719 entry->key = dummy;
Raymond Hettingerb335dfe2015-01-25 16:38:52 -0800720 entry->hash = -1;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000721 so->used--;
Raymond Hettingercf5863f2018-11-09 02:31:56 -0800722 so->finger = entry - so->table + 1; /* next place to start */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000723 return key;
Raymond Hettingerce8185e2005-08-13 09:28:48 +0000724}
725
Benjamin Petersonf10a79a2008-10-11 00:49:57 +0000726PyDoc_STRVAR(pop_doc, "Remove and return an arbitrary set element.\n\
727Raises KeyError if the set is empty.");
Raymond Hettingerce8185e2005-08-13 09:28:48 +0000728
729static int
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000730set_traverse(PySetObject *so, visitproc visit, void *arg)
Raymond Hettingerce8185e2005-08-13 09:28:48 +0000731{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000732 Py_ssize_t pos = 0;
733 setentry *entry;
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000734
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000735 while (set_next(so, &pos, &entry))
736 Py_VISIT(entry->key);
737 return 0;
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000738}
739
Raymond Hettingerfbffdef2015-08-01 09:53:00 -0700740/* Work to increase the bit dispersion for closely spaced hash values.
741 This is important because some use cases have many combinations of a
742 small number of elements with nearby hashes so that many distinct
743 combinations collapse to only a handful of distinct hash values. */
744
745static Py_uhash_t
746_shuffle_bits(Py_uhash_t h)
747{
748 return ((h ^ 89869747UL) ^ (h << 16)) * 3644798167UL;
749}
750
751/* Most of the constants in this hash algorithm are randomly chosen
752 large primes with "interesting bit patterns" and that passed tests
753 for good collision statistics on a variety of problematic datasets
754 including powersets and graph structures (such as David Eppstein's
755 graph recipes in Lib/test/test_set.py) */
756
Benjamin Peterson8f67d082010-10-17 20:54:53 +0000757static Py_hash_t
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000758frozenset_hash(PyObject *self)
759{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000760 PySetObject *so = (PySetObject *)self;
Raymond Hettinger41481952015-08-07 00:43:39 -0700761 Py_uhash_t hash = 0;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000762 setentry *entry;
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000763
Raymond Hettingerb501a272015-08-06 22:15:22 -0700764 if (so->hash != -1)
765 return so->hash;
766
Raymond Hettingerfbffdef2015-08-01 09:53:00 -0700767 /* Xor-in shuffled bits from every entry's hash field because xor is
768 commutative and a frozenset hash should be independent of order.
769
770 For speed, include null entries and dummy entries and then
771 subtract out their effect afterwards so that the final hash
772 depends only on active entries. This allows the code to be
773 vectorized by the compiler and it saves the unpredictable
774 branches that would arise when trying to exclude null and dummy
775 entries on every iteration. */
776
777 for (entry = so->table; entry <= &so->table[so->mask]; entry++)
778 hash ^= _shuffle_bits(entry->hash);
779
Raymond Hettingera286a512015-08-01 11:07:11 -0700780 /* Remove the effect of an odd number of NULL entries */
Raymond Hettingerfbffdef2015-08-01 09:53:00 -0700781 if ((so->mask + 1 - so->fill) & 1)
782 hash ^= _shuffle_bits(0);
783
784 /* Remove the effect of an odd number of dummy entries */
785 if ((so->fill - so->used) & 1)
786 hash ^= _shuffle_bits(-1);
787
Raymond Hettinger41481952015-08-07 00:43:39 -0700788 /* Factor in the number of active entries */
789 hash ^= ((Py_uhash_t)PySet_GET_SIZE(self) + 1) * 1927868237UL;
790
Raymond Hettingerfbffdef2015-08-01 09:53:00 -0700791 /* Disperse patterns arising in nested frozensets */
Raymond Hettingerfa788062018-01-18 13:23:27 -0800792 hash ^= (hash >> 11) ^ (hash >> 25);
Gregory P. Smithc2176e42012-12-10 18:32:53 -0800793 hash = hash * 69069U + 907133923UL;
Raymond Hettingerfbffdef2015-08-01 09:53:00 -0700794
Raymond Hettinger36c05002015-08-01 10:57:42 -0700795 /* -1 is reserved as an error code */
Victor Stinner12174a52014-08-15 23:17:38 +0200796 if (hash == (Py_uhash_t)-1)
Gregory P. Smith27cbcd62012-12-10 18:15:46 -0800797 hash = 590923713UL;
Raymond Hettinger36c05002015-08-01 10:57:42 -0700798
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000799 so->hash = hash;
800 return hash;
Raymond Hettingerf408ddf2005-08-17 00:27:42 +0000801}
802
Raymond Hettingera9d99362005-08-05 00:01:15 +0000803/***** Set iterator type ***********************************************/
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000804
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000805typedef struct {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000806 PyObject_HEAD
807 PySetObject *si_set; /* Set to NULL when iterator is exhausted */
808 Py_ssize_t si_used;
Raymond Hettinger9632a7d2015-07-07 15:29:24 -0700809 Py_ssize_t si_pos;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000810 Py_ssize_t len;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000811} setiterobject;
812
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000813static void
814setiter_dealloc(setiterobject *si)
815{
INADA Naokia6296d32017-08-24 14:55:17 +0900816 /* bpo-31095: UnTrack is needed before calling any callbacks */
817 _PyObject_GC_UNTRACK(si);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000818 Py_XDECREF(si->si_set);
819 PyObject_GC_Del(si);
Antoine Pitrou7ddda782009-01-01 15:35:33 +0000820}
821
822static int
823setiter_traverse(setiterobject *si, visitproc visit, void *arg)
824{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000825 Py_VISIT(si->si_set);
826 return 0;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000827}
828
Raymond Hettinger6b27cda2005-09-24 21:23:05 +0000829static PyObject *
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +0530830setiter_len(setiterobject *si, PyObject *Py_UNUSED(ignored))
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000831{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000832 Py_ssize_t len = 0;
833 if (si->si_set != NULL && si->si_used == si->si_set->used)
834 len = si->len;
Antoine Pitrou671b4d92010-08-17 17:55:07 +0000835 return PyLong_FromSsize_t(len);
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000836}
837
Armin Rigof5b3e362006-02-11 21:32:43 +0000838PyDoc_STRVAR(length_hint_doc, "Private method returning an estimate of len(list(it)).");
Raymond Hettinger6b27cda2005-09-24 21:23:05 +0000839
Kristján Valur Jónsson31668b82012-04-03 10:49:41 +0000840static PyObject *setiter_iternext(setiterobject *si);
841
842static PyObject *
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +0530843setiter_reduce(setiterobject *si, PyObject *Py_UNUSED(ignored))
Kristján Valur Jónsson31668b82012-04-03 10:49:41 +0000844{
Serhiy Storchakabb86bf42018-12-11 08:28:18 +0200845 _Py_IDENTIFIER(iter);
Ezio Melotti0e1af282012-09-28 16:43:40 +0300846 /* copy the iterator state */
Sergey Fedoseev63958442018-10-20 05:43:33 +0500847 setiterobject tmp = *si;
Kristján Valur Jónsson31668b82012-04-03 10:49:41 +0000848 Py_XINCREF(tmp.si_set);
Ezio Melotti0e1af282012-09-28 16:43:40 +0300849
Kristján Valur Jónsson31668b82012-04-03 10:49:41 +0000850 /* iterate the temporary into a list */
Sergey Fedoseev63958442018-10-20 05:43:33 +0500851 PyObject *list = PySequence_List((PyObject*)&tmp);
Kristján Valur Jónsson31668b82012-04-03 10:49:41 +0000852 Py_XDECREF(tmp.si_set);
Sergey Fedoseev63958442018-10-20 05:43:33 +0500853 if (list == NULL) {
Kristján Valur Jónsson31668b82012-04-03 10:49:41 +0000854 return NULL;
855 }
Serhiy Storchakabb86bf42018-12-11 08:28:18 +0200856 return Py_BuildValue("N(N)", _PyEval_GetBuiltinId(&PyId_iter), list);
Kristján Valur Jónsson31668b82012-04-03 10:49:41 +0000857}
858
859PyDoc_STRVAR(reduce_doc, "Return state information for pickling.");
860
Raymond Hettinger6b27cda2005-09-24 21:23:05 +0000861static PyMethodDef setiter_methods[] = {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000862 {"__length_hint__", (PyCFunction)setiter_len, METH_NOARGS, length_hint_doc},
Kristján Valur Jónsson31668b82012-04-03 10:49:41 +0000863 {"__reduce__", (PyCFunction)setiter_reduce, METH_NOARGS, reduce_doc},
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000864 {NULL, NULL} /* sentinel */
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000865};
866
Raymond Hettinger06d8cf82005-07-31 15:36:06 +0000867static PyObject *setiter_iternext(setiterobject *si)
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000868{
Raymond Hettinger9632a7d2015-07-07 15:29:24 -0700869 PyObject *key;
870 Py_ssize_t i, mask;
Antoine Pitrou9ed5f272013-08-13 20:18:52 +0200871 setentry *entry;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000872 PySetObject *so = si->si_set;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000873
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000874 if (so == NULL)
875 return NULL;
876 assert (PyAnySet_Check(so));
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000877
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000878 if (si->si_used != so->used) {
879 PyErr_SetString(PyExc_RuntimeError,
880 "Set changed size during iteration");
881 si->si_used = -1; /* Make this state sticky */
882 return NULL;
883 }
Raymond Hettinger9632a7d2015-07-07 15:29:24 -0700884
885 i = si->si_pos;
886 assert(i>=0);
887 entry = so->table;
888 mask = so->mask;
889 while (i <= mask && (entry[i].key == NULL || entry[i].key == dummy))
890 i++;
891 si->si_pos = i+1;
892 if (i > mask)
893 goto fail;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000894 si->len--;
Raymond Hettinger9632a7d2015-07-07 15:29:24 -0700895 key = entry[i].key;
896 Py_INCREF(key);
897 return key;
898
899fail:
Raymond Hettinger9632a7d2015-07-07 15:29:24 -0700900 si->si_set = NULL;
Serhiy Storchakafbb1c5e2016-03-30 20:40:02 +0300901 Py_DECREF(so);
Raymond Hettinger9632a7d2015-07-07 15:29:24 -0700902 return NULL;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000903}
904
Christian Heimesa22e8bd2007-11-29 22:35:39 +0000905PyTypeObject PySetIter_Type = {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000906 PyVarObject_HEAD_INIT(&PyType_Type, 0)
907 "set_iterator", /* tp_name */
908 sizeof(setiterobject), /* tp_basicsize */
909 0, /* tp_itemsize */
910 /* methods */
911 (destructor)setiter_dealloc, /* tp_dealloc */
Jeroen Demeyer530f5062019-05-31 04:13:39 +0200912 0, /* tp_vectorcall_offset */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000913 0, /* tp_getattr */
914 0, /* tp_setattr */
Jeroen Demeyer530f5062019-05-31 04:13:39 +0200915 0, /* tp_as_async */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000916 0, /* tp_repr */
917 0, /* tp_as_number */
918 0, /* tp_as_sequence */
919 0, /* tp_as_mapping */
920 0, /* tp_hash */
921 0, /* tp_call */
922 0, /* tp_str */
923 PyObject_GenericGetAttr, /* tp_getattro */
924 0, /* tp_setattro */
925 0, /* tp_as_buffer */
Raymond Hettinger5d2385f2015-07-08 11:52:27 -0700926 Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC, /* tp_flags */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000927 0, /* tp_doc */
928 (traverseproc)setiter_traverse, /* tp_traverse */
929 0, /* tp_clear */
930 0, /* tp_richcompare */
931 0, /* tp_weaklistoffset */
932 PyObject_SelfIter, /* tp_iter */
933 (iternextfunc)setiter_iternext, /* tp_iternext */
934 setiter_methods, /* tp_methods */
935 0,
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000936};
937
Thomas Wouters49fd7fa2006-04-21 10:40:58 +0000938static PyObject *
939set_iter(PySetObject *so)
940{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000941 setiterobject *si = PyObject_GC_New(setiterobject, &PySetIter_Type);
942 if (si == NULL)
943 return NULL;
944 Py_INCREF(so);
945 si->si_set = so;
946 si->si_used = so->used;
Raymond Hettinger9632a7d2015-07-07 15:29:24 -0700947 si->si_pos = 0;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000948 si->len = so->used;
949 _PyObject_GC_TRACK(si);
950 return (PyObject *)si;
Thomas Wouters49fd7fa2006-04-21 10:40:58 +0000951}
952
Raymond Hettingerd7946662005-08-01 21:39:29 +0000953static int
Raymond Hettingerd7946662005-08-01 21:39:29 +0000954set_update_internal(PySetObject *so, PyObject *other)
Raymond Hettingera690a992003-11-16 16:17:49 +0000955{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000956 PyObject *key, *it;
Raymond Hettingera690a992003-11-16 16:17:49 +0000957
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000958 if (PyAnySet_Check(other))
959 return set_merge(so, other);
Raymond Hettingera690a992003-11-16 16:17:49 +0000960
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000961 if (PyDict_CheckExact(other)) {
962 PyObject *value;
963 Py_ssize_t pos = 0;
Benjamin Peterson8f67d082010-10-17 20:54:53 +0000964 Py_hash_t hash;
Serhiy Storchaka5ab81d72016-12-16 16:18:57 +0200965 Py_ssize_t dictsize = PyDict_GET_SIZE(other);
Thomas Wouterscf297e42007-02-23 15:07:44 +0000966
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000967 /* Do one big resize at the start, rather than
968 * incrementally resizing as we insert new keys. Expect
969 * that there will be no (or few) overlapping keys.
