12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894 |
- #include "Python.h"
- #include "pycore_bitutils.h" // _Py_popcount32
- #include "pycore_hamt.h"
- #include "pycore_initconfig.h" // _PyStatus_OK()
- #include "pycore_object.h" // _PyObject_GC_TRACK()
- #include <stddef.h> // offsetof()
- /*
- This file provides an implementation of an immutable mapping using the
- Hash Array Mapped Trie (or HAMT) datastructure.
- This design allows to have:
- 1. Efficient copy: immutable mappings can be copied by reference,
- making it an O(1) operation.
- 2. Efficient mutations: due to structural sharing, only a portion of
- the trie needs to be copied when the collection is mutated. The
- cost of set/delete operations is O(log N).
- 3. Efficient lookups: O(log N).
- (where N is number of key/value items in the immutable mapping.)
- HAMT
- ====
- The core idea of HAMT is that the shape of the trie is encoded into the
- hashes of keys.
- Say we want to store a K/V pair in our mapping. First, we calculate the
- hash of K, let's say it's 19830128, or in binary:
- 0b1001011101001010101110000 = 19830128
- Now let's partition this bit representation of the hash into blocks of
- 5 bits each:
- 0b00_00000_10010_11101_00101_01011_10000 = 19830128
- (6) (5) (4) (3) (2) (1)
- Each block of 5 bits represents a number between 0 and 31. So if we have
- a tree that consists of nodes, each of which is an array of 32 pointers,
- those 5-bit blocks will encode a position on a single tree level.
- For example, storing the key K with hash 19830128, results in the following
- tree structure:
- (array of 32 pointers)
- +---+ -- +----+----+----+ -- +----+
- root node | 0 | .. | 15 | 16 | 17 | .. | 31 | 0b10000 = 16 (1)
- (level 1) +---+ -- +----+----+----+ -- +----+
- |
- +---+ -- +----+----+----+ -- +----+
- a 2nd level node | 0 | .. | 10 | 11 | 12 | .. | 31 | 0b01011 = 11 (2)
- +---+ -- +----+----+----+ -- +----+
- |
- +---+ -- +----+----+----+ -- +----+
- a 3rd level node | 0 | .. | 04 | 05 | 06 | .. | 31 | 0b00101 = 5 (3)
- +---+ -- +----+----+----+ -- +----+
- |
- +---+ -- +----+----+----+----+
- a 4th level node | 0 | .. | 04 | 29 | 30 | 31 | 0b11101 = 29 (4)
- +---+ -- +----+----+----+----+
- |
- +---+ -- +----+----+----+ -- +----+
- a 5th level node | 0 | .. | 17 | 18 | 19 | .. | 31 | 0b10010 = 18 (5)
- +---+ -- +----+----+----+ -- +----+
- |
- +--------------+
- |
- +---+ -- +----+----+----+ -- +----+
- a 6th level node | 0 | .. | 15 | 16 | 17 | .. | 31 | 0b00000 = 0 (6)
- +---+ -- +----+----+----+ -- +----+
- |
- V -- our value (or collision)
- To rehash: for a K/V pair, the hash of K encodes where in the tree V will
- be stored.
- To optimize memory footprint and handle hash collisions, our implementation
- uses three different types of nodes:
- * A Bitmap node;
- * An Array node;
- * A Collision node.
- Because we implement an immutable dictionary, our nodes are also
- immutable. Therefore, when we need to modify a node, we copy it, and
- do that modification to the copy.
- Array Nodes
- -----------
- These nodes are very simple. Essentially they are arrays of 32 pointers
- we used to illustrate the high-level idea in the previous section.
- We use Array nodes only when we need to store more than 16 pointers
- in a single node.
- Array nodes do not store key objects or value objects. They are used
- only as an indirection level - their pointers point to other nodes in
- the tree.
- Bitmap Node
- -----------
- Allocating a new 32-pointers array for every node of our tree would be
- very expensive. Unless we store millions of keys, most of tree nodes would
- be very sparse.
- When we have less than 16 elements in a node, we don't want to use the
- Array node, that would mean that we waste a lot of memory. Instead,
- we can use bitmap compression and can have just as many pointers
- as we need!
- Bitmap nodes consist of two fields:
- 1. An array of pointers. If a Bitmap node holds N elements, the
- array will be of N pointers.
- 2. A 32bit integer -- a bitmap field. If an N-th bit is set in the
- bitmap, it means that the node has an N-th element.
- For example, say we need to store a 3 elements sparse array:
- +---+ -- +---+ -- +----+ -- +----+
- | 0 | .. | 4 | .. | 11 | .. | 17 |
- +---+ -- +---+ -- +----+ -- +----+
- | | |
- o1 o2 o3
- We allocate a three-pointer Bitmap node. Its bitmap field will be
- then set to:
- 0b_00100_00010_00000_10000 == (1 << 17) | (1 << 11) | (1 << 4)
- To check if our Bitmap node has an I-th element we can do:
- bitmap & (1 << I)
- And here's a formula to calculate a position in our pointer array
- which would correspond to an I-th element:
- popcount(bitmap & ((1 << I) - 1))
- Let's break it down:
- * `popcount` is a function that returns a number of bits set to 1;
- * `((1 << I) - 1)` is a mask to filter the bitmask to contain bits
- set to the *right* of our bit.
