Repository navigation
Expand file tree
/
Copy pathodictobject.c
More file actions
2438 lines (2077 loc) · 76.8 KB
/
Copy pathodictobject.c
File metadata and controls
2438 lines (2077 loc) · 76.8 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/* Ordered Dictionary object implementation.
This implementation is necessarily explicitly equivalent to the pure Python
OrderedDict class in Lib/collections/__init__.py. The strategy there
involves using a doubly-linked-list to capture the order. We keep to that
strategy, using a lower-level linked-list.
About the Linked-List
=====================
For the linked list we use a basic doubly-linked-list. Using a circularly-
linked-list does have some benefits, but they don't apply so much here
since OrderedDict is focused on the ends of the list (for the most part).
Furthermore, there are some features of generic linked-lists that we simply
don't need for OrderedDict. Thus a simple custom implementation meets our
needs. Alternatives to our simple approach include the QCIRCLE_*
macros from BSD's queue.h, and the linux's list.h.
Getting O(1) Node Lookup
------------------------
One invariant of Python's OrderedDict is that it preserves time complexity
of dict's methods, particularly the O(1) operations. Simply adding a
linked-list on top of dict is not sufficient here; operations for nodes in
the middle of the linked-list implicitly require finding the node first.
With a simple linked-list like we're using, that is an O(n) operation.
Consequently, methods like __delitem__() would change from O(1) to O(n),
which is unacceptable.
In order to preserve O(1) performance for node removal (finding nodes), we
must do better than just looping through the linked-list. Here are options
we've considered:
1. use a second dict to map keys to nodes (a la the pure Python version).
2. keep a simple hash table mirroring the order of dict's, mapping each key
to the corresponding node in the linked-list.
3. use a version of shared keys (split dict) that allows non-unicode keys.
4. have the value stored for each key be a (value, node) pair, and adjust
__getitem__(), get(), etc. accordingly.
The approach with the least performance impact (time and space) is #2,
mirroring the key order of dict's dk_entries with an array of node pointers.
While lookdict() and friends (dk_lookup) don't give us the index into the
array, we make use of pointer arithmetic to get that index. An alternative
would be to refactor lookdict() to provide the index, explicitly exposing
the implementation detail. We could even just use a custom lookup function
for OrderedDict that facilitates our need. However, both approaches are
significantly more complicated than just using pointer arithmetic.
The catch with mirroring the hash table ordering is that we have to keep
the ordering in sync through any dict resizes. However, that order only
matters during node lookup. We can simply defer any potential resizing
until we need to do a lookup.
Linked-List Nodes
-----------------
The current implementation stores a pointer to the associated key only.
One alternative would be to store a pointer to the PyDictKeyEntry instead.
This would save one pointer de-reference per item, which is nice during
calls to values() and items(). However, it adds unnecessary overhead
otherwise, so we stick with just the key.
Linked-List API
---------------
As noted, the linked-list implemented here does not have all the bells and
whistles. However, we recognize that the implementation may need to
change to accommodate performance improvements or extra functionality. To
that end, We use a simple API to interact with the linked-list. Here's a
summary of the methods/macros:
Node info:
* _odictnode_KEY(node)
* _odictnode_VALUE(od, node)
* _odictnode_PREV(node)
* _odictnode_NEXT(node)
Linked-List info:
* _odict_FIRST(od)
* _odict_LAST(od)
* _odict_EMPTY(od)
* _odict_FOREACH(od, node) - used in place of `for (node=...)`
For adding nodes:
* _odict_add_head(od, node)
* _odict_add_tail(od, node)
* _odict_add_new_node(od, key, hash)
For removing nodes:
* _odict_clear_node(od, node, key, hash)
* _odict_clear_nodes(od, clear_each)
Others:
* _odict_find_node_hash(od, key, hash)
* _odict_find_node(od, key)
* _odict_keys_equal(od1, od2)
Used, but specific to the linked-list implementation:
* _odict_free_fast_nodes(od)
And here's a look at how the linked-list relates to the OrderedDict API:
============ === === ==== ==== ==== === ==== ===== ==== ==== === ==== === ===
method key val prev next mem 1st last empty iter find add rmv clr keq
============ === === ==== ==== ==== === ==== ===== ==== ==== === ==== === ===
__del__ ~ X
__delitem__ free ~ node
__eq__ ~ X
__iter__ X X
__new__ X X
__reduce__ X ~ X
__repr__ X X X
__reversed__ X X
__setitem__ key
__sizeof__ size X
clear ~ ~ X
copy X X X
items X X X
keys X X
move_to_end X X X ~ h/t key
pop free key
popitem X X free X X node
setdefault ~ ? ~
values X X
============ === === ==== ==== ==== === ==== ===== ==== ==== === ==== === ===
__delitem__ is the only method that directly relies on finding an arbitrary
node in the linked-list. Everything else is iteration or relates to the
ends of the linked-list.