970 */
Raymond Hettingerc2480dc2015-07-04 08:46:31 -0700971 if (dictsize < 0)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000972 return -1;
Raymond Hettinger5cd87a82017-02-04 02:43:42 -0800973 if ((so->fill + dictsize)*5 >= so->mask*3) {
INADA Naokie82cf862017-04-01 17:20:25 +0900974 if (set_table_resize(so, (so->used + dictsize)*2) != 0)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000975 return -1;
976 }
977 while (_PyDict_Next(other, &pos, &key, &value, &hash)) {
Raymond Hettinger73799b12015-07-05 16:06:10 -0700978 if (set_add_entry(so, key, hash))
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000979 return -1;
980 }
981 return 0;
982 }
Raymond Hettinger9f1a6792005-07-31 01:16:36 +0000983
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000984 it = PyObject_GetIter(other);
985 if (it == NULL)
986 return -1;
Raymond Hettingera690a992003-11-16 16:17:49 +0000987
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000988 while ((key = PyIter_Next(it)) != NULL) {
Raymond Hettinger5af9e132015-05-13 01:44:36 -0700989 if (set_add_key(so, key)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +0000990 Py_DECREF(it);
991 Py_DECREF(key);
992 return -1;
993 }
994 Py_DECREF(key);
995 }
996 Py_DECREF(it);
997 if (PyErr_Occurred())
998 return -1;
999 return 0;
Raymond Hettingerd7946662005-08-01 21:39:29 +00001000}
1001
1002static PyObject *
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001003set_update(PySetObject *so, PyObject *args)
Raymond Hettingerd7946662005-08-01 21:39:29 +00001004{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001005 Py_ssize_t i;
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001006
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001007 for (i=0 ; i<PyTuple_GET_SIZE(args) ; i++) {
1008 PyObject *other = PyTuple_GET_ITEM(args, i);
Raymond Hettinger5af9e132015-05-13 01:44:36 -07001009 if (set_update_internal(so, other))
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001010 return NULL;
1011 }
1012 Py_RETURN_NONE;
Raymond Hettingera38123e2003-11-24 22:18:49 +00001013}
1014
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001015PyDoc_STRVAR(update_doc,
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001016"Update a set with the union of itself and others.");
Raymond Hettingera38123e2003-11-24 22:18:49 +00001017
Raymond Hettinger426d9952014-05-18 21:40:20 +01001018/* XXX Todo:
1019 If aligned memory allocations become available, make the
1020 set object 64 byte aligned so that most of the fields
1021 can be retrieved or updated in a single cache line.
1022*/
1023
Raymond Hettingera38123e2003-11-24 22:18:49 +00001024static PyObject *
1025make_new_set(PyTypeObject *type, PyObject *iterable)
1026{
Raymond Hettinger8421d712016-04-29 01:37:05 -07001027 PySetObject *so;
Raymond Hettingera38123e2003-11-24 22:18:49 +00001028
Raymond Hettinger8f8839e2013-09-07 20:26:50 -07001029 so = (PySetObject *)type->tp_alloc(type, 0);
1030 if (so == NULL)
1031 return NULL;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +00001032
Raymond Hettinger4ea90802013-09-07 21:01:29 -07001033 so->fill = 0;
1034 so->used = 0;
1035 so->mask = PySet_MINSIZE - 1;
1036 so->table = so->smalltable;
Raymond Hettinger4ea90802013-09-07 21:01:29 -07001037 so->hash = -1;
Raymond Hettinger1202a472015-01-18 13:12:42 -08001038 so->finger = 0;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001039 so->weakreflist = NULL;
Raymond Hettingera690a992003-11-16 16:17:49 +00001040
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001041 if (iterable != NULL) {
Raymond Hettinger5af9e132015-05-13 01:44:36 -07001042 if (set_update_internal(so, iterable)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001043 Py_DECREF(so);
1044 return NULL;
1045 }
1046 }
Raymond Hettingera38123e2003-11-24 22:18:49 +00001047
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001048 return (PyObject *)so;
Raymond Hettingera690a992003-11-16 16:17:49 +00001049}
1050
Raymond Hettinger7d99f092008-11-16 11:44:54 +00001051static PyObject *
1052make_new_set_basetype(PyTypeObject *type, PyObject *iterable)
1053{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001054 if (type != &PySet_Type && type != &PyFrozenSet_Type) {
1055 if (PyType_IsSubtype(type, &PySet_Type))
1056 type = &PySet_Type;
1057 else
1058 type = &PyFrozenSet_Type;
1059 }
1060 return make_new_set(type, iterable);
Raymond Hettinger7d99f092008-11-16 11:44:54 +00001061}
1062
Raymond Hettingerd7946662005-08-01 21:39:29 +00001063/* The empty frozenset is a singleton */
1064static PyObject *emptyfrozenset = NULL;
1065
Raymond Hettingera690a992003-11-16 16:17:49 +00001066static PyObject *
Raymond Hettinger50a4bb32003-11-17 16:42:33 +00001067frozenset_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
Raymond Hettingera690a992003-11-16 16:17:49 +00001068{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001069 PyObject *iterable = NULL, *result;
Raymond Hettingera690a992003-11-16 16:17:49 +00001070
Serhiy Storchaka6cca5c82017-06-08 14:41:19 +03001071 if (type == &PyFrozenSet_Type && !_PyArg_NoKeywords("frozenset", kwds))
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001072 return NULL;
Georg Brandl02c42872005-08-26 06:42:30 +00001073
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001074 if (!PyArg_UnpackTuple(args, type->tp_name, 0, 1, &iterable))
1075 return NULL;
Raymond Hettingerd7946662005-08-01 21:39:29 +00001076
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001077 if (type != &PyFrozenSet_Type)
1078 return make_new_set(type, iterable);
Raymond Hettingerd7946662005-08-01 21:39:29 +00001079
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001080 if (iterable != NULL) {
1081 /* frozenset(f) is idempotent */
1082 if (PyFrozenSet_CheckExact(iterable)) {
1083 Py_INCREF(iterable);
1084 return iterable;
1085 }
1086 result = make_new_set(type, iterable);
1087 if (result == NULL || PySet_GET_SIZE(result))
1088 return result;
1089 Py_DECREF(result);
1090 }
1091 /* The empty frozenset is a singleton */
1092 if (emptyfrozenset == NULL)
1093 emptyfrozenset = make_new_set(type, NULL);
1094 Py_XINCREF(emptyfrozenset);
1095 return emptyfrozenset;
Raymond Hettingerd7946662005-08-01 21:39:29 +00001096}
1097
Raymond Hettinger50a4bb32003-11-17 16:42:33 +00001098static PyObject *
1099set_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
1100{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001101 return make_new_set(type, NULL);
Raymond Hettinger50a4bb32003-11-17 16:42:33 +00001102}
1103
Raymond Hettinger934d63e2005-07-31 01:33:10 +00001104/* set_swap_bodies() switches the contents of any two sets by moving their
1105 internal data pointers and, if needed, copying the internal smalltables.
1106 Semantically equivalent to:
1107
1108 t=set(a); a.clear(); a.update(b); b.clear(); b.update(t); del t
1109
1110 The function always succeeds and it leaves both objects in a stable state.
Raymond Hettinger4d45c102015-01-25 16:27:40 -08001111 Useful for operations that update in-place (by allowing an intermediate
1112 result to be swapped into one of the original inputs).