- So for our 17, 11, and 4 indexes:
- * bitmap & ((1 << 17) - 1) == 0b100000010000 => 2 bits are set => index is 2.
- * bitmap & ((1 << 11) - 1) == 0b10000 => 1 bit is set => index is 1.
- * bitmap & ((1 << 4) - 1) == 0b0 => 0 bits are set => index is 0.
- To conclude: Bitmap nodes are just like Array nodes -- they can store
- a number of pointers, but use bitmap compression to eliminate unused
- pointers.
- Bitmap nodes have two pointers for each item:
- +----+----+----+----+ -- +----+----+
- | k1 | v1 | k2 | v2 | .. | kN | vN |
- +----+----+----+----+ -- +----+----+
- When kI == NULL, vI points to another tree level.
- When kI != NULL, the actual key object is stored in kI, and its
- value is stored in vI.
- Collision Nodes
- ---------------
- Collision nodes are simple arrays of pointers -- two pointers per
- key/value. When there's a hash collision, say for k1/v1 and k2/v2
- we have `hash(k1)==hash(k2)`. Then our collision node will be:
- +----+----+----+----+
- | k1 | v1 | k2 | v2 |
- +----+----+----+----+
- Tree Structure
- --------------
- All nodes are PyObjects.
- The `PyHamtObject` object has a pointer to the root node (h_root),
- and has a length field (h_count).
- High-level functions accept a PyHamtObject object and dispatch to
- lower-level functions depending on what kind of node h_root points to.
- Operations
- ==========
- There are three fundamental operations on an immutable dictionary:
- 1. "o.assoc(k, v)" will return a new immutable dictionary, that will be
- a copy of "o", but with the "k/v" item set.
- Functions in this file:
- hamt_node_assoc, hamt_node_bitmap_assoc,
- hamt_node_array_assoc, hamt_node_collision_assoc
- `hamt_node_assoc` function accepts a node object, and calls
- other functions depending on its actual type.
- 2. "o.find(k)" will lookup key "k" in "o".
- Functions:
- hamt_node_find, hamt_node_bitmap_find,
- hamt_node_array_find, hamt_node_collision_find
- 3. "o.without(k)" will return a new immutable dictionary, that will be
- a copy of "o", buth without the "k" key.
- Functions:
- hamt_node_without, hamt_node_bitmap_without,
- hamt_node_array_without, hamt_node_collision_without
- Further Reading
- ===============
- 1. http://blog.higher-order.net/2009/09/08/understanding-clojures-persistenthashmap-deftwice.html
- 2. http://blog.higher-order.net/2010/08/16/assoc-and-clojures-persistenthashmap-part-ii.html
- 3. Clojure's PersistentHashMap implementation:
- https://github.com/clojure/clojure/blob/master/src/jvm/clojure/lang/PersistentHashMap.java
- Debug
- =====
- The HAMT datatype is accessible for testing purposes under the
- `_testcapi` module:
- >>> from _testcapi import hamt
- >>> h = hamt()
- >>> h2 = h.set('a', 2)
- >>> h3 = h2.set('b', 3)
- >>> list(h3)
- ['a', 'b']
- When CPython is built in debug mode, a '__dump__()' method is available
- to introspect the tree:
- >>> print(h3.__dump__())
- HAMT(len=2):
- BitmapNode(size=4 count=2 bitmap=0b110 id=0x10eb9d9e8):
- 'a': 2
- 'b': 3
- */
- #define IS_ARRAY_NODE(node) Py_IS_TYPE(node, &_PyHamt_ArrayNode_Type)
- #define IS_BITMAP_NODE(node) Py_IS_TYPE(node, &_PyHamt_BitmapNode_Type)
- #define IS_COLLISION_NODE(node) Py_IS_TYPE(node, &_PyHamt_CollisionNode_Type)
- /* Return type for 'find' (lookup a key) functions.
- * F_ERROR - an error occurred;
- * F_NOT_FOUND - the key was not found;
- * F_FOUND - the key was found.
- */
- typedef enum {F_ERROR, F_NOT_FOUND, F_FOUND} hamt_find_t;
- /* Return type for 'without' (delete a key) functions.
- * W_ERROR - an error occurred;
- * W_NOT_FOUND - the key was not found: there's nothing to delete;
- * W_EMPTY - the key was found: the node/tree would be empty
- if the key is deleted;
- * W_NEWNODE - the key was found: a new node/tree is returned
- without that key.
- */
- typedef enum {W_ERROR, W_NOT_FOUND, W_EMPTY, W_NEWNODE} hamt_without_t;
- /* Low-level iterator protocol type.