Situation that Endangers Consistency
------------------------------------
Using a raw linked-list for OrderedDict exposes a key situation that can
cause problems. If a node is stored in a variable, there is a chance that
the node may have been deallocated before the variable gets used, thus
potentially leading to a segmentation fault. A key place where this shows
up is during iteration through the linked list (via _odict_FOREACH or
otherwise).
A number of solutions are available to resolve this situation:
* defer looking up the node until as late as possible and certainly after
any code that could possibly result in a deletion;
* if the node is needed both before and after a point where the node might
be removed, do a check before using the node at the "after" location to
see if the node is still valid;
* like the last one, but simply pull the node again to ensure it's right;
* keep the key in the variable instead of the node and then look up the
node using the key at the point where the node is needed (this is what
we do for the iterators).
Another related problem, preserving consistent ordering during iteration,
is described below. That one is not exclusive to using linked-lists.
Challenges from Subclassing dict
================================
OrderedDict subclasses dict, which is an unusual relationship between two
builtin types (other than the base object type). Doing so results in
some complication and deserves further explanation. There are two things
to consider here. First, in what circumstances or with what adjustments
can OrderedDict be used as a drop-in replacement for dict (at the C level)?
Second, how can the OrderedDict implementation leverage the dict
implementation effectively without introducing unnecessary coupling or
inefficiencies?
This second point is reflected here and in the implementation, so the
further focus is on the first point. It is worth noting that for
overridden methods, the dict implementation is deferred to as much as
possible. Furthermore, coupling is limited to as little as is reasonable.
Concrete API Compatibility
--------------------------
Use of the concrete C-API for dict (PyDict_*) with OrderedDict is
problematic. (See http://bugs.python.org/issue10977.) The concrete API
has a number of hard-coded assumptions tied to the dict implementation.
This is, in part, due to performance reasons, which is understandable
given the part dict plays in Python.
Any attempt to replace dict with OrderedDict for any role in the
interpreter (e.g. **kwds) faces a challenge. Such any effort must
recognize that the instances in affected locations currently interact with
the concrete API.
Here are some ways to address this challenge:
1. Change the relevant usage of the concrete API in CPython and add
PyDict_CheckExact() calls to each of the concrete API functions.
2. Adjust the relevant concrete API functions to explicitly accommodate
OrderedDict.
3. As with #1, add the checks, but improve the abstract API with smart fast
paths for dict and OrderedDict, and refactor CPython to use the abstract
API. Improvements to the abstract API would be valuable regardless.
Adding the checks to the concrete API would help make any interpreter
switch to OrderedDict less painful for extension modules. However, this
won't work. The equivalent C API call to `dict.__setitem__(obj, k, v)`
is 'PyDict_SetItem(obj, k, v)`. This illustrates how subclasses in C call
the base class's methods, since there is no equivalent of super() in the
C API. Calling into Python for parent class API would work, but some
extension modules already rely on this feature of the concrete API.