Raymond Hettinger934d63e2005-07-31 01:33:10 +00001113*/
1114
Raymond Hettinger9f1a6792005-07-31 01:16:36 +00001115static void
1116set_swap_bodies(PySetObject *a, PySetObject *b)
Raymond Hettingera690a992003-11-16 16:17:49 +00001117{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001118 Py_ssize_t t;
1119 setentry *u;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001120 setentry tab[PySet_MINSIZE];
Antoine Pitroufbb1c612010-10-23 17:37:54 +00001121 Py_hash_t h;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +00001122
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001123 t = a->fill; a->fill = b->fill; b->fill = t;
1124 t = a->used; a->used = b->used; b->used = t;
1125 t = a->mask; a->mask = b->mask; b->mask = t;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +00001126
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001127 u = a->table;
1128 if (a->table == a->smalltable)
1129 u = b->smalltable;
1130 a->table = b->table;
1131 if (b->table == b->smalltable)
1132 a->table = a->smalltable;
1133 b->table = u;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +00001134
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001135 if (a->table == a->smalltable || b->table == b->smalltable) {
1136 memcpy(tab, a->smalltable, sizeof(tab));
1137 memcpy(a->smalltable, b->smalltable, sizeof(tab));
1138 memcpy(b->smalltable, tab, sizeof(tab));
1139 }
Raymond Hettinger9f1a6792005-07-31 01:16:36 +00001140
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001141 if (PyType_IsSubtype(Py_TYPE(a), &PyFrozenSet_Type) &&
1142 PyType_IsSubtype(Py_TYPE(b), &PyFrozenSet_Type)) {
1143 h = a->hash; a->hash = b->hash; b->hash = h;
1144 } else {
1145 a->hash = -1;
1146 b->hash = -1;
1147 }
Raymond Hettingera690a992003-11-16 16:17:49 +00001148}
1149
Raymond Hettinger8f5cdaa2003-12-13 11:26:12 +00001150static PyObject *
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301151set_copy(PySetObject *so, PyObject *Py_UNUSED(ignored))
Raymond Hettingera690a992003-11-16 16:17:49 +00001152{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001153 return make_new_set_basetype(Py_TYPE(so), (PyObject *)so);
Raymond Hettingera690a992003-11-16 16:17:49 +00001154}
1155
Raymond Hettinger49ba4c32003-11-23 02:49:05 +00001156static PyObject *
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301157frozenset_copy(PySetObject *so, PyObject *Py_UNUSED(ignored))
Raymond Hettinger49ba4c32003-11-23 02:49:05 +00001158{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001159 if (PyFrozenSet_CheckExact(so)) {
1160 Py_INCREF(so);
1161 return (PyObject *)so;
1162 }
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301163 return set_copy(so, NULL);
Raymond Hettinger49ba4c32003-11-23 02:49:05 +00001164}
1165
Raymond Hettingera690a992003-11-16 16:17:49 +00001166PyDoc_STRVAR(copy_doc, "Return a shallow copy of a set.");
1167
1168static PyObject *
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301169set_clear(PySetObject *so, PyObject *Py_UNUSED(ignored))
Raymond Hettingerc991db22005-08-11 07:58:45 +00001170{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001171 set_clear_internal(so);
1172 Py_RETURN_NONE;
Raymond Hettingerc991db22005-08-11 07:58:45 +00001173}
1174
1175PyDoc_STRVAR(clear_doc, "Remove all elements from this set.");
1176
1177static PyObject *
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001178set_union(PySetObject *so, PyObject *args)
Raymond Hettingerf5f41bf2003-11-24 02:57:33 +00001179{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001180 PySetObject *result;
1181 PyObject *other;
1182 Py_ssize_t i;
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001183
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301184 result = (PySetObject *)set_copy(so, NULL);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001185 if (result == NULL)
1186 return NULL;
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001187
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001188 for (i=0 ; i<PyTuple_GET_SIZE(args) ; i++) {
1189 other = PyTuple_GET_ITEM(args, i);
1190 if ((PyObject *)so == other)
1191 continue;
Raymond Hettinger5af9e132015-05-13 01:44:36 -07001192 if (set_update_internal(result, other)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001193 Py_DECREF(result);
1194 return NULL;
1195 }
1196 }
1197 return (PyObject *)result;
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001198}
1199
1200PyDoc_STRVAR(union_doc,
1201 "Return the union of sets as a new set.\n\
1202\n\
1203(i.e. all elements that are in either set.)");
1204
1205static PyObject *
1206set_or(PySetObject *so, PyObject *other)
1207{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001208 PySetObject *result;
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001209
Brian Curtindfc80e32011-08-10 20:28:54 -05001210 if (!PyAnySet_Check(so) || !PyAnySet_Check(other))
1211 Py_RETURN_NOTIMPLEMENTED;
Raymond Hettingerf5f41bf2003-11-24 02:57:33 +00001212
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301213 result = (PySetObject *)set_copy(so, NULL);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001214 if (result == NULL)
1215 return NULL;
1216 if ((PyObject *)so == other)
1217 return (PyObject *)result;
Raymond Hettinger5af9e132015-05-13 01:44:36 -07001218 if (set_update_internal(result, other)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001219 Py_DECREF(result);
1220 return NULL;
1221 }
1222 return (PyObject *)result;
Raymond Hettingerf5f41bf2003-11-24 02:57:33 +00001223}
1224
Raymond Hettingera690a992003-11-16 16:17:49 +00001225static PyObject *
1226set_ior(PySetObject *so, PyObject *other)
1227{
Brian Curtindfc80e32011-08-10 20:28:54 -05001228 if (!PyAnySet_Check(other))
1229 Py_RETURN_NOTIMPLEMENTED;
1230
Raymond Hettinger5af9e132015-05-13 01:44:36 -07001231 if (set_update_internal(so, other))
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001232 return NULL;
1233 Py_INCREF(so);
1234 return (PyObject *)so;
Raymond Hettingera690a992003-11-16 16:17:49 +00001235}
1236
1237static PyObject *
1238set_intersection(PySetObject *so, PyObject *other)
1239{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001240 PySetObject *result;
1241 PyObject *key, *it, *tmp;
Raymond Hettinger73799b12015-07-05 16:06:10 -07001242 Py_hash_t hash;
1243 int rv;
Raymond Hettingera690a992003-11-16 16:17:49 +00001244
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001245 if ((PyObject *)so == other)
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301246 return set_copy(so, NULL);
Raymond Hettingerc991db22005-08-11 07:58:45 +00001247
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001248 result = (PySetObject *)make_new_set_basetype(Py_TYPE(so), NULL);
1249 if (result == NULL)
1250 return NULL;
Raymond Hettingerf5f41bf2003-11-24 02:57:33 +00001251
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001252 if (PyAnySet_Check(other)) {
1253 Py_ssize_t pos = 0;
1254 setentry *entry;
Raymond Hettingerb02c35e2005-08-12 20:48:39 +00001255
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001256 if (PySet_GET_SIZE(other) > PySet_GET_SIZE(so)) {
1257 tmp = (PyObject *)so;
1258 so = (PySetObject *)other;
1259 other = tmp;
1260 }
Raymond Hettingerb02c35e2005-08-12 20:48:39 +00001261
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001262 while (set_next((PySetObject *)other, &pos, &entry)) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001263 key = entry->key;
1264 hash = entry->hash;
1265 rv = set_contains_entry(so, key, hash);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001266 if (rv < 0) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001267 Py_DECREF(result);
1268 return NULL;
1269 }
1270 if (rv) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001271 if (set_add_entry(result, key, hash)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001272 Py_DECREF(result);
1273 return NULL;
1274 }
1275 }
1276 }
1277 return (PyObject *)result;
1278 }
Raymond Hettingerf5f41bf2003-11-24 02:57:33 +00001279
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001280 it = PyObject_GetIter(other);
1281 if (it == NULL) {
1282 Py_DECREF(result);
1283 return NULL;
1284 }
Raymond Hettingera690a992003-11-16 16:17:49 +00001285
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001286 while ((key = PyIter_Next(it)) != NULL) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001287 hash = PyObject_Hash(key);
Raymond Hettinger11ce8e62015-07-06 19:08:49 -07001288 if (hash == -1)
1289 goto error;
Raymond Hettinger73799b12015-07-05 16:06:10 -07001290 rv = set_contains_entry(so, key, hash);
Raymond Hettinger11ce8e62015-07-06 19:08:49 -07001291 if (rv < 0)
1292 goto error;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001293 if (rv) {
Raymond Hettinger11ce8e62015-07-06 19:08:49 -07001294 if (set_add_entry(result, key, hash))
1295 goto error;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001296 }
1297 Py_DECREF(key);
1298 }
1299 Py_DECREF(it);
1300 if (PyErr_Occurred()) {
1301 Py_DECREF(result);
1302 return NULL;
1303 }
1304 return (PyObject *)result;
Raymond Hettinger11ce8e62015-07-06 19:08:49 -07001305 error:
1306 Py_DECREF(it);
1307 Py_DECREF(result);
1308 Py_DECREF(key);
1309 return NULL;
Raymond Hettingera690a992003-11-16 16:17:49 +00001310}
1311
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001312static PyObject *
1313set_intersection_multi(PySetObject *so, PyObject *args)
1314{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001315 Py_ssize_t i;
1316 PyObject *result = (PyObject *)so;
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001317
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001318 if (PyTuple_GET_SIZE(args) == 0)
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301319 return set_copy(so, NULL);
Amaury Forgeot d'Arcfdfe62d2008-06-17 20:36:03 +00001320
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001321 Py_INCREF(so);
1322 for (i=0 ; i<PyTuple_GET_SIZE(args) ; i++) {
1323 PyObject *other = PyTuple_GET_ITEM(args, i);
1324 PyObject *newresult = set_intersection((PySetObject *)result, other);
1325 if (newresult == NULL) {
1326 Py_DECREF(result);
1327 return NULL;
1328 }
1329 Py_DECREF(result);
1330 result = newresult;
1331 }
1332 return result;
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001333}
1334
Raymond Hettingera690a992003-11-16 16:17:49 +00001335PyDoc_STRVAR(intersection_doc,
1336"Return the intersection of two sets as a new set.\n\
1337\n\
1338(i.e. all elements that are in both sets.)");
1339
1340static PyObject *
1341set_intersection_update(PySetObject *so, PyObject *other)
1342{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001343 PyObject *tmp;
Raymond Hettingera690a992003-11-16 16:17:49 +00001344
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001345 tmp = set_intersection(so, other);
1346 if (tmp == NULL)
1347 return NULL;
1348 set_swap_bodies(so, (PySetObject *)tmp);
1349 Py_DECREF(tmp);
1350 Py_RETURN_NONE;
Raymond Hettingera690a992003-11-16 16:17:49 +00001351}
1352
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001353static PyObject *
1354set_intersection_update_multi(PySetObject *so, PyObject *args)
1355{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001356 PyObject *tmp;
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001357
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001358 tmp = set_intersection_multi(so, args);
1359 if (tmp == NULL)
1360 return NULL;
1361 set_swap_bodies(so, (PySetObject *)tmp);
1362 Py_DECREF(tmp);
1363 Py_RETURN_NONE;
Georg Brandlc28e1fa2008-06-10 19:20:26 +00001364}
1365
Raymond Hettingera690a992003-11-16 16:17:49 +00001366PyDoc_STRVAR(intersection_update_doc,
1367"Update a set with the intersection of itself and another.");
1368
1369static PyObject *
1370set_and(PySetObject *so, PyObject *other)
1371{
Brian Curtindfc80e32011-08-10 20:28:54 -05001372 if (!PyAnySet_Check(so) || !PyAnySet_Check(other))
1373 Py_RETURN_NOTIMPLEMENTED;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001374 return set_intersection(so, other);
Raymond Hettingera690a992003-11-16 16:17:49 +00001375}
1376
1377static PyObject *
1378set_iand(PySetObject *so, PyObject *other)
1379{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001380 PyObject *result;
Raymond Hettingera690a992003-11-16 16:17:49 +00001381
Brian Curtindfc80e32011-08-10 20:28:54 -05001382 if (!PyAnySet_Check(other))
1383 Py_RETURN_NOTIMPLEMENTED;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001384 result = set_intersection_update(so, other);
1385 if (result == NULL)
1386 return NULL;
1387 Py_DECREF(result);
1388 Py_INCREF(so);
1389 return (PyObject *)so;
Raymond Hettingera690a992003-11-16 16:17:49 +00001390}
1391
Guido van Rossum58da9312007-11-10 23:39:45 +00001392static PyObject *
1393set_isdisjoint(PySetObject *so, PyObject *other)
1394{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001395 PyObject *key, *it, *tmp;
Raymond Hettinger73799b12015-07-05 16:06:10 -07001396 int rv;
Guido van Rossum58da9312007-11-10 23:39:45 +00001397
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001398 if ((PyObject *)so == other) {
1399 if (PySet_GET_SIZE(so) == 0)
1400 Py_RETURN_TRUE;
1401 else
1402 Py_RETURN_FALSE;
1403 }
Guido van Rossum58da9312007-11-10 23:39:45 +00001404
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001405 if (PyAnySet_CheckExact(other)) {
1406 Py_ssize_t pos = 0;
1407 setentry *entry;
Guido van Rossum58da9312007-11-10 23:39:45 +00001408
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001409 if (PySet_GET_SIZE(other) > PySet_GET_SIZE(so)) {
1410 tmp = (PyObject *)so;
1411 so = (PySetObject *)other;
1412 other = tmp;
1413 }
1414 while (set_next((PySetObject *)other, &pos, &entry)) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001415 rv = set_contains_entry(so, entry->key, entry->hash);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001416 if (rv < 0)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001417 return NULL;
1418 if (rv)
1419 Py_RETURN_FALSE;
1420 }
1421 Py_RETURN_TRUE;
1422 }
Guido van Rossum58da9312007-11-10 23:39:45 +00001423
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001424 it = PyObject_GetIter(other);
1425 if (it == NULL)
1426 return NULL;
Guido van Rossum58da9312007-11-10 23:39:45 +00001427
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001428 while ((key = PyIter_Next(it)) != NULL) {
Benjamin Peterson8f67d082010-10-17 20:54:53 +00001429 Py_hash_t hash = PyObject_Hash(key);
Guido van Rossum58da9312007-11-10 23:39:45 +00001430
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001431 if (hash == -1) {
1432 Py_DECREF(key);
1433 Py_DECREF(it);
1434 return NULL;
1435 }
Raymond Hettinger73799b12015-07-05 16:06:10 -07001436 rv = set_contains_entry(so, key, hash);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001437 Py_DECREF(key);
Raymond Hettingerc2480dc2015-07-04 08:46:31 -07001438 if (rv < 0) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001439 Py_DECREF(it);
1440 return NULL;
1441 }
1442 if (rv) {
1443 Py_DECREF(it);
1444 Py_RETURN_FALSE;
1445 }
1446 }
1447 Py_DECREF(it);
1448 if (PyErr_Occurred())
1449 return NULL;
1450 Py_RETURN_TRUE;
Guido van Rossum58da9312007-11-10 23:39:45 +00001451}
1452
1453PyDoc_STRVAR(isdisjoint_doc,
1454"Return True if two sets have a null intersection.");
1455
Neal Norwitz6576bd82005-11-13 18:41:28 +00001456static int
Raymond Hettingerc991db22005-08-11 07:58:45 +00001457set_difference_update_internal(PySetObject *so, PyObject *other)
1458{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001459 if ((PyObject *)so == other)
1460 return set_clear_internal(so);
Raymond Hettingerc991db22005-08-11 07:58:45 +00001461
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001462 if (PyAnySet_Check(other)) {
1463 setentry *entry;
1464 Py_ssize_t pos = 0;
Raymond Hettingerc991db22005-08-11 07:58:45 +00001465
Raymond Hettinger88ea1662019-08-29 09:02:58 -07001466 /* Optimization: When the other set is more than 8 times
1467 larger than the base set, replace the other set with
1468 interesection of the two sets.