- * I_ITEM - a new item has been yielded;
- * I_END - the whole tree was visited (similar to StopIteration).
- */
- typedef enum {I_ITEM, I_END} hamt_iter_t;
- #define HAMT_ARRAY_NODE_SIZE 32
- typedef struct {
- PyObject_HEAD
- PyHamtNode *a_array[HAMT_ARRAY_NODE_SIZE];
- Py_ssize_t a_count;
- } PyHamtNode_Array;
- typedef struct {
- PyObject_VAR_HEAD
- int32_t c_hash;
- PyObject *c_array[1];
- } PyHamtNode_Collision;
- static PyHamtObject *
- hamt_alloc(void);
- static PyHamtNode *
- hamt_node_assoc(PyHamtNode *node,
- uint32_t shift, int32_t hash,
- PyObject *key, PyObject *val, int* added_leaf);
- static hamt_without_t
- hamt_node_without(PyHamtNode *node,
- uint32_t shift, int32_t hash,
- PyObject *key,
- PyHamtNode **new_node);
- static hamt_find_t
- hamt_node_find(PyHamtNode *node,
- uint32_t shift, int32_t hash,
- PyObject *key, PyObject **val);
- #ifdef Py_DEBUG
- static int
- hamt_node_dump(PyHamtNode *node,
- _PyUnicodeWriter *writer, int level);
- #endif
- static PyHamtNode *
- hamt_node_array_new(Py_ssize_t);
- static PyHamtNode *
- hamt_node_collision_new(int32_t hash, Py_ssize_t size);
- static inline Py_ssize_t
- hamt_node_collision_count(PyHamtNode_Collision *node);
- #ifdef Py_DEBUG
- static void
- _hamt_node_array_validate(void *obj_raw)
- {
- PyObject *obj = _PyObject_CAST(obj_raw);
- assert(IS_ARRAY_NODE(obj));
- PyHamtNode_Array *node = (PyHamtNode_Array*)obj;
- Py_ssize_t i = 0, count = 0;
- for (; i < HAMT_ARRAY_NODE_SIZE; i++) {
- if (node->a_array[i] != NULL) {
- count++;
- }
- }
- assert(count == node->a_count);
- }
- #define VALIDATE_ARRAY_NODE(NODE) \
- do { _hamt_node_array_validate(NODE); } while (0);
- #else
- #define VALIDATE_ARRAY_NODE(NODE)
- #endif
- /* Returns -1 on error */
- static inline int32_t
- hamt_hash(PyObject *o)
- {
- Py_hash_t hash = PyObject_Hash(o);
- #if SIZEOF_PY_HASH_T <= 4
- return hash;
- #else
- if (hash == -1) {
- /* exception */
- return -1;
- }
- /* While it's somewhat suboptimal to reduce Python's 64 bit hash to
- 32 bits via XOR, it seems that the resulting hash function
- is good enough (this is also how Long type is hashed in Java.)
- Storing 10, 100, 1000 Python strings results in a relatively
- shallow and uniform tree structure.
- Also it's worth noting that it would be possible to adapt the tree
- structure to 64 bit hashes, but that would increase memory pressure
- and provide little to no performance benefits for collections with
- fewer than billions of key/value pairs.
- Important: do not change this hash reducing function. There are many
- tests that need an exact tree shape to cover all code paths and
- we do that by specifying concrete values for test data's `__hash__`.
- If this function is changed most of the regression tests would
- become useless.
- */
- int32_t xored = (int32_t)(hash & 0xffffffffl) ^ (int32_t)(hash >> 32);
- return xored == -1 ? -2 : xored;
- #endif
- }
- static inline uint32_t
- hamt_mask(int32_t hash, uint32_t shift)
- {
- return (((uint32_t)hash >> shift) & 0x01f);
- }
- static inline uint32_t
- hamt_bitpos(int32_t hash, uint32_t shift)
- {
- return (uint32_t)1 << hamt_mask(hash, shift);
- }
- static inline uint32_t
- hamt_bitindex(uint32_t bitmap, uint32_t bit)
- {
- return (uint32_t)_Py_popcount32(bitmap & (bit - 1));
- }
- /////////////////////////////////// Dump Helpers
- #ifdef Py_DEBUG
- static int
- _hamt_dump_ident(_PyUnicodeWriter *writer, int level)
- {
- /* Write `' ' * level` to the `writer` */
- PyObject *str = NULL;
- PyObject *num = NULL;
- PyObject *res = NULL;
- int ret = -1;
- str = PyUnicode_FromString(" ");
- if (str == NULL) {
- goto error;
- }
- num = PyLong_FromLong((long)level);
- if (num == NULL) {
- goto error;
- }
- res = PyNumber_Multiply(str, num);
- if (res == NULL) {
- goto error;
- }
- ret = _PyUnicodeWriter_WriteStr(writer, res);
- error:
- Py_XDECREF(res);
- Py_XDECREF(str);
- Py_XDECREF(num);
- return ret;
- }
- static int
- _hamt_dump_format(_PyUnicodeWriter *writer, const char *format, ...)