For reference, here is a breakdown of some of the dict concrete API:
========================== ============= =======================
concrete API uses abstract API
========================== ============= =======================
PyDict_Check PyMapping_Check
(PyDict_CheckExact) -
(PyDict_New) -
(PyDictProxy_New) -
PyDict_Clear -
PyDict_Contains PySequence_Contains
PyDict_Copy -
PyDict_SetItem PyObject_SetItem
PyDict_SetItemString PyMapping_SetItemString
PyDict_DelItem PyMapping_DelItem
PyDict_DelItemString PyMapping_DelItemString
PyDict_GetItem -
PyDict_GetItemWithError PyObject_GetItem
_PyDict_GetItemIdWithError -
PyDict_GetItemString PyMapping_GetItemString
PyDict_Items PyMapping_Items
PyDict_Keys PyMapping_Keys
PyDict_Values PyMapping_Values
PyDict_Size PyMapping_Size
PyMapping_Length
PyDict_Next PyIter_Next
_PyDict_Next -
PyDict_Merge -
PyDict_Update -
PyDict_MergeFromSeq2 -
PyDict_ClearFreeList -
- PyMapping_HasKeyString
- PyMapping_HasKey
========================== ============= =======================
The dict Interface Relative to OrderedDict
==========================================
Since OrderedDict subclasses dict, understanding the various methods and
attributes of dict is important for implementing OrderedDict.
Relevant Type Slots
-------------------
================= ================ =================== ================
slot attribute object dict
================= ================ =================== ================
tp_dealloc - object_dealloc dict_dealloc
tp_repr __repr__ object_repr dict_repr
sq_contains __contains__ - dict_contains
mp_length __len__ - dict_length
mp_subscript __getitem__ - dict_subscript
mp_ass_subscript __setitem__ - dict_ass_sub
__delitem__
tp_hash __hash__ _Py_HashPointer ..._HashNotImpl
tp_str __str__ object_str -
tp_getattro __getattribute__ ..._GenericGetAttr (repeated)
__getattr__
tp_setattro __setattr__ ..._GenericSetAttr (disabled)
tp_doc __doc__ (literal) dictionary_doc
tp_traverse - - dict_traverse
tp_clear - - dict_tp_clear
tp_richcompare __eq__ object_richcompare dict_richcompare
__ne__
tp_weaklistoffset (__weakref__) - -
tp_iter __iter__ - dict_iter
tp_dictoffset (__dict__) - -
tp_init __init__ object_init dict_init
tp_alloc - PyType_GenericAlloc (repeated)
tp_new __new__ object_new dict_new
tp_free - PyObject_Del PyObject_GC_Del
================= ================ =================== ================
Relevant Methods
----------------
================ =================== ===============
method object dict
================ =================== ===============
__reduce__ object_reduce -
__sizeof__ object_sizeof dict_sizeof
clear - dict_clear
copy - dict_copy
fromkeys - dict_fromkeys
get - dict_get
items - dictitems_new
keys - dictkeys_new
pop - dict_pop
popitem - dict_popitem
setdefault - dict_setdefault
update - dict_update
values - dictvalues_new
================ =================== ===============
Pure Python OrderedDict
=======================
As already noted, compatibility with the pure Python OrderedDict
implementation is a key goal of this C implementation. To further that
goal, here's a summary of how OrderedDict-specific methods are implemented
in collections/__init__.py. Also provided is an indication of which
methods directly mutate or iterate the object, as well as any relationship
with the underlying linked-list.