1469 */
1470 if ((PySet_GET_SIZE(other) >> 3) > PySet_GET_SIZE(so)) {
1471 other = set_intersection(so, other);
1472 if (other == NULL)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001473 return -1;
Raymond Hettinger88ea1662019-08-29 09:02:58 -07001474 } else {
1475 Py_INCREF(other);
1476 }
1477
1478 while (set_next((PySetObject *)other, &pos, &entry))
1479 if (set_discard_entry(so, entry->key, entry->hash) < 0) {
1480 Py_DECREF(other);
1481 return -1;
1482 }
1483
1484 Py_DECREF(other);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001485 } else {
1486 PyObject *key, *it;
1487 it = PyObject_GetIter(other);
1488 if (it == NULL)
1489 return -1;
1490
1491 while ((key = PyIter_Next(it)) != NULL) {
Raymond Hettingerb3223262015-07-03 23:37:16 -07001492 if (set_discard_key(so, key) < 0) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001493 Py_DECREF(it);
1494 Py_DECREF(key);
1495 return -1;
1496 }
1497 Py_DECREF(key);
1498 }
1499 Py_DECREF(it);
1500 if (PyErr_Occurred())
1501 return -1;
1502 }
Raymond Hettingere186c762015-07-04 11:28:35 -07001503 /* If more than 1/4th are dummies, then resize them away. */
1504 if ((size_t)(so->fill - so->used) <= (size_t)so->mask / 4)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001505 return 0;
INADA Naokie82cf862017-04-01 17:20:25 +09001506 return set_table_resize(so, so->used>50000 ? so->used*2 : so->used*4);
Raymond Hettingerc991db22005-08-11 07:58:45 +00001507}
1508
Raymond Hettingera690a992003-11-16 16:17:49 +00001509static PyObject *
Amaury Forgeot d'Arcfdfe62d2008-06-17 20:36:03 +00001510set_difference_update(PySetObject *so, PyObject *args)
Raymond Hettingera690a992003-11-16 16:17:49 +00001511{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001512 Py_ssize_t i;
Amaury Forgeot d'Arcfdfe62d2008-06-17 20:36:03 +00001513
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001514 for (i=0 ; i<PyTuple_GET_SIZE(args) ; i++) {
1515 PyObject *other = PyTuple_GET_ITEM(args, i);
Raymond Hettinger5af9e132015-05-13 01:44:36 -07001516 if (set_difference_update_internal(so, other))
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001517 return NULL;
1518 }
1519 Py_RETURN_NONE;
Raymond Hettingera690a992003-11-16 16:17:49 +00001520}
1521
1522PyDoc_STRVAR(difference_update_doc,
1523"Remove all elements of another set from this set.");
1524
1525static PyObject *
Antoine Pitrou715f3cd2010-11-30 22:23:20 +00001526set_copy_and_difference(PySetObject *so, PyObject *other)
1527{
1528 PyObject *result;
1529
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301530 result = set_copy(so, NULL);
Antoine Pitrou715f3cd2010-11-30 22:23:20 +00001531 if (result == NULL)
1532 return NULL;
Raymond Hettingerb3223262015-07-03 23:37:16 -07001533 if (set_difference_update_internal((PySetObject *) result, other) == 0)
Antoine Pitrou715f3cd2010-11-30 22:23:20 +00001534 return result;
1535 Py_DECREF(result);
1536 return NULL;
1537}
1538
1539static PyObject *
Raymond Hettingerfb4e33a2003-12-15 13:23:55 +00001540set_difference(PySetObject *so, PyObject *other)
1541{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001542 PyObject *result;
Raymond Hettinger73799b12015-07-05 16:06:10 -07001543 PyObject *key;
1544 Py_hash_t hash;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001545 setentry *entry;
Serhiy Storchakabf623ae2017-04-19 20:03:52 +03001546 Py_ssize_t pos = 0, other_size;
Raymond Hettinger73799b12015-07-05 16:06:10 -07001547 int rv;
Raymond Hettingerfb4e33a2003-12-15 13:23:55 +00001548
Serhiy Storchakabf623ae2017-04-19 20:03:52 +03001549 if (PyAnySet_Check(other)) {
1550 other_size = PySet_GET_SIZE(other);
1551 }
1552 else if (PyDict_CheckExact(other)) {
1553 other_size = PyDict_GET_SIZE(other);
1554 }
1555 else {
Antoine Pitrou715f3cd2010-11-30 22:23:20 +00001556 return set_copy_and_difference(so, other);
1557 }
1558
1559 /* If len(so) much more than len(other), it's more efficient to simply copy
1560 * so and then iterate other looking for common elements. */
Serhiy Storchakabf623ae2017-04-19 20:03:52 +03001561 if ((PySet_GET_SIZE(so) >> 2) > other_size) {
Antoine Pitrou715f3cd2010-11-30 22:23:20 +00001562 return set_copy_and_difference(so, other);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001563 }
Raymond Hettingerfb4e33a2003-12-15 13:23:55 +00001564
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001565 result = make_new_set_basetype(Py_TYPE(so), NULL);
1566 if (result == NULL)
1567 return NULL;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +00001568
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001569 if (PyDict_CheckExact(other)) {
1570 while (set_next(so, &pos, &entry)) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001571 key = entry->key;
1572 hash = entry->hash;
Raymond Hettinger15f08692015-07-03 17:21:17 -07001573 rv = _PyDict_Contains(other, key, hash);
Serhiy Storchakafa494fd2015-05-30 17:45:22 +03001574 if (rv < 0) {
1575 Py_DECREF(result);
1576 return NULL;
1577 }
1578 if (!rv) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001579 if (set_add_entry((PySetObject *)result, key, hash)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001580 Py_DECREF(result);
1581 return NULL;
1582 }
1583 }
1584 }
1585 return result;
1586 }
1587
Antoine Pitrou715f3cd2010-11-30 22:23:20 +00001588 /* Iterate over so, checking for common elements in other. */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001589 while (set_next(so, &pos, &entry)) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001590 key = entry->key;
1591 hash = entry->hash;
1592 rv = set_contains_entry((PySetObject *)other, key, hash);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001593 if (rv < 0) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001594 Py_DECREF(result);
1595 return NULL;
1596 }
1597 if (!rv) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001598 if (set_add_entry((PySetObject *)result, key, hash)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001599 Py_DECREF(result);
1600 return NULL;
1601 }
1602 }
1603 }
1604 return result;
Raymond Hettingerfb4e33a2003-12-15 13:23:55 +00001605}
1606
Amaury Forgeot d'Arcfdfe62d2008-06-17 20:36:03 +00001607static PyObject *
1608set_difference_multi(PySetObject *so, PyObject *args)
1609{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001610 Py_ssize_t i;
1611 PyObject *result, *other;
Amaury Forgeot d'Arcfdfe62d2008-06-17 20:36:03 +00001612
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001613 if (PyTuple_GET_SIZE(args) == 0)
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301614 return set_copy(so, NULL);
Amaury Forgeot d'Arcfdfe62d2008-06-17 20:36:03 +00001615
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001616 other = PyTuple_GET_ITEM(args, 0);
1617 result = set_difference(so, other);
1618 if (result == NULL)
1619 return NULL;
Amaury Forgeot d'Arcfdfe62d2008-06-17 20:36:03 +00001620
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001621 for (i=1 ; i<PyTuple_GET_SIZE(args) ; i++) {
1622 other = PyTuple_GET_ITEM(args, i);
Raymond Hettinger5af9e132015-05-13 01:44:36 -07001623 if (set_difference_update_internal((PySetObject *)result, other)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001624 Py_DECREF(result);
1625 return NULL;
1626 }
1627 }
1628 return result;
Amaury Forgeot d'Arcfdfe62d2008-06-17 20:36:03 +00001629}
1630
Raymond Hettingerfb4e33a2003-12-15 13:23:55 +00001631PyDoc_STRVAR(difference_doc,
Amaury Forgeot d'Arcfdfe62d2008-06-17 20:36:03 +00001632"Return the difference of two or more sets as a new set.\n\
Raymond Hettingerfb4e33a2003-12-15 13:23:55 +00001633\n\
Amaury Forgeot d'Arcfdfe62d2008-06-17 20:36:03 +00001634(i.e. all elements that are in this set but not the others.)");
Raymond Hettingerfb4e33a2003-12-15 13:23:55 +00001635static PyObject *
Raymond Hettingera690a992003-11-16 16:17:49 +00001636set_sub(PySetObject *so, PyObject *other)
1637{
Brian Curtindfc80e32011-08-10 20:28:54 -05001638 if (!PyAnySet_Check(so) || !PyAnySet_Check(other))
1639 Py_RETURN_NOTIMPLEMENTED;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001640 return set_difference(so, other);
Raymond Hettingera690a992003-11-16 16:17:49 +00001641}
1642
1643static PyObject *
1644set_isub(PySetObject *so, PyObject *other)
1645{
Brian Curtindfc80e32011-08-10 20:28:54 -05001646 if (!PyAnySet_Check(other))
1647 Py_RETURN_NOTIMPLEMENTED;
Raymond Hettinger5af9e132015-05-13 01:44:36 -07001648 if (set_difference_update_internal(so, other))
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001649 return NULL;
1650 Py_INCREF(so);
1651 return (PyObject *)so;
Raymond Hettingera690a992003-11-16 16:17:49 +00001652}
1653
1654static PyObject *
Raymond Hettingera690a992003-11-16 16:17:49 +00001655set_symmetric_difference_update(PySetObject *so, PyObject *other)
1656{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001657 PySetObject *otherset;
1658 PyObject *key;
1659 Py_ssize_t pos = 0;
Raymond Hettinger73799b12015-07-05 16:06:10 -07001660 Py_hash_t hash;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001661 setentry *entry;
Raymond Hettinger73799b12015-07-05 16:06:10 -07001662 int rv;
Raymond Hettingerc991db22005-08-11 07:58:45 +00001663
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001664 if ((PyObject *)so == other)
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301665 return set_clear(so, NULL);
Raymond Hettingera690a992003-11-16 16:17:49 +00001666
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001667 if (PyDict_CheckExact(other)) {
1668 PyObject *value;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001669 while (_PyDict_Next(other, &pos, &key, &value, &hash)) {
Raymond Hettingerfaf7b7f2010-09-03 10:00:50 +00001670 Py_INCREF(key);
Raymond Hettinger73799b12015-07-05 16:06:10 -07001671 rv = set_discard_entry(so, key, hash);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001672 if (rv < 0) {
Raymond Hettingerfaf7b7f2010-09-03 10:00:50 +00001673 Py_DECREF(key);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001674 return NULL;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001675 }
Raymond Hettingerfaf7b7f2010-09-03 10:00:50 +00001676 if (rv == DISCARD_NOTFOUND) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001677 if (set_add_entry(so, key, hash)) {
Raymond Hettingerfaf7b7f2010-09-03 10:00:50 +00001678 Py_DECREF(key);
1679 return NULL;
1680 }
1681 }
Georg Brandl2d444492010-09-03 10:52:55 +00001682 Py_DECREF(key);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001683 }