- {
- /* A convenient helper combining _PyUnicodeWriter_WriteStr and
- PyUnicode_FromFormatV.
- */
- PyObject* msg;
- int ret;
- va_list vargs;
- va_start(vargs, format);
- msg = PyUnicode_FromFormatV(format, vargs);
- va_end(vargs);
- if (msg == NULL) {
- return -1;
- }
- ret = _PyUnicodeWriter_WriteStr(writer, msg);
- Py_DECREF(msg);
- return ret;
- }
- #endif /* Py_DEBUG */
- /////////////////////////////////// Bitmap Node
- static PyHamtNode *
- hamt_node_bitmap_new(Py_ssize_t size)
- {
- /* Create a new bitmap node of size 'size' */
- PyHamtNode_Bitmap *node;
- Py_ssize_t i;
- if (size == 0) {
- /* Since bitmap nodes are immutable, we can cache the instance
- for size=0 and reuse it whenever we need an empty bitmap node.
- */
- return (PyHamtNode *)Py_NewRef(&_Py_SINGLETON(hamt_bitmap_node_empty));
- }
- assert(size >= 0);
- assert(size % 2 == 0);
- /* No freelist; allocate a new bitmap node */
- node = PyObject_GC_NewVar(
- PyHamtNode_Bitmap, &_PyHamt_BitmapNode_Type, size);
- if (node == NULL) {
- return NULL;
- }
- Py_SET_SIZE(node, size);
- for (i = 0; i < size; i++) {
- node->b_array[i] = NULL;
- }
- node->b_bitmap = 0;
- _PyObject_GC_TRACK(node);
- return (PyHamtNode *)node;
- }
- static inline Py_ssize_t
- hamt_node_bitmap_count(PyHamtNode_Bitmap *node)
- {
- return Py_SIZE(node) / 2;
- }
- static PyHamtNode_Bitmap *
- hamt_node_bitmap_clone(PyHamtNode_Bitmap *node)
- {
- /* Clone a bitmap node; return a new one with the same child notes. */
- PyHamtNode_Bitmap *clone;
- Py_ssize_t i;
- clone = (PyHamtNode_Bitmap *)hamt_node_bitmap_new(Py_SIZE(node));
- if (clone == NULL) {
- return NULL;
- }
- for (i = 0; i < Py_SIZE(node); i++) {
- clone->b_array[i] = Py_XNewRef(node->b_array[i]);
- }
- clone->b_bitmap = node->b_bitmap;
- return clone;
- }
- static PyHamtNode_Bitmap *
- hamt_node_bitmap_clone_without(PyHamtNode_Bitmap *o, uint32_t bit)
- {
- assert(bit & o->b_bitmap);
- assert(hamt_node_bitmap_count(o) > 1);
- PyHamtNode_Bitmap *new = (PyHamtNode_Bitmap *)hamt_node_bitmap_new(
- Py_SIZE(o) - 2);
- if (new == NULL) {
- return NULL;
- }
- uint32_t idx = hamt_bitindex(o->b_bitmap, bit);
- uint32_t key_idx = 2 * idx;
- uint32_t val_idx = key_idx + 1;
- uint32_t i;
- for (i = 0; i < key_idx; i++) {
- new->b_array[i] = Py_XNewRef(o->b_array[i]);
- }
- assert(Py_SIZE(o) >= 0 && Py_SIZE(o) <= 32);
- for (i = val_idx + 1; i < (uint32_t)Py_SIZE(o); i++) {
- new->b_array[i - 2] = Py_XNewRef(o->b_array[i]);
- }
- new->b_bitmap = o->b_bitmap & ~bit;
- return new;
- }
- static PyHamtNode *
- hamt_node_new_bitmap_or_collision(uint32_t shift,
- PyObject *key1, PyObject *val1,
- int32_t key2_hash,
- PyObject *key2, PyObject *val2)
- {
- /* Helper method. Creates a new node for key1/val and key2/val2
- pairs.
- If key1 hash is equal to the hash of key2, a Collision node
- will be created. If they are not equal, a Bitmap node is
- created.