============= ============== == ================ === === ====
method impl used ll uses inq mut iter
============= ============== == ================ === === ====
__contains__ dict - - X
__delitem__ OrderedDict Y dict.__delitem__ X
__eq__ OrderedDict N OrderedDict ~
dict.__eq__
__iter__
__getitem__ dict - - X
__iter__ OrderedDict Y - X
__init__ OrderedDict N update
__len__ dict - - X
__ne__ MutableMapping - __eq__ ~
__reduce__ OrderedDict N OrderedDict ~
__iter__
__getitem__
__repr__ OrderedDict N __class__ ~
items
__reversed__ OrderedDict Y - X
__setitem__ OrderedDict Y __contains__ X
dict.__setitem__
__sizeof__ OrderedDict Y __len__ ~
__dict__
clear OrderedDict Y dict.clear X
copy OrderedDict N __class__
__init__
fromkeys OrderedDict N __setitem__
get dict - - ~
items MutableMapping - ItemsView X
keys MutableMapping - KeysView X
move_to_end OrderedDict Y - X
pop OrderedDict N __contains__ X
__getitem__
__delitem__
popitem OrderedDict Y dict.pop X
setdefault OrderedDict N __contains__ ~
__getitem__
__setitem__
update MutableMapping - __setitem__ ~
values MutableMapping - ValuesView X
============= ============== == ================ === === ====
__reversed__ and move_to_end are both exclusive to OrderedDict.
C OrderedDict Implementation
============================
================= ================
slot impl
================= ================
tp_dealloc odict_dealloc
tp_repr odict_repr
mp_ass_subscript odict_ass_sub
tp_doc odict_doc
tp_traverse odict_traverse
tp_clear odict_tp_clear
tp_richcompare odict_richcompare
tp_weaklistoffset (offset)
tp_iter odict_iter
tp_dictoffset (offset)
tp_init odict_init
tp_alloc (repeated)
tp_new odict_new
================= ================
================= ================
method impl
================= ================
__reduce__ odict_reduce
__sizeof__ odict_sizeof
clear odict_clear
copy odict_copy
fromkeys odict_fromkeys
items odictitems_new
keys odictkeys_new
pop odict_pop
popitem odict_popitem
setdefault odict_setdefault
update odict_update
values odictvalues_new
================= ================
Inherited unchanged from object/dict:
================ ==========================
method type field
================ ==========================
- tp_free
__contains__ tp_as_sequence.sq_contains
__getattr__ tp_getattro
__getattribute__ tp_getattro
__getitem__ tp_as_mapping.mp_subscript
__hash__ tp_hash
__len__ tp_as_mapping.mp_length
__setattr__ tp_setattro
__str__ tp_str
get -
================ ==========================
Other Challenges
================
Preserving Ordering During Iteration
------------------------------------
During iteration through an OrderedDict, it is possible that items could
get added, removed, or reordered. For a linked-list implementation, as
with some other implementations, that situation may lead to undefined
behavior. The documentation for dict mentions this in the `iter()` section
of http://docs.python.org/3.4/library/stdtypes.html#dictionary-view-objects.
In this implementation we follow dict's lead (as does the pure Python
implementation) for __iter__(), keys(), values(), and items().
For internal iteration (using _odict_FOREACH or not), there is still the
risk that not all nodes that we expect to be seen in the loop actually get
seen. Thus, we are careful in each of those places to ensure that they
are. This comes, of course, at a small price at each location. The
solutions are much the same as those detailed in the `Situation that
Endangers Consistency` section above.
Potential Optimizations
=======================
* Allocate the nodes as a block via od_fast_nodes instead of individually.
- Set node->key to NULL to indicate the node is not-in-use.
- Add _odict_EXISTS()?
- How to maintain consistency across resizes? Existing node pointers
would be invalidate after a resize, which is particularly problematic
for the iterators.
* Use a more stream-lined implementation of update() and, likely indirectly,
__init__().
*/
/* TODO
sooner:
- reentrancy (make sure everything is at a thread-safe state when calling
into Python). I've already checked this multiple times, but want to
make one more pass.
- add unit tests for reentrancy?
later:
- make the dict views support the full set API (the pure Python impl does)
- implement a fuller MutableMapping API in C?
- move the MutableMapping implementation to abstract.c?
- optimize mutablemapping_update
- use PyObject_MALLOC (small object allocator) for odict nodes?