1684 Py_RETURN_NONE;
1685 }
Raymond Hettinger9f1a6792005-07-31 01:16:36 +00001686
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001687 if (PyAnySet_Check(other)) {
1688 Py_INCREF(other);
1689 otherset = (PySetObject *)other;
1690 } else {
1691 otherset = (PySetObject *)make_new_set_basetype(Py_TYPE(so), other);
1692 if (otherset == NULL)
1693 return NULL;
1694 }
Raymond Hettingera690a992003-11-16 16:17:49 +00001695
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001696 while (set_next(otherset, &pos, &entry)) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001697 key = entry->key;
1698 hash = entry->hash;
1699 rv = set_discard_entry(so, key, hash);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001700 if (rv < 0) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001701 Py_DECREF(otherset);
1702 return NULL;
1703 }
1704 if (rv == DISCARD_NOTFOUND) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001705 if (set_add_entry(so, key, hash)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001706 Py_DECREF(otherset);
1707 return NULL;
1708 }
1709 }
1710 }
1711 Py_DECREF(otherset);
1712 Py_RETURN_NONE;
Raymond Hettingera690a992003-11-16 16:17:49 +00001713}
1714
1715PyDoc_STRVAR(symmetric_difference_update_doc,
1716"Update a set with the symmetric difference of itself and another.");
1717
1718static PyObject *
Raymond Hettingerf5f41bf2003-11-24 02:57:33 +00001719set_symmetric_difference(PySetObject *so, PyObject *other)
1720{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001721 PyObject *rv;
1722 PySetObject *otherset;
Raymond Hettingerf5f41bf2003-11-24 02:57:33 +00001723
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001724 otherset = (PySetObject *)make_new_set_basetype(Py_TYPE(so), other);
1725 if (otherset == NULL)
1726 return NULL;
1727 rv = set_symmetric_difference_update(otherset, (PyObject *)so);
lekma491bbed2018-05-02 11:29:10 +02001728 if (rv == NULL) {
1729 Py_DECREF(otherset);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001730 return NULL;
lekma491bbed2018-05-02 11:29:10 +02001731 }
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001732 Py_DECREF(rv);
1733 return (PyObject *)otherset;
Raymond Hettingerf5f41bf2003-11-24 02:57:33 +00001734}
1735
1736PyDoc_STRVAR(symmetric_difference_doc,
1737"Return the symmetric difference of two sets as a new set.\n\
1738\n\
1739(i.e. all elements that are in exactly one of the sets.)");
1740
1741static PyObject *
Raymond Hettingera690a992003-11-16 16:17:49 +00001742set_xor(PySetObject *so, PyObject *other)
1743{
Brian Curtindfc80e32011-08-10 20:28:54 -05001744 if (!PyAnySet_Check(so) || !PyAnySet_Check(other))
1745 Py_RETURN_NOTIMPLEMENTED;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001746 return set_symmetric_difference(so, other);
Raymond Hettingera690a992003-11-16 16:17:49 +00001747}
1748
1749static PyObject *
1750set_ixor(PySetObject *so, PyObject *other)
1751{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001752 PyObject *result;
Raymond Hettingera690a992003-11-16 16:17:49 +00001753
Brian Curtindfc80e32011-08-10 20:28:54 -05001754 if (!PyAnySet_Check(other))
1755 Py_RETURN_NOTIMPLEMENTED;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001756 result = set_symmetric_difference_update(so, other);
1757 if (result == NULL)
1758 return NULL;
1759 Py_DECREF(result);
1760 Py_INCREF(so);
1761 return (PyObject *)so;
Raymond Hettingera690a992003-11-16 16:17:49 +00001762}
1763
1764static PyObject *
1765set_issubset(PySetObject *so, PyObject *other)
1766{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001767 setentry *entry;
1768 Py_ssize_t pos = 0;
Raymond Hettinger73799b12015-07-05 16:06:10 -07001769 int rv;
Raymond Hettingera690a992003-11-16 16:17:49 +00001770
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001771 if (!PyAnySet_Check(other)) {
1772 PyObject *tmp, *result;
1773 tmp = make_new_set(&PySet_Type, other);
1774 if (tmp == NULL)
1775 return NULL;
1776 result = set_issubset(so, tmp);
1777 Py_DECREF(tmp);
1778 return result;
1779 }
1780 if (PySet_GET_SIZE(so) > PySet_GET_SIZE(other))
1781 Py_RETURN_FALSE;
Raymond Hettingera690a992003-11-16 16:17:49 +00001782
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001783 while (set_next(so, &pos, &entry)) {
Raymond Hettinger73799b12015-07-05 16:06:10 -07001784 rv = set_contains_entry((PySetObject *)other, entry->key, entry->hash);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001785 if (rv < 0)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001786 return NULL;
1787 if (!rv)
1788 Py_RETURN_FALSE;
1789 }
1790 Py_RETURN_TRUE;
Raymond Hettingera690a992003-11-16 16:17:49 +00001791}
1792
1793PyDoc_STRVAR(issubset_doc, "Report whether another set contains this set.");
1794
1795static PyObject *
1796set_issuperset(PySetObject *so, PyObject *other)
1797{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001798 PyObject *tmp, *result;
Raymond Hettinger3fbec702003-11-21 07:56:36 +00001799
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001800 if (!PyAnySet_Check(other)) {
1801 tmp = make_new_set(&PySet_Type, other);
1802 if (tmp == NULL)
1803 return NULL;
1804 result = set_issuperset(so, tmp);
1805 Py_DECREF(tmp);
1806 return result;
1807 }
1808 return set_issubset((PySetObject *)other, (PyObject *)so);
Raymond Hettingera690a992003-11-16 16:17:49 +00001809}
1810
1811PyDoc_STRVAR(issuperset_doc, "Report whether this set contains another set.");
1812
Raymond Hettingera690a992003-11-16 16:17:49 +00001813static PyObject *
1814set_richcompare(PySetObject *v, PyObject *w, int op)
1815{
Serhiy Storchakafa494fd2015-05-30 17:45:22 +03001816 PyObject *r1;
1817 int r2;
Raymond Hettinger9f1a6792005-07-31 01:16:36 +00001818
Brian Curtindfc80e32011-08-10 20:28:54 -05001819 if(!PyAnySet_Check(w))
1820 Py_RETURN_NOTIMPLEMENTED;
1821
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001822 switch (op) {
1823 case Py_EQ:
1824 if (PySet_GET_SIZE(v) != PySet_GET_SIZE(w))
1825 Py_RETURN_FALSE;
1826 if (v->hash != -1 &&
1827 ((PySetObject *)w)->hash != -1 &&
1828 v->hash != ((PySetObject *)w)->hash)
1829 Py_RETURN_FALSE;
1830 return set_issubset(v, w);
1831 case Py_NE:
1832 r1 = set_richcompare(v, w, Py_EQ);
1833 if (r1 == NULL)
1834 return NULL;
Serhiy Storchakafa494fd2015-05-30 17:45:22 +03001835 r2 = PyObject_IsTrue(r1);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001836 Py_DECREF(r1);
Serhiy Storchakafa494fd2015-05-30 17:45:22 +03001837 if (r2 < 0)
1838 return NULL;
1839 return PyBool_FromLong(!r2);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001840 case Py_LE:
1841 return set_issubset(v, w);
1842 case Py_GE:
1843 return set_issuperset(v, w);
1844 case Py_LT:
1845 if (PySet_GET_SIZE(v) >= PySet_GET_SIZE(w))
1846 Py_RETURN_FALSE;
1847 return set_issubset(v, w);
1848 case Py_GT:
1849 if (PySet_GET_SIZE(v) <= PySet_GET_SIZE(w))
1850 Py_RETURN_FALSE;
1851 return set_issuperset(v, w);
1852 }
Brian Curtindfc80e32011-08-10 20:28:54 -05001853 Py_RETURN_NOTIMPLEMENTED;
Raymond Hettingera690a992003-11-16 16:17:49 +00001854}
1855
Raymond Hettingera690a992003-11-16 16:17:49 +00001856static PyObject *
Raymond Hettinger06d8cf82005-07-31 15:36:06 +00001857set_add(PySetObject *so, PyObject *key)
Raymond Hettingera690a992003-11-16 16:17:49 +00001858{
Raymond Hettinger5af9e132015-05-13 01:44:36 -07001859 if (set_add_key(so, key))
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001860 return NULL;
1861 Py_RETURN_NONE;
Raymond Hettingera690a992003-11-16 16:17:49 +00001862}
1863
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001864PyDoc_STRVAR(add_doc,
Raymond Hettingera690a992003-11-16 16:17:49 +00001865"Add an element to a set.\n\
1866\n\
1867This has no effect if the element is already present.");
1868
Raymond Hettingerce8185e2005-08-13 09:28:48 +00001869static int
1870set_contains(PySetObject *so, PyObject *key)
1871{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001872 PyObject *tmpkey;
1873 int rv;
Raymond Hettingerce8185e2005-08-13 09:28:48 +00001874
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001875 rv = set_contains_key(so, key);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001876 if (rv < 0) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001877 if (!PySet_Check(key) || !PyErr_ExceptionMatches(PyExc_TypeError))
1878 return -1;
1879 PyErr_Clear();
Raymond Hettinger38bf2cc2010-08-06 09:52:17 +00001880 tmpkey = make_new_set(&PyFrozenSet_Type, key);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001881 if (tmpkey == NULL)
1882 return -1;
Petri Lehtinen5acc27e2011-10-30 13:56:41 +02001883 rv = set_contains_key(so, tmpkey);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001884 Py_DECREF(tmpkey);
1885 }
1886 return rv;
Raymond Hettingerce8185e2005-08-13 09:28:48 +00001887}
1888
1889static PyObject *
1890set_direct_contains(PySetObject *so, PyObject *key)
1891{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001892 long result;
Raymond Hettingerce8185e2005-08-13 09:28:48 +00001893
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001894 result = set_contains(so, key);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001895 if (result < 0)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001896 return NULL;
1897 return PyBool_FromLong(result);
Raymond Hettingerce8185e2005-08-13 09:28:48 +00001898}
1899
1900PyDoc_STRVAR(contains_doc, "x.__contains__(y) <==> y in x.");
1901
Raymond Hettingera690a992003-11-16 16:17:49 +00001902static PyObject *
Raymond Hettinger06d8cf82005-07-31 15:36:06 +00001903set_remove(PySetObject *so, PyObject *key)
Raymond Hettingera690a992003-11-16 16:17:49 +00001904{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001905 PyObject *tmpkey;
1906 int rv;
Raymond Hettingerbfd334a2003-11-22 03:55:23 +00001907
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001908 rv = set_discard_key(so, key);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001909 if (rv < 0) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001910 if (!PySet_Check(key) || !PyErr_ExceptionMatches(PyExc_TypeError))
1911 return NULL;
1912 PyErr_Clear();
Raymond Hettinger38bf2cc2010-08-06 09:52:17 +00001913 tmpkey = make_new_set(&PyFrozenSet_Type, key);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001914 if (tmpkey == NULL)
1915 return NULL;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001916 rv = set_discard_key(so, tmpkey);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001917 Py_DECREF(tmpkey);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001918 if (rv < 0)