- */
- int32_t key1_hash = hamt_hash(key1);
- if (key1_hash == -1) {
- return NULL;
- }
- if (key1_hash == key2_hash) {
- PyHamtNode_Collision *n;
- n = (PyHamtNode_Collision *)hamt_node_collision_new(key1_hash, 4);
- if (n == NULL) {
- return NULL;
- }
- n->c_array[0] = Py_NewRef(key1);
- n->c_array[1] = Py_NewRef(val1);
- n->c_array[2] = Py_NewRef(key2);
- n->c_array[3] = Py_NewRef(val2);
- return (PyHamtNode *)n;
- }
- else {
- int added_leaf = 0;
- PyHamtNode *n = hamt_node_bitmap_new(0);
- if (n == NULL) {
- return NULL;
- }
- PyHamtNode *n2 = hamt_node_assoc(
- n, shift, key1_hash, key1, val1, &added_leaf);
- Py_DECREF(n);
- if (n2 == NULL) {
- return NULL;
- }
- n = hamt_node_assoc(n2, shift, key2_hash, key2, val2, &added_leaf);
- Py_DECREF(n2);
- if (n == NULL) {
- return NULL;
- }
- return n;
- }
- }
- static PyHamtNode *
- hamt_node_bitmap_assoc(PyHamtNode_Bitmap *self,
- uint32_t shift, int32_t hash,
- PyObject *key, PyObject *val, int* added_leaf)
- {
- /* assoc operation for bitmap nodes.
- Return: a new node, or self if key/val already is in the
- collection.
- 'added_leaf' is later used in '_PyHamt_Assoc' to determine if
- `hamt.set(key, val)` increased the size of the collection.
- */
- uint32_t bit = hamt_bitpos(hash, shift);
- uint32_t idx = hamt_bitindex(self->b_bitmap, bit);
- /* Bitmap node layout:
- +------+------+------+------+ --- +------+------+
- | key1 | val1 | key2 | val2 | ... | keyN | valN |
- +------+------+------+------+ --- +------+------+
- where `N < Py_SIZE(node)`.
- The `node->b_bitmap` field is a bitmap. For a given
- `(shift, hash)` pair we can determine:
- - If this node has the corresponding key/val slots.
- - The index of key/val slots.
- */
- if (self->b_bitmap & bit) {
- /* The key is set in this node */
- uint32_t key_idx = 2 * idx;
- uint32_t val_idx = key_idx + 1;
- assert(val_idx < (size_t)Py_SIZE(self));
- PyObject *key_or_null = self->b_array[key_idx];
- PyObject *val_or_node = self->b_array[val_idx];
- if (key_or_null == NULL) {
- /* key is NULL. This means that we have a few keys
- that have the same (hash, shift) pair. */
- assert(val_or_node != NULL);
- PyHamtNode *sub_node = hamt_node_assoc(
- (PyHamtNode *)val_or_node,
- shift + 5, hash, key, val, added_leaf);
- if (sub_node == NULL) {
- return NULL;
- }
- if (val_or_node == (PyObject *)sub_node) {
- Py_DECREF(sub_node);
- return (PyHamtNode *)Py_NewRef(self);
- }
- PyHamtNode_Bitmap *ret = hamt_node_bitmap_clone(self);
- if (ret == NULL) {
- return NULL;
- }
- Py_SETREF(ret->b_array[val_idx], (PyObject*)sub_node);
- return (PyHamtNode *)ret;
- }
- assert(key != NULL);
- /* key is not NULL. This means that we have only one other
- key in this collection that matches our hash for this shift. */
- int comp_err = PyObject_RichCompareBool(key, key_or_null, Py_EQ);
- if (comp_err < 0) { /* exception in __eq__ */
- return NULL;
- }
- if (comp_err == 1) { /* key == key_or_null */
- if (val == val_or_node) {
- /* we already have the same key/val pair; return self. */
- return (PyHamtNode *)Py_NewRef(self);
- }
- /* We're setting a new value for the key we had before.
- Make a new bitmap node with a replaced value, and return it. */
- PyHamtNode_Bitmap *ret = hamt_node_bitmap_clone(self);
- if (ret == NULL) {
- return NULL;
- }
- Py_SETREF(ret->b_array[val_idx], Py_NewRef(val));
- return (PyHamtNode *)ret;
- }
- /* It's a new key, and it has the same index as *one* another key.
- We have a collision. We need to create a new node which will
- combine the existing key and the key we're adding.
- `hamt_node_new_bitmap_or_collision` will either create a new
- Collision node if the keys have identical hashes, or
- a new Bitmap node.
- */
- PyHamtNode *sub_node = hamt_node_new_bitmap_or_collision(
- shift + 5,
- key_or_null, val_or_node, /* existing key/val */
- hash,
- key, val /* new key/val */
- );
- if (sub_node == NULL) {
- return NULL;
- }
- PyHamtNode_Bitmap *ret = hamt_node_bitmap_clone(self);
- if (ret == NULL) {
- Py_DECREF(sub_node);
- return NULL;
- }
- Py_SETREF(ret->b_array[key_idx], NULL);
- Py_SETREF(ret->b_array[val_idx], (PyObject *)sub_node);
- *added_leaf = 1;
- return (PyHamtNode *)ret;
- }
- else {
- /* There was no key before with the same (shift,hash). */
- uint32_t n = (uint32_t)_Py_popcount32(self->b_bitmap);
- if (n >= 16) {
- /* When we have a situation where we want to store more
- than 16 nodes at one level of the tree, we no longer
- want to use the Bitmap node with bitmap encoding.