- support subclasses better (e.g. in odict_richcompare)
*/
#include "Python.h"
#include "structmember.h"
#include "dict-common.h"
#include <stddef.h>
typedef struct _odictnode _ODictNode;
/* PyODictObject */
struct _odictobject {
PyDictObject od_dict; /* the underlying dict */
_ODictNode *od_first; /* first node in the linked list, if any */
_ODictNode *od_last; /* last node in the linked list, if any */
/* od_fast_nodes, od_fast_nodes_size and od_resize_sentinel are managed
* by _odict_resize().
* Note that we rely on implementation details of dict for both. */
_ODictNode **od_fast_nodes; /* hash table that mirrors the dict table */
Py_ssize_t od_fast_nodes_size;
void *od_resize_sentinel; /* changes if odict should be resized */
size_t od_state; /* incremented whenever the LL changes */
PyObject *od_inst_dict; /* OrderedDict().__dict__ */
PyObject *od_weakreflist; /* holds weakrefs to the odict */
};
/* ----------------------------------------------
* odict keys (a simple doubly-linked list)
*/
struct _odictnode {
PyObject *key;
Py_hash_t hash;
_ODictNode *next;
_ODictNode *prev;
};
#define _odictnode_KEY(node) \
(node->key)
#define _odictnode_HASH(node) \
(node->hash)
/* borrowed reference */
#define _odictnode_VALUE(node, od) \
PyODict_GetItemWithError((PyObject *)od, _odictnode_KEY(node))
#define _odictnode_PREV(node) (node->prev)
#define _odictnode_NEXT(node) (node->next)
#define _odict_FIRST(od) (((PyODictObject *)od)->od_first)
#define _odict_LAST(od) (((PyODictObject *)od)->od_last)
#define _odict_EMPTY(od) (_odict_FIRST(od) == NULL)
#define _odict_FOREACH(od, node) \
for (node = _odict_FIRST(od); node != NULL; node = _odictnode_NEXT(node))
#define _odict_FAST_SIZE(od) ((PyDictObject *)od)->ma_keys->dk_size
static void
_odict_free_fast_nodes(PyODictObject *od) {
if (od->od_fast_nodes) {
PyMem_FREE(od->od_fast_nodes);
}
}
/* Return the index into the hash table, regardless of a valid node. */
static Py_ssize_t
_odict_get_index_raw(PyODictObject *od, PyObject *key, Py_hash_t hash)
{
PyObject **value_addr = NULL;
PyDictKeysObject *keys = ((PyDictObject *)od)->ma_keys;
Py_ssize_t ix;
ix = (keys->dk_lookup)((PyDictObject *)od, key, hash, &value_addr, NULL);
if (ix == DKIX_EMPTY) {
return keys->dk_nentries; /* index of new entry */
}
if (ix < 0)
return -1;
/* We use pointer arithmetic to get the entry's index into the table. */
return ix;
}
/* Replace od->od_fast_nodes with a new table matching the size of dict's. */
static int
_odict_resize(PyODictObject *od) {
Py_ssize_t size, i;
_ODictNode **fast_nodes, *node;
/* Initialize a new "fast nodes" table. */
size = ((PyDictObject *)od)->ma_keys->dk_size;
fast_nodes = PyMem_NEW(_ODictNode *, size);
if (fast_nodes == NULL) {
PyErr_NoMemory();
return -1;
}
for (i = 0; i < size; i++)
fast_nodes[i] = NULL;
/* Copy the current nodes into the table. */
_odict_FOREACH(od, node) {
i = _odict_get_index_raw(od, _odictnode_KEY(node),
_odictnode_HASH(node));
if (i < 0) {
PyMem_FREE(fast_nodes);
return -1;
}
fast_nodes[i] = node;
}
/* Replace the old fast nodes table. */
_odict_free_fast_nodes(od);
od->od_fast_nodes = fast_nodes;
od->od_fast_nodes_size = size;
od->od_resize_sentinel = ((PyDictObject *)od)->ma_keys;
return 0;
}
/* Return the index into the hash table, regardless of a valid node. */