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001919 return NULL;
1920 }
Benjamin Petersonf10a79a2008-10-11 00:49:57 +00001921
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001922 if (rv == DISCARD_NOTFOUND) {
Raymond Hettinger69492da2013-09-02 15:59:26 -07001923 _PyErr_SetKeyError(key);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001924 return NULL;
1925 }
1926 Py_RETURN_NONE;
Raymond Hettingera690a992003-11-16 16:17:49 +00001927}
1928
1929PyDoc_STRVAR(remove_doc,
1930"Remove an element from a set; it must be a member.\n\
1931\n\
1932If the element is not a member, raise a KeyError.");
1933
1934static PyObject *
Raymond Hettinger06d8cf82005-07-31 15:36:06 +00001935set_discard(PySetObject *so, PyObject *key)
Raymond Hettingera690a992003-11-16 16:17:49 +00001936{
Benjamin Peterson2b50a012011-10-30 14:24:44 -04001937 PyObject *tmpkey;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001938 int rv;
Raymond Hettinger0deab622003-12-13 18:53:18 +00001939
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001940 rv = set_discard_key(so, key);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001941 if (rv < 0) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001942 if (!PySet_Check(key) || !PyErr_ExceptionMatches(PyExc_TypeError))
1943 return NULL;
1944 PyErr_Clear();
Raymond Hettinger38bf2cc2010-08-06 09:52:17 +00001945 tmpkey = make_new_set(&PyFrozenSet_Type, key);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001946 if (tmpkey == NULL)
1947 return NULL;
Petri Lehtinene0aa8032011-10-30 14:31:27 +02001948 rv = set_discard_key(so, tmpkey);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001949 Py_DECREF(tmpkey);
Raymond Hettingerb3223262015-07-03 23:37:16 -07001950 if (rv < 0)
Petri Lehtinene0aa8032011-10-30 14:31:27 +02001951 return NULL;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001952 }
1953 Py_RETURN_NONE;
Raymond Hettingera690a992003-11-16 16:17:49 +00001954}
1955
1956PyDoc_STRVAR(discard_doc,
1957"Remove an element from a set if it is a member.\n\
1958\n\
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001959If the element is not a member, do nothing.");
Raymond Hettingera690a992003-11-16 16:17:49 +00001960
1961static PyObject *
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301962set_reduce(PySetObject *so, PyObject *Py_UNUSED(ignored))
Raymond Hettingera690a992003-11-16 16:17:49 +00001963{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001964 PyObject *keys=NULL, *args=NULL, *result=NULL, *dict=NULL;
Martin v. Löwisbd928fe2011-10-14 10:20:37 +02001965 _Py_IDENTIFIER(__dict__);
Raymond Hettingera690a992003-11-16 16:17:49 +00001966
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001967 keys = PySequence_List((PyObject *)so);
1968 if (keys == NULL)
1969 goto done;
1970 args = PyTuple_Pack(1, keys);
1971 if (args == NULL)
1972 goto done;
Serhiy Storchaka41c57b32019-09-01 12:03:39 +03001973 if (_PyObject_LookupAttrId((PyObject *)so, &PyId___dict__, &dict) < 0) {
1974 goto done;
1975 }
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001976 if (dict == NULL) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001977 dict = Py_None;
1978 Py_INCREF(dict);
1979 }
1980 result = PyTuple_Pack(3, Py_TYPE(so), args, dict);
Raymond Hettingera690a992003-11-16 16:17:49 +00001981done:
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001982 Py_XDECREF(args);
1983 Py_XDECREF(keys);
1984 Py_XDECREF(dict);
1985 return result;
Raymond Hettingera690a992003-11-16 16:17:49 +00001986}
1987
Robert Schuppeniesfbe94c52008-07-14 10:13:31 +00001988static PyObject *
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05301989set_sizeof(PySetObject *so, PyObject *Py_UNUSED(ignored))
Robert Schuppeniesfbe94c52008-07-14 10:13:31 +00001990{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001991 Py_ssize_t res;
Robert Schuppeniesfbe94c52008-07-14 10:13:31 +00001992
Serhiy Storchaka5c4064e2015-12-19 20:05:25 +02001993 res = _PyObject_SIZE(Py_TYPE(so));
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00001994 if (so->table != so->smalltable)
1995 res = res + (so->mask + 1) * sizeof(setentry);
1996 return PyLong_FromSsize_t(res);
Robert Schuppeniesfbe94c52008-07-14 10:13:31 +00001997}
1998
1999PyDoc_STRVAR(sizeof_doc, "S.__sizeof__() -> size of S in memory, in bytes");
Raymond Hettinger50a4bb32003-11-17 16:42:33 +00002000static int
2001set_init(PySetObject *self, PyObject *args, PyObject *kwds)
2002{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002003 PyObject *iterable = NULL;
Raymond Hettinger50a4bb32003-11-17 16:42:33 +00002004
Serhiy Storchaka6cca5c82017-06-08 14:41:19 +03002005 if (!_PyArg_NoKeywords("set", kwds))
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002006 return -1;
2007 if (!PyArg_UnpackTuple(args, Py_TYPE(self)->tp_name, 0, 1, &iterable))
2008 return -1;
Raymond Hettingerb72e21b2016-03-25 02:29:59 -07002009 if (self->fill)
2010 set_clear_internal(self);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002011 self->hash = -1;
2012 if (iterable == NULL)
2013 return 0;
2014 return set_update_internal(self, iterable);
Raymond Hettinger50a4bb32003-11-17 16:42:33 +00002015}
2016
Raymond Hettingera690a992003-11-16 16:17:49 +00002017static PySequenceMethods set_as_sequence = {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002018 set_len, /* sq_length */
2019 0, /* sq_concat */
2020 0, /* sq_repeat */
2021 0, /* sq_item */
2022 0, /* sq_slice */
2023 0, /* sq_ass_item */
2024 0, /* sq_ass_slice */
2025 (objobjproc)set_contains, /* sq_contains */
Raymond Hettingera690a992003-11-16 16:17:49 +00002026};
2027
2028/* set object ********************************************************/
2029
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002030#ifdef Py_DEBUG
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05302031static PyObject *test_c_api(PySetObject *so, PyObject *Py_UNUSED(ignored));
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002032
2033PyDoc_STRVAR(test_c_api_doc, "Exercises C API. Returns True.\n\
2034All is well if assertions don't fail.");
2035#endif
2036
Raymond Hettingera690a992003-11-16 16:17:49 +00002037static PyMethodDef set_methods[] = {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002038 {"add", (PyCFunction)set_add, METH_O,
2039 add_doc},
2040 {"clear", (PyCFunction)set_clear, METH_NOARGS,
2041 clear_doc},
2042 {"__contains__",(PyCFunction)set_direct_contains, METH_O | METH_COEXIST,
2043 contains_doc},
2044 {"copy", (PyCFunction)set_copy, METH_NOARGS,
2045 copy_doc},
2046 {"discard", (PyCFunction)set_discard, METH_O,
2047 discard_doc},
2048 {"difference", (PyCFunction)set_difference_multi, METH_VARARGS,
2049 difference_doc},
2050 {"difference_update", (PyCFunction)set_difference_update, METH_VARARGS,
2051 difference_update_doc},
2052 {"intersection",(PyCFunction)set_intersection_multi, METH_VARARGS,
2053 intersection_doc},
2054 {"intersection_update",(PyCFunction)set_intersection_update_multi, METH_VARARGS,
2055 intersection_update_doc},
2056 {"isdisjoint", (PyCFunction)set_isdisjoint, METH_O,
2057 isdisjoint_doc},
2058 {"issubset", (PyCFunction)set_issubset, METH_O,
2059 issubset_doc},
2060 {"issuperset", (PyCFunction)set_issuperset, METH_O,
2061 issuperset_doc},
2062 {"pop", (PyCFunction)set_pop, METH_NOARGS,
2063 pop_doc},
2064 {"__reduce__", (PyCFunction)set_reduce, METH_NOARGS,
2065 reduce_doc},
2066 {"remove", (PyCFunction)set_remove, METH_O,
2067 remove_doc},
2068 {"__sizeof__", (PyCFunction)set_sizeof, METH_NOARGS,
2069 sizeof_doc},
2070 {"symmetric_difference",(PyCFunction)set_symmetric_difference, METH_O,
2071 symmetric_difference_doc},
2072 {"symmetric_difference_update",(PyCFunction)set_symmetric_difference_update, METH_O,
2073 symmetric_difference_update_doc},
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002074#ifdef Py_DEBUG
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002075 {"test_c_api", (PyCFunction)test_c_api, METH_NOARGS,
2076 test_c_api_doc},
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002077#endif
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002078 {"union", (PyCFunction)set_union, METH_VARARGS,
2079 union_doc},
2080 {"update", (PyCFunction)set_update, METH_VARARGS,
2081 update_doc},
2082 {NULL, NULL} /* sentinel */
Raymond Hettingera690a992003-11-16 16:17:49 +00002083};
2084
2085static PyNumberMethods set_as_number = {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002086 0, /*nb_add*/
2087 (binaryfunc)set_sub, /*nb_subtract*/
2088 0, /*nb_multiply*/
2089 0, /*nb_remainder*/
2090 0, /*nb_divmod*/
2091 0, /*nb_power*/
2092 0, /*nb_negative*/
2093 0, /*nb_positive*/
2094 0, /*nb_absolute*/
2095 0, /*nb_bool*/
2096 0, /*nb_invert*/
2097 0, /*nb_lshift*/
2098 0, /*nb_rshift*/
2099 (binaryfunc)set_and, /*nb_and*/
2100 (binaryfunc)set_xor, /*nb_xor*/
2101 (binaryfunc)set_or, /*nb_or*/
2102 0, /*nb_int*/
2103 0, /*nb_reserved*/
2104 0, /*nb_float*/
2105 0, /*nb_inplace_add*/
2106 (binaryfunc)set_isub, /*nb_inplace_subtract*/
2107 0, /*nb_inplace_multiply*/
2108 0, /*nb_inplace_remainder*/
2109 0, /*nb_inplace_power*/
2110 0, /*nb_inplace_lshift*/
2111 0, /*nb_inplace_rshift*/
2112 (binaryfunc)set_iand, /*nb_inplace_and*/
2113 (binaryfunc)set_ixor, /*nb_inplace_xor*/
2114 (binaryfunc)set_ior, /*nb_inplace_or*/
Raymond Hettingera690a992003-11-16 16:17:49 +00002115};
2116
2117PyDoc_STRVAR(set_doc,
Ezio Melotti7f807b72010-03-01 04:08:34 +00002118"set() -> new empty set object\n\
2119set(iterable) -> new set object\n\
Raymond Hettingera690a992003-11-16 16:17:49 +00002120\n\
Thomas Wouters0e3f5912006-08-11 14:57:12 +00002121Build an unordered collection of unique elements.");
Raymond Hettingera690a992003-11-16 16:17:49 +00002122
2123PyTypeObject PySet_Type = {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002124 PyVarObject_HEAD_INIT(&PyType_Type, 0)
2125 "set", /* tp_name */
2126 sizeof(PySetObject), /* tp_basicsize */
2127 0, /* tp_itemsize */
2128 /* methods */
2129 (destructor)set_dealloc, /* tp_dealloc */
Jeroen Demeyer530f5062019-05-31 04:13:39 +02002130 0, /* tp_vectorcall_offset */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002131 0, /* tp_getattr */
2132 0, /* tp_setattr */
Jeroen Demeyer530f5062019-05-31 04:13:39 +02002133 0, /* tp_as_async */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002134 (reprfunc)set_repr, /* tp_repr */
2135 &set_as_number, /* tp_as_number */