- Instead we start using an Array node, which has
- simpler (faster) implementation at the expense of
- having preallocated 32 pointers for its keys/values
- pairs.
- Small hamt objects (<30 keys) usually don't have any
- Array nodes at all. Between ~30 and ~400 keys hamt
- objects usually have one Array node, and usually it's
- a root node.
- */
- uint32_t jdx = hamt_mask(hash, shift);
- /* 'jdx' is the index of where the new key should be added
- in the new Array node we're about to create. */
- PyHamtNode *empty = NULL;
- PyHamtNode_Array *new_node = NULL;
- PyHamtNode *res = NULL;
- /* Create a new Array node. */
- new_node = (PyHamtNode_Array *)hamt_node_array_new(n + 1);
- if (new_node == NULL) {
- goto fin;
- }
- /* Create an empty bitmap node for the next
- hamt_node_assoc call. */
- empty = hamt_node_bitmap_new(0);
- if (empty == NULL) {
- goto fin;
- }
- /* Make a new bitmap node for the key/val we're adding.
- Set that bitmap node to new-array-node[jdx]. */
- new_node->a_array[jdx] = hamt_node_assoc(
- empty, shift + 5, hash, key, val, added_leaf);
- if (new_node->a_array[jdx] == NULL) {
- goto fin;
- }
- /* Copy existing key/value pairs from the current Bitmap
- node to the new Array node we've just created. */
- Py_ssize_t i, j;
- for (i = 0, j = 0; i < HAMT_ARRAY_NODE_SIZE; i++) {
- if (((self->b_bitmap >> i) & 1) != 0) {
- /* Ensure we don't accidentally override `jdx` element
- we set few lines above.
- */
- assert(new_node->a_array[i] == NULL);
- if (self->b_array[j] == NULL) {
- new_node->a_array[i] =
- (PyHamtNode *)Py_NewRef(self->b_array[j + 1]);
- }
- else {
- int32_t rehash = hamt_hash(self->b_array[j]);
- if (rehash == -1) {
- goto fin;
- }
- new_node->a_array[i] = hamt_node_assoc(
- empty, shift + 5,
- rehash,
- self->b_array[j],
- self->b_array[j + 1],
- added_leaf);
- if (new_node->a_array[i] == NULL) {
- goto fin;
- }
- }
- j += 2;
- }
- }
- VALIDATE_ARRAY_NODE(new_node)
- /* That's it! */
- res = (PyHamtNode *)new_node;
- fin:
- Py_XDECREF(empty);
- if (res == NULL) {
- Py_XDECREF(new_node);
- }
- return res;
- }
- else {
- /* We have less than 16 keys at this level; let's just
- create a new bitmap node out of this node with the
- new key/val pair added. */
- uint32_t key_idx = 2 * idx;
- uint32_t val_idx = key_idx + 1;
- uint32_t i;
- *added_leaf = 1;
- /* Allocate new Bitmap node which can have one more key/val
- pair in addition to what we have already. */
- PyHamtNode_Bitmap *new_node =
- (PyHamtNode_Bitmap *)hamt_node_bitmap_new(2 * (n + 1));
- if (new_node == NULL) {
- return NULL;
- }
- /* Copy all keys/values that will be before the new key/value
- we are adding. */
- for (i = 0; i < key_idx; i++) {
- new_node->b_array[i] = Py_XNewRef(self->b_array[i]);
- }
- /* Set the new key/value to the new Bitmap node. */
- new_node->b_array[key_idx] = Py_NewRef(key);
- new_node->b_array[val_idx] = Py_NewRef(val);
- /* Copy all keys/values that will be after the new key/value
- we are adding. */
- assert(Py_SIZE(self) >= 0 && Py_SIZE(self) <= 32);
- for (i = key_idx; i < (uint32_t)Py_SIZE(self); i++) {
- new_node->b_array[i + 2] = Py_XNewRef(self->b_array[i]);
- }
- new_node->b_bitmap = self->b_bitmap | bit;
- return (PyHamtNode *)new_node;
- }
- }
- }
- static hamt_without_t
- hamt_node_bitmap_without(PyHamtNode_Bitmap *self,
- uint32_t shift, int32_t hash,
- PyObject *key,
- PyHamtNode **new_node)
- {
- uint32_t bit = hamt_bitpos(hash, shift);
- if ((self->b_bitmap & bit) == 0) {
- return W_NOT_FOUND;
- }
- uint32_t idx = hamt_bitindex(self->b_bitmap, bit);
- uint32_t key_idx = 2 * idx;
- uint32_t val_idx = key_idx + 1;
- PyObject *key_or_null = self->b_array[key_idx];
- PyObject *val_or_node = self->b_array[val_idx];
- if (key_or_null == NULL) {
- /* key == NULL means that 'value' is another tree node. */
- PyHamtNode *sub_node = NULL;
- hamt_without_t res = hamt_node_without(
- (PyHamtNode *)val_or_node,
- shift + 5, hash, key, &sub_node);
- switch (res) {
- case W_EMPTY:
- /* It's impossible for us to receive a W_EMPTY here:
- - Array nodes are converted to Bitmap nodes when
- we delete 16th item from them;
- - Collision nodes are converted to Bitmap when
- there is one item in them;
- - Bitmap node's without() inlines single-item
- sub-nodes.