static Py_ssize_t
_odict_get_index(PyODictObject *od, PyObject *key, Py_hash_t hash)
{
PyDictKeysObject *keys;
assert(key != NULL);
keys = ((PyDictObject *)od)->ma_keys;
/* Ensure od_fast_nodes and dk_entries are in sync. */
if (od->od_resize_sentinel != keys ||
od->od_fast_nodes_size != keys->dk_size) {
int resize_res = _odict_resize(od);
if (resize_res < 0)
return -1;
}
return _odict_get_index_raw(od, key, hash);
}
/* Returns NULL if there was some error or the key was not found. */
static _ODictNode *
_odict_find_node_hash(PyODictObject *od, PyObject *key, Py_hash_t hash)
{
Py_ssize_t index;
if (_odict_EMPTY(od))
return NULL;
index = _odict_get_index(od, key, hash);
if (index < 0)
return NULL;
return od->od_fast_nodes[index];
}
static _ODictNode *
_odict_find_node(PyODictObject *od, PyObject *key)
{
Py_ssize_t index;
Py_hash_t hash;
if (_odict_EMPTY(od))
return NULL;
hash = PyObject_Hash(key);
if (hash == -1)
return NULL;
index = _odict_get_index(od, key, hash);
if (index < 0)
return NULL;
return od->od_fast_nodes[index];
}
static void
_odict_add_head(PyODictObject *od, _ODictNode *node)
{
_odictnode_PREV(node) = NULL;
_odictnode_NEXT(node) = _odict_FIRST(od);
if (_odict_FIRST(od) == NULL)
_odict_LAST(od) = node;
else
_odictnode_PREV(_odict_FIRST(od)) = node;
_odict_FIRST(od) = node;
od->od_state++;
}
static void
_odict_add_tail(PyODictObject *od, _ODictNode *node)
{
_odictnode_PREV(node) = _odict_LAST(od);
_odictnode_NEXT(node) = NULL;
if (_odict_LAST(od) == NULL)
_odict_FIRST(od) = node;
else
_odictnode_NEXT(_odict_LAST(od)) = node;
_odict_LAST(od) = node;
od->od_state++;
}
/* adds the node to the end of the list */
static int
_odict_add_new_node(PyODictObject *od, PyObject *key, Py_hash_t hash)
{
Py_ssize_t i;
_ODictNode *node;
Py_INCREF(key);
i = _odict_get_index(od, key, hash);
if (i < 0) {
if (!PyErr_Occurred())
PyErr_SetObject(PyExc_KeyError, key);
Py_DECREF(key);
return -1;
}
else if (od->od_fast_nodes[i] != NULL) {
/* We already have a node for the key so there's no need to add one. */
Py_DECREF(key);
return 0;
}
/* must not be added yet */
node = (_ODictNode *)PyMem_MALLOC(sizeof(_ODictNode));
if (node == NULL) {
Py_DECREF(key);
PyErr_NoMemory();
return -1;
}
_odictnode_KEY(node) = key;
_odictnode_HASH(node) = hash;
_odict_add_tail(od, node);
od->od_fast_nodes[i] = node;
return 0;
}
/* Putting the decref after the free causes problems. */
#define _odictnode_DEALLOC(node) \
do { \
Py_DECREF(_odictnode_KEY(node)); \
PyMem_FREE((void *)node); \
} while (0)
/* Repeated calls on the same node are no-ops. */
static void
_odict_remove_node(PyODictObject *od, _ODictNode *node)
{
if (_odict_FIRST(od) == node)
_odict_FIRST(od) = _odictnode_NEXT(node);
else if (_odictnode_PREV(node) != NULL)
_odictnode_NEXT(_odictnode_PREV(node)) = _odictnode_NEXT(node);
if (_odict_LAST(od) == node)
_odict_LAST(od) = _odictnode_PREV(node);
else if (_odictnode_NEXT(node) != NULL)
_odictnode_PREV(_odictnode_NEXT(node)) = _odictnode_PREV(node);
_odictnode_PREV(node) = NULL;
_odictnode_NEXT(node) = NULL;
od->od_state++;
}
/* If someone calls PyDict_DelItem() directly on an OrderedDict, we'll
get all sorts of problems here. In PyODict_DelItem we make sure to
call _odict_clear_node first.