2136 &set_as_sequence, /* tp_as_sequence */
2137 0, /* tp_as_mapping */
Georg Brandl00da4e02010-10-18 07:32:48 +00002138 PyObject_HashNotImplemented, /* tp_hash */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002139 0, /* tp_call */
2140 0, /* tp_str */
2141 PyObject_GenericGetAttr, /* tp_getattro */
2142 0, /* tp_setattro */
2143 0, /* tp_as_buffer */
2144 Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC |
2145 Py_TPFLAGS_BASETYPE, /* tp_flags */
2146 set_doc, /* tp_doc */
2147 (traverseproc)set_traverse, /* tp_traverse */
2148 (inquiry)set_clear_internal, /* tp_clear */
2149 (richcmpfunc)set_richcompare, /* tp_richcompare */
Georg Brandl00da4e02010-10-18 07:32:48 +00002150 offsetof(PySetObject, weakreflist), /* tp_weaklistoffset */
2151 (getiterfunc)set_iter, /* tp_iter */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002152 0, /* tp_iternext */
2153 set_methods, /* tp_methods */
2154 0, /* tp_members */
2155 0, /* tp_getset */
2156 0, /* tp_base */
2157 0, /* tp_dict */
2158 0, /* tp_descr_get */
2159 0, /* tp_descr_set */
2160 0, /* tp_dictoffset */
2161 (initproc)set_init, /* tp_init */
2162 PyType_GenericAlloc, /* tp_alloc */
2163 set_new, /* tp_new */
2164 PyObject_GC_Del, /* tp_free */
Raymond Hettingera690a992003-11-16 16:17:49 +00002165};
2166
2167/* frozenset object ********************************************************/
2168
2169
2170static PyMethodDef frozenset_methods[] = {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002171 {"__contains__",(PyCFunction)set_direct_contains, METH_O | METH_COEXIST,
2172 contains_doc},
2173 {"copy", (PyCFunction)frozenset_copy, METH_NOARGS,
2174 copy_doc},
2175 {"difference", (PyCFunction)set_difference_multi, METH_VARARGS,
2176 difference_doc},
Raymond Hettinger5d2385f2015-07-08 11:52:27 -07002177 {"intersection", (PyCFunction)set_intersection_multi, METH_VARARGS,
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002178 intersection_doc},
2179 {"isdisjoint", (PyCFunction)set_isdisjoint, METH_O,
2180 isdisjoint_doc},
2181 {"issubset", (PyCFunction)set_issubset, METH_O,
2182 issubset_doc},
2183 {"issuperset", (PyCFunction)set_issuperset, METH_O,
2184 issuperset_doc},
2185 {"__reduce__", (PyCFunction)set_reduce, METH_NOARGS,
2186 reduce_doc},
2187 {"__sizeof__", (PyCFunction)set_sizeof, METH_NOARGS,
2188 sizeof_doc},
2189 {"symmetric_difference",(PyCFunction)set_symmetric_difference, METH_O,
2190 symmetric_difference_doc},
2191 {"union", (PyCFunction)set_union, METH_VARARGS,
2192 union_doc},
2193 {NULL, NULL} /* sentinel */
Raymond Hettingera690a992003-11-16 16:17:49 +00002194};
2195
2196static PyNumberMethods frozenset_as_number = {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002197 0, /*nb_add*/
2198 (binaryfunc)set_sub, /*nb_subtract*/
2199 0, /*nb_multiply*/
2200 0, /*nb_remainder*/
2201 0, /*nb_divmod*/
2202 0, /*nb_power*/
2203 0, /*nb_negative*/
2204 0, /*nb_positive*/
2205 0, /*nb_absolute*/
2206 0, /*nb_bool*/
2207 0, /*nb_invert*/
2208 0, /*nb_lshift*/
2209 0, /*nb_rshift*/
2210 (binaryfunc)set_and, /*nb_and*/
2211 (binaryfunc)set_xor, /*nb_xor*/
2212 (binaryfunc)set_or, /*nb_or*/
Raymond Hettingera690a992003-11-16 16:17:49 +00002213};
2214
2215PyDoc_STRVAR(frozenset_doc,
Ezio Melotti7f807b72010-03-01 04:08:34 +00002216"frozenset() -> empty frozenset object\n\
2217frozenset(iterable) -> frozenset object\n\
Raymond Hettingera690a992003-11-16 16:17:49 +00002218\n\
Thomas Wouters0e3f5912006-08-11 14:57:12 +00002219Build an immutable unordered collection of unique elements.");
Raymond Hettingera690a992003-11-16 16:17:49 +00002220
2221PyTypeObject PyFrozenSet_Type = {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002222 PyVarObject_HEAD_INIT(&PyType_Type, 0)
2223 "frozenset", /* tp_name */
2224 sizeof(PySetObject), /* tp_basicsize */
2225 0, /* tp_itemsize */
2226 /* methods */
2227 (destructor)set_dealloc, /* tp_dealloc */
Jeroen Demeyer530f5062019-05-31 04:13:39 +02002228 0, /* tp_vectorcall_offset */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002229 0, /* tp_getattr */
2230 0, /* tp_setattr */
Jeroen Demeyer530f5062019-05-31 04:13:39 +02002231 0, /* tp_as_async */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002232 (reprfunc)set_repr, /* tp_repr */
2233 &frozenset_as_number, /* tp_as_number */
2234 &set_as_sequence, /* tp_as_sequence */
2235 0, /* tp_as_mapping */
2236 frozenset_hash, /* tp_hash */
2237 0, /* tp_call */
2238 0, /* tp_str */
2239 PyObject_GenericGetAttr, /* tp_getattro */
2240 0, /* tp_setattro */
2241 0, /* tp_as_buffer */
2242 Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC |
2243 Py_TPFLAGS_BASETYPE, /* tp_flags */
2244 frozenset_doc, /* tp_doc */
2245 (traverseproc)set_traverse, /* tp_traverse */
2246 (inquiry)set_clear_internal, /* tp_clear */
2247 (richcmpfunc)set_richcompare, /* tp_richcompare */
Raymond Hettinger5d2385f2015-07-08 11:52:27 -07002248 offsetof(PySetObject, weakreflist), /* tp_weaklistoffset */
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002249 (getiterfunc)set_iter, /* tp_iter */
2250 0, /* tp_iternext */
2251 frozenset_methods, /* tp_methods */
2252 0, /* tp_members */
2253 0, /* tp_getset */
2254 0, /* tp_base */
2255 0, /* tp_dict */
2256 0, /* tp_descr_get */
2257 0, /* tp_descr_set */
2258 0, /* tp_dictoffset */
2259 0, /* tp_init */
2260 PyType_GenericAlloc, /* tp_alloc */
2261 frozenset_new, /* tp_new */
2262 PyObject_GC_Del, /* tp_free */
Raymond Hettingera690a992003-11-16 16:17:49 +00002263};
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002264
2265
2266/***** C API functions *************************************************/
2267
2268PyObject *
2269PySet_New(PyObject *iterable)
2270{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002271 return make_new_set(&PySet_Type, iterable);
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002272}
2273
2274PyObject *
2275PyFrozenSet_New(PyObject *iterable)
2276{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002277 return make_new_set(&PyFrozenSet_Type, iterable);
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002278}
2279
Neal Norwitz8c49c822006-03-04 18:41:19 +00002280Py_ssize_t
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002281PySet_Size(PyObject *anyset)
2282{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002283 if (!PyAnySet_Check(anyset)) {
2284 PyErr_BadInternalCall();
2285 return -1;
2286 }
2287 return PySet_GET_SIZE(anyset);
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002288}
2289
2290int
Thomas Wouters49fd7fa2006-04-21 10:40:58 +00002291PySet_Clear(PyObject *set)
2292{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002293 if (!PySet_Check(set)) {
2294 PyErr_BadInternalCall();
2295 return -1;
2296 }
2297 return set_clear_internal((PySetObject *)set);
Thomas Wouters49fd7fa2006-04-21 10:40:58 +00002298}
2299
2300int
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002301PySet_Contains(PyObject *anyset, PyObject *key)
2302{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002303 if (!PyAnySet_Check(anyset)) {
2304 PyErr_BadInternalCall();
2305 return -1;
2306 }
2307 return set_contains_key((PySetObject *)anyset, key);
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002308}
2309
2310int
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002311PySet_Discard(PyObject *set, PyObject *key)
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002312{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002313 if (!PySet_Check(set)) {
2314 PyErr_BadInternalCall();
2315 return -1;
2316 }
2317 return set_discard_key((PySetObject *)set, key);
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002318}
2319
2320int
Christian Heimesfd66e512008-01-29 12:18:50 +00002321PySet_Add(PyObject *anyset, PyObject *key)
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002322{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002323 if (!PySet_Check(anyset) &&
2324 (!PyFrozenSet_Check(anyset) || Py_REFCNT(anyset) != 1)) {
2325 PyErr_BadInternalCall();
2326 return -1;
2327 }
2328 return set_add_key((PySetObject *)anyset, key);
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002329}
2330
Thomas Wouters49fd7fa2006-04-21 10:40:58 +00002331int
Raymond Hettinger3625af52016-03-27 01:15:07 -07002332PySet_ClearFreeList(void)
2333{
2334 return 0;
2335}
2336
2337void
Victor Stinnerbed48172019-08-27 00:12:32 +02002338_PySet_Fini(void)
Raymond Hettinger3625af52016-03-27 01:15:07 -07002339{
2340 Py_CLEAR(emptyfrozenset);
2341}
2342
2343int
Benjamin Peterson8f67d082010-10-17 20:54:53 +00002344_PySet_NextEntry(PyObject *set, Py_ssize_t *pos, PyObject **key, Py_hash_t *hash)
Guido van Rossumd8faa362007-04-27 19:54:29 +00002345{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002346 setentry *entry;
Guido van Rossumd8faa362007-04-27 19:54:29 +00002347
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002348 if (!PyAnySet_Check(set)) {
2349 PyErr_BadInternalCall();
2350 return -1;
2351 }
2352 if (set_next((PySetObject *)set, pos, &entry) == 0)
2353 return 0;
2354 *key = entry->key;
Benjamin Peterson8f67d082010-10-17 20:54:53 +00002355 *hash = entry->hash;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002356 return 1;
Thomas Wouters49fd7fa2006-04-21 10:40:58 +00002357}
2358
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002359PyObject *
2360PySet_Pop(PyObject *set)
2361{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002362 if (!PySet_Check(set)) {
2363 PyErr_BadInternalCall();
2364 return NULL;
2365 }
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05302366 return set_pop((PySetObject *)set, NULL);
Raymond Hettingerbeb31012005-08-16 03:47:52 +00002367}
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002368
Thomas Wouters49fd7fa2006-04-21 10:40:58 +00002369int
2370_PySet_Update(PyObject *set, PyObject *iterable)
2371{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002372 if (!PySet_Check(set)) {
2373 PyErr_BadInternalCall();
2374 return -1;
2375 }
2376 return set_update_internal((PySetObject *)set, iterable);
Thomas Wouters49fd7fa2006-04-21 10:40:58 +00002377}
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002378
Raymond Hettingere259f132013-12-15 11:56:14 -08002379/* Exported for the gdb plugin's benefit. */
2380PyObject *_PySet_Dummy = dummy;
2381
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002382#ifdef Py_DEBUG
2383
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002384/* Test code to be called with any three element set.