- So in no situation we can have a single-item
- Bitmap child of another Bitmap node.
- */
- Py_UNREACHABLE();
- case W_NEWNODE: {
- assert(sub_node != NULL);
- if (IS_BITMAP_NODE(sub_node)) {
- PyHamtNode_Bitmap *sub_tree = (PyHamtNode_Bitmap *)sub_node;
- if (hamt_node_bitmap_count(sub_tree) == 1 &&
- sub_tree->b_array[0] != NULL)
- {
- /* A bitmap node with one key/value pair. Just
- merge it into this node.
- Note that we don't inline Bitmap nodes that
- have a NULL key -- those nodes point to another
- tree level, and we cannot simply move tree levels
- up or down.
- */
- PyHamtNode_Bitmap *clone = hamt_node_bitmap_clone(self);
- if (clone == NULL) {
- Py_DECREF(sub_node);
- return W_ERROR;
- }
- PyObject *key = sub_tree->b_array[0];
- PyObject *val = sub_tree->b_array[1];
- Py_XSETREF(clone->b_array[key_idx], Py_NewRef(key));
- Py_SETREF(clone->b_array[val_idx], Py_NewRef(val));
- Py_DECREF(sub_tree);
- *new_node = (PyHamtNode *)clone;
- return W_NEWNODE;
- }
- }
- #ifdef Py_DEBUG
- /* Ensure that Collision.without implementation
- converts to Bitmap nodes itself.
- */
- if (IS_COLLISION_NODE(sub_node)) {
- assert(hamt_node_collision_count(
- (PyHamtNode_Collision*)sub_node) > 1);
- }
- #endif
- PyHamtNode_Bitmap *clone = hamt_node_bitmap_clone(self);
- if (clone == NULL) {
- return W_ERROR;
- }
- Py_SETREF(clone->b_array[val_idx],
- (PyObject *)sub_node); /* borrow */
- *new_node = (PyHamtNode *)clone;
- return W_NEWNODE;
- }
- case W_ERROR:
- case W_NOT_FOUND:
- assert(sub_node == NULL);
- return res;
- default:
- Py_UNREACHABLE();
- }
- }
- else {
- /* We have a regular key/value pair */
- int cmp = PyObject_RichCompareBool(key_or_null, key, Py_EQ);
- if (cmp < 0) {
- return W_ERROR;
- }
- if (cmp == 0) {
- return W_NOT_FOUND;
- }
- if (hamt_node_bitmap_count(self) == 1) {
- return W_EMPTY;
- }
- *new_node = (PyHamtNode *)
- hamt_node_bitmap_clone_without(self, bit);
- if (*new_node == NULL) {
- return W_ERROR;
- }
- return W_NEWNODE;
- }
- }
- static hamt_find_t
- hamt_node_bitmap_find(PyHamtNode_Bitmap *self,
- uint32_t shift, int32_t hash,
- PyObject *key, PyObject **val)
- {
- /* Lookup a key in a Bitmap node. */
- uint32_t bit = hamt_bitpos(hash, shift);
- uint32_t idx;
- uint32_t key_idx;
- uint32_t val_idx;
- PyObject *key_or_null;
- PyObject *val_or_node;
- int comp_err;
- if ((self->b_bitmap & bit) == 0) {
- return F_NOT_FOUND;
- }
- idx = hamt_bitindex(self->b_bitmap, bit);
- key_idx = idx * 2;
- val_idx = key_idx + 1;
- assert(val_idx < (size_t)Py_SIZE(self));
- key_or_null = self->b_array[key_idx];
- val_or_node = self->b_array[val_idx];
- if (key_or_null == NULL) {
- /* There are a few keys that have the same hash at the current shift
- that match our key. Dispatch the lookup further down the tree. */
- assert(val_or_node != NULL);
- return hamt_node_find((PyHamtNode *)val_or_node,
- shift + 5, hash, key, val);
- }
- /* We have only one key -- a potential match. Let's compare if the
- key we are looking at is equal to the key we are looking for. */
- assert(key != NULL);
- comp_err = PyObject_RichCompareBool(key, key_or_null, Py_EQ);
- if (comp_err < 0) { /* exception in __eq__ */
- return F_ERROR;
- }
- if (comp_err == 1) { /* key == key_or_null */
- *val = val_or_node;
- return F_FOUND;
- }
- return F_NOT_FOUND;
- }
- static int
- hamt_node_bitmap_traverse(PyHamtNode_Bitmap *self, visitproc visit, void *arg)
- {
- /* Bitmap's tp_traverse */
- Py_ssize_t i;
- for (i = Py_SIZE(self); --i >= 0; ) {