This matters in the case of colliding keys. Suppose we add 3 keys:
[A, B, C], where the hash of C collides with A and the next possible
index in the hash table is occupied by B. If we remove B then for C
the dict's looknode func will give us the old index of B instead of
the index we got before deleting B. However, the node for C in
od_fast_nodes is still at the old dict index of C. Thus to be sure
things don't get out of sync, we clear the node in od_fast_nodes
*before* calling PyDict_DelItem.
The same must be done for any other OrderedDict operations where
we modify od_fast_nodes.
*/
static int
_odict_clear_node(PyODictObject *od, _ODictNode *node, PyObject *key,
Py_hash_t hash)
{
Py_ssize_t i;
assert(key != NULL);
if (_odict_EMPTY(od)) {
/* Let later code decide if this is a KeyError. */
return 0;
}
i = _odict_get_index(od, key, hash);
if (i < 0)
return PyErr_Occurred() ? -1 : 0;
if (node == NULL)
node = od->od_fast_nodes[i];
assert(node == od->od_fast_nodes[i]);
if (node == NULL) {
/* Let later code decide if this is a KeyError. */
return 0;
}
// Now clear the node.
od->od_fast_nodes[i] = NULL;
_odict_remove_node(od, node);
_odictnode_DEALLOC(node);
return 0;
}
static void
_odict_clear_nodes(PyODictObject *od)
{
_ODictNode *node, *next;
_odict_free_fast_nodes(od);
od->od_fast_nodes = NULL;
node = _odict_FIRST(od);
_odict_FIRST(od) = NULL;
_odict_LAST(od) = NULL;
while (node != NULL) {
next = _odictnode_NEXT(node);
_odictnode_DEALLOC(node);
node = next;
}
}
/* There isn't any memory management of nodes past this point. */
#undef _odictnode_DEALLOC
static int
_odict_keys_equal(PyODictObject *a, PyODictObject *b)
{
_ODictNode *node_a, *node_b;
node_a = _odict_FIRST(a);
node_b = _odict_FIRST(b);
while (1) {
if (node_a == NULL && node_b == NULL)
/* success: hit the end of each at the same time */
return 1;
else if (node_a == NULL || node_b == NULL)
/* unequal length */
return 0;
else {
int res = PyObject_RichCompareBool(
(PyObject *)_odictnode_KEY(node_a),
(PyObject *)_odictnode_KEY(node_b),
Py_EQ);
if (res < 0)
return res;
else if (res == 0)
return 0;
/* otherwise it must match, so move on to the next one */
node_a = _odictnode_NEXT(node_a);
node_b = _odictnode_NEXT(node_b);
}
}
}
/* ----------------------------------------------
* OrderedDict mapping methods
*/
/* mp_ass_subscript: __setitem__() and __delitem__() */
static int
odict_mp_ass_sub(PyODictObject *od, PyObject *v, PyObject *w)
{
if (w == NULL)
return PyODict_DelItem((PyObject *)od, v);
else
return PyODict_SetItem((PyObject *)od, v, w);
}
/* tp_as_mapping */
static PyMappingMethods odict_as_mapping = {
0, /*mp_length*/
0, /*mp_subscript*/
(objobjargproc)odict_mp_ass_sub, /*mp_ass_subscript*/
};
/* ----------------------------------------------
* OrderedDict methods
*/
/* __delitem__() */
PyDoc_STRVAR(odict_delitem__doc__, "od.__delitem__(y) <==> del od[y]");
/* __eq__() */
PyDoc_STRVAR(odict_eq__doc__,
"od.__eq__(y) <==> od==y. Comparison to another OD is order-sensitive \n\
while comparison to a regular mapping is order-insensitive.\n\
");
/* forward */