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002385 Returns True and original set is restored. */
2386
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002387#define assertRaises(call_return_value, exception) \
2388 do { \
2389 assert(call_return_value); \
2390 assert(PyErr_ExceptionMatches(exception)); \
2391 PyErr_Clear(); \
2392 } while(0)
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002393
2394static PyObject *
Siddhesh Poyarekar55edd0c2018-04-30 00:29:33 +05302395test_c_api(PySetObject *so, PyObject *Py_UNUSED(ignored))
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002396{
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002397 Py_ssize_t count;
Serhiy Storchaka85b0f5b2016-11-20 10:16:47 +02002398 const char *s;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002399 Py_ssize_t i;
Raymond Hettinger583cd032013-09-07 22:06:35 -07002400 PyObject *elem=NULL, *dup=NULL, *t, *f, *dup2, *x=NULL;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002401 PyObject *ob = (PyObject *)so;
Benjamin Peterson8f67d082010-10-17 20:54:53 +00002402 Py_hash_t hash;
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002403 PyObject *str;
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002404
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002405 /* Verify preconditions */
2406 assert(PyAnySet_Check(ob));
2407 assert(PyAnySet_CheckExact(ob));
2408 assert(!PyFrozenSet_CheckExact(ob));
Victor Stinner08b36bd2010-03-13 00:19:17 +00002409
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002410 /* so.clear(); so |= set("abc"); */
2411 str = PyUnicode_FromString("abc");
2412 if (str == NULL)
2413 return NULL;
2414 set_clear_internal(so);
Raymond Hettinger5af9e132015-05-13 01:44:36 -07002415 if (set_update_internal(so, str)) {
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002416 Py_DECREF(str);
2417 return NULL;
2418 }
2419 Py_DECREF(str);
Victor Stinner08b36bd2010-03-13 00:19:17 +00002420
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002421 /* Exercise type/size checks */
2422 assert(PySet_Size(ob) == 3);
2423 assert(PySet_GET_SIZE(ob) == 3);
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002424
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002425 /* Raise TypeError for non-iterable constructor arguments */
2426 assertRaises(PySet_New(Py_None) == NULL, PyExc_TypeError);
2427 assertRaises(PyFrozenSet_New(Py_None) == NULL, PyExc_TypeError);
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002428
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002429 /* Raise TypeError for unhashable key */
2430 dup = PySet_New(ob);
2431 assertRaises(PySet_Discard(ob, dup) == -1, PyExc_TypeError);
2432 assertRaises(PySet_Contains(ob, dup) == -1, PyExc_TypeError);
2433 assertRaises(PySet_Add(ob, dup) == -1, PyExc_TypeError);
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002434
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002435 /* Exercise successful pop, contains, add, and discard */
2436 elem = PySet_Pop(ob);
2437 assert(PySet_Contains(ob, elem) == 0);
2438 assert(PySet_GET_SIZE(ob) == 2);
2439 assert(PySet_Add(ob, elem) == 0);
2440 assert(PySet_Contains(ob, elem) == 1);
2441 assert(PySet_GET_SIZE(ob) == 3);
2442 assert(PySet_Discard(ob, elem) == 1);
2443 assert(PySet_GET_SIZE(ob) == 2);
2444 assert(PySet_Discard(ob, elem) == 0);
2445 assert(PySet_GET_SIZE(ob) == 2);
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002446
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002447 /* Exercise clear */
2448 dup2 = PySet_New(dup);
2449 assert(PySet_Clear(dup2) == 0);
2450 assert(PySet_Size(dup2) == 0);
2451 Py_DECREF(dup2);
Thomas Wouters49fd7fa2006-04-21 10:40:58 +00002452
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002453 /* Raise SystemError on clear or update of frozen set */
2454 f = PyFrozenSet_New(dup);
2455 assertRaises(PySet_Clear(f) == -1, PyExc_SystemError);
2456 assertRaises(_PySet_Update(f, dup) == -1, PyExc_SystemError);
2457 assert(PySet_Add(f, elem) == 0);
2458 Py_INCREF(f);
2459 assertRaises(PySet_Add(f, elem) == -1, PyExc_SystemError);
2460 Py_DECREF(f);
2461 Py_DECREF(f);
Thomas Wouters49fd7fa2006-04-21 10:40:58 +00002462
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002463 /* Exercise direct iteration */
2464 i = 0, count = 0;
2465 while (_PySet_NextEntry((PyObject *)dup, &i, &x, &hash)) {
Serhiy Storchaka06515832016-11-20 09:13:07 +02002466 s = PyUnicode_AsUTF8(x);
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002467 assert(s && (s[0] == 'a' || s[0] == 'b' || s[0] == 'c'));
2468 count++;
2469 }
2470 assert(count == 3);
Thomas Wouters49fd7fa2006-04-21 10:40:58 +00002471
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002472 /* Exercise updates */
2473 dup2 = PySet_New(NULL);
2474 assert(_PySet_Update(dup2, dup) == 0);
2475 assert(PySet_Size(dup2) == 3);
2476 assert(_PySet_Update(dup2, dup) == 0);
2477 assert(PySet_Size(dup2) == 3);
2478 Py_DECREF(dup2);
Thomas Wouters49fd7fa2006-04-21 10:40:58 +00002479
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002480 /* Raise SystemError when self argument is not a set or frozenset. */
2481 t = PyTuple_New(0);
2482 assertRaises(PySet_Size(t) == -1, PyExc_SystemError);
2483 assertRaises(PySet_Contains(t, elem) == -1, PyExc_SystemError);
2484 Py_DECREF(t);
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002485
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002486 /* Raise SystemError when self argument is not a set. */
2487 f = PyFrozenSet_New(dup);
2488 assert(PySet_Size(f) == 3);
2489 assert(PyFrozenSet_CheckExact(f));
2490 assertRaises(PySet_Discard(f, elem) == -1, PyExc_SystemError);
2491 assertRaises(PySet_Pop(f) == NULL, PyExc_SystemError);
2492 Py_DECREF(f);
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002493
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002494 /* Raise KeyError when popping from an empty set */
2495 assert(PyNumber_InPlaceSubtract(ob, ob) == ob);
2496 Py_DECREF(ob);
2497 assert(PySet_GET_SIZE(ob) == 0);
2498 assertRaises(PySet_Pop(ob) == NULL, PyExc_KeyError);
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002499
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002500 /* Restore the set from the copy using the PyNumber API */
2501 assert(PyNumber_InPlaceOr(ob, dup) == ob);
2502 Py_DECREF(ob);
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002503
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002504 /* Verify constructors accept NULL arguments */
2505 f = PySet_New(NULL);
2506 assert(f != NULL);
2507 assert(PySet_GET_SIZE(f) == 0);
2508 Py_DECREF(f);
2509 f = PyFrozenSet_New(NULL);
2510 assert(f != NULL);
2511 assert(PyFrozenSet_CheckExact(f));
2512 assert(PySet_GET_SIZE(f) == 0);
2513 Py_DECREF(f);
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002514
Antoine Pitrouf95a1b32010-05-09 15:52:27 +00002515 Py_DECREF(elem);
2516 Py_DECREF(dup);
2517 Py_RETURN_TRUE;
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002518}
2519
Raymond Hettinger9bda1d62005-09-16 07:14:21 +00002520#undef assertRaises
2521
Raymond Hettingerc47e01d2005-08-16 10:44:15 +00002522#endif
Raymond Hettingerbfc1e1a2013-08-23 03:22:15 -05002523
2524/***** Dummy Struct *************************************************/
2525
2526static PyObject *
2527dummy_repr(PyObject *op)
2528{
2529 return PyUnicode_FromString("<dummy key>");
2530}
2531
Victor Stinner2a4903f2020-01-30 13:09:11 +01002532static void _Py_NO_RETURN
Raymond Hettingerbfc1e1a2013-08-23 03:22:15 -05002533dummy_dealloc(PyObject* ignore)
2534{
2535 Py_FatalError("deallocating <dummy key>");
2536}
2537
2538static PyTypeObject _PySetDummy_Type = {
2539 PyVarObject_HEAD_INIT(&PyType_Type, 0)
2540 "<dummy key> type",
2541 0,
2542 0,
2543 dummy_dealloc, /*tp_dealloc*/ /*never called*/
Jeroen Demeyer530f5062019-05-31 04:13:39 +02002544 0, /*tp_vectorcall_offset*/
Raymond Hettingerbfc1e1a2013-08-23 03:22:15 -05002545 0, /*tp_getattr*/
2546 0, /*tp_setattr*/
Jeroen Demeyer530f5062019-05-31 04:13:39 +02002547 0, /*tp_as_async*/
Raymond Hettingerbfc1e1a2013-08-23 03:22:15 -05002548 dummy_repr, /*tp_repr*/
2549 0, /*tp_as_number*/
2550 0, /*tp_as_sequence*/
2551 0, /*tp_as_mapping*/
2552 0, /*tp_hash */
2553 0, /*tp_call */
2554 0, /*tp_str */
2555 0, /*tp_getattro */
2556 0, /*tp_setattro */
2557 0, /*tp_as_buffer */
2558 Py_TPFLAGS_DEFAULT, /*tp_flags */
2559};
2560
2561static PyObject _dummy_struct = {
2562 _PyObject_EXTRA_INIT
2563 2, &_PySetDummy_Type
2564};
2565