static PyObject * odict_richcompare(PyObject *v, PyObject *w, int op);
static PyObject *
odict_eq(PyObject *a, PyObject *b)
{
return odict_richcompare(a, b, Py_EQ);
}
/* __init__() */
PyDoc_STRVAR(odict_init__doc__,
"Initialize an ordered dictionary. The signature is the same as\n\
regular dictionaries, but keyword arguments are not recommended because\n\
their insertion order is arbitrary.\n\
\n\
");
/* forward */
static int odict_init(PyObject *self, PyObject *args, PyObject *kwds);
/* __iter__() */
PyDoc_STRVAR(odict_iter__doc__, "od.__iter__() <==> iter(od)");
static PyObject * odict_iter(PyODictObject *self); /* forward */
/* __ne__() */
/* Mapping.__ne__() does not have a docstring. */
PyDoc_STRVAR(odict_ne__doc__, "");
static PyObject *
odict_ne(PyObject *a, PyObject *b)
{
return odict_richcompare(a, b, Py_NE);
}
/* __repr__() */
PyDoc_STRVAR(odict_repr__doc__, "od.__repr__() <==> repr(od)");
static PyObject * odict_repr(PyODictObject *self); /* forward */
/* __setitem__() */
PyDoc_STRVAR(odict_setitem__doc__, "od.__setitem__(i, y) <==> od[i]=y");
/* fromkeys() */
PyDoc_STRVAR(odict_fromkeys__doc__,
"OD.fromkeys(S[, v]) -> New ordered dictionary with keys from S.\n\
If not specified, the value defaults to None.\n\
\n\
");
static PyObject *
odict_fromkeys(PyObject *cls, PyObject *args, PyObject *kwargs)
{
static char *kwlist[] = {"iterable", "value", 0};
PyObject *seq;
PyObject *value = Py_None;
/* both borrowed */
if (!PyArg_ParseTupleAndKeywords(args, kwargs, "O|O:fromkeys", kwlist,
&seq, &value)) {
return NULL;
}
return _PyDict_FromKeys(cls, seq, value);
}
/* __sizeof__() */
/* OrderedDict.__sizeof__() does not have a docstring. */
PyDoc_STRVAR(odict_sizeof__doc__, "");
static PyObject *
odict_sizeof(PyODictObject *od)
{
Py_ssize_t res = _PyDict_SizeOf((PyDictObject *)od);
res += sizeof(_ODictNode *) * _odict_FAST_SIZE(od); /* od_fast_nodes */
if (!_odict_EMPTY(od)) {
res += sizeof(_ODictNode) * PyODict_SIZE(od); /* linked-list */
}
return PyLong_FromSsize_t(res);
}
/* __reduce__() */
PyDoc_STRVAR(odict_reduce__doc__, "Return state information for pickling");
static PyObject *
odict_reduce(register PyODictObject *od)
{
_Py_IDENTIFIER(__dict__);
_Py_IDENTIFIER(items);
PyObject *dict = NULL, *result = NULL;
PyObject *items_iter, *items, *args = NULL;
/* capture any instance state */
dict = _PyObject_GetAttrId((PyObject *)od, &PyId___dict__);
if (dict == NULL)
goto Done;
else {
/* od.__dict__ isn't necessarily a dict... */
Py_ssize_t dict_len = PyObject_Length(dict);
if (dict_len == -1)
goto Done;
if (!dict_len) {
/* nothing to pickle in od.__dict__ */
Py_CLEAR(dict);
}
}
/* build the result */
args = PyTuple_New(0);
if (args == NULL)
goto Done;
items = _PyObject_CallMethodIdObjArgs((PyObject *)od, &PyId_items, NULL);
if (items == NULL)
goto Done;
items_iter = PyObject_GetIter(items);
Py_DECREF(items);
if (items_iter == NULL)
goto Done;
result = PyTuple_Pack(5, Py_TYPE(od), args, dict ? dict : Py_None, Py_None, items_iter);
Py_DECREF(items_iter);
Done:
Py_XDECREF(dict);
Py_XDECREF(args);
return result;