- Py_VISIT(self->b_array[i]);
- }
- return 0;
- }
- static void
- hamt_node_bitmap_dealloc(PyHamtNode_Bitmap *self)
- {
- /* Bitmap's tp_dealloc */
- Py_ssize_t len = Py_SIZE(self);
- Py_ssize_t i;
- if (Py_SIZE(self) == 0) {
- /* The empty node is statically allocated. */
- assert(self == &_Py_SINGLETON(hamt_bitmap_node_empty));
- #ifdef Py_DEBUG
- _Py_FatalRefcountError("deallocating the empty hamt node bitmap singleton");
- #else
- return;
- #endif
- }
- PyObject_GC_UnTrack(self);
- Py_TRASHCAN_BEGIN(self, hamt_node_bitmap_dealloc)
- if (len > 0) {
- i = len;
- while (--i >= 0) {
- Py_XDECREF(self->b_array[i]);
- }
- }
- Py_TYPE(self)->tp_free((PyObject *)self);
- Py_TRASHCAN_END
- }
- #ifdef Py_DEBUG
- static int
- hamt_node_bitmap_dump(PyHamtNode_Bitmap *node,
- _PyUnicodeWriter *writer, int level)
- {
- /* Debug build: __dump__() method implementation for Bitmap nodes. */
- Py_ssize_t i;
- PyObject *tmp1;
- PyObject *tmp2;
- if (_hamt_dump_ident(writer, level + 1)) {
- goto error;
- }
- if (_hamt_dump_format(writer, "BitmapNode(size=%zd count=%zd ",
- Py_SIZE(node), Py_SIZE(node) / 2))
- {
- goto error;
- }
- tmp1 = PyLong_FromUnsignedLong(node->b_bitmap);
- if (tmp1 == NULL) {
- goto error;
- }
- tmp2 = _PyLong_Format(tmp1, 2);
- Py_DECREF(tmp1);
- if (tmp2 == NULL) {
- goto error;
- }
- if (_hamt_dump_format(writer, "bitmap=%S id=%p):\n", tmp2, node)) {
- Py_DECREF(tmp2);
- goto error;
- }
- Py_DECREF(tmp2);
- for (i = 0; i < Py_SIZE(node); i += 2) {
- PyObject *key_or_null = node->b_array[i];
- PyObject *val_or_node = node->b_array[i + 1];
- if (_hamt_dump_ident(writer, level + 2)) {
- goto error;
- }
- if (key_or_null == NULL) {
- if (_hamt_dump_format(writer, "NULL:\n")) {
- goto error;
- }
- if (hamt_node_dump((PyHamtNode *)val_or_node,
- writer, level + 2))
- {
- goto error;
- }
- }
- else {
- if (_hamt_dump_format(writer, "%R: %R", key_or_null,
- val_or_node))
- {
- goto error;
- }
- }
- if (_hamt_dump_format(writer, "\n")) {
- goto error;
- }
- }
- return 0;
- error:
- return -1;
- }
- #endif /* Py_DEBUG */
- /////////////////////////////////// Collision Node
- static PyHamtNode *
- hamt_node_collision_new(int32_t hash, Py_ssize_t size)
- {
- /* Create a new Collision node. */
- PyHamtNode_Collision *node;
- Py_ssize_t i;
- assert(size >= 4);
- assert(size % 2 == 0);
- node = PyObject_GC_NewVar(
- PyHamtNode_Collision, &_PyHamt_CollisionNode_Type, size);
- if (node == NULL) {
- return NULL;
- }
- for (i = 0; i < size; i++) {
- node->c_array[i] = NULL;
- }
- Py_SET_SIZE(node, size);
- node->c_hash = hash;
- _PyObject_GC_TRACK(node);
- return (PyHamtNode *)node;
- }
- static hamt_find_t
- hamt_node_collision_find_index(PyHamtNode_Collision *self, PyObject *key,
- Py_ssize_t *idx)
- {
- /* Lookup `key` in the Collision node `self`. Set the index of the
- found key to 'idx'. */
- Py_ssize_t i;
- PyObject *el;
- for (i = 0; i < Py_SIZE(self); i += 2) {
- el = self->c_array[i];
- assert(el != NULL);
- int cmp = PyObject_RichCompareBool(key, el, Py_EQ);
- if (cmp < 0) {
- return F_ERROR;
- }
- if (cmp == 1) {
- *idx = i;
- return F_FOUND;
- }
- }
- return F_NOT_FOUND;
- }
- static PyHamtNode *
- hamt_node_collision_assoc(PyHamtNode_Collision *self,
- uint32_t shift, int32_t hash,
- PyObject *key, PyObject *val, int* added_leaf)
- {
- /* Set a new key to this level (currently a Collision node)
- of the tree. */
- if (hash == self->c_hash) {
- /* The hash of the 'key' we are adding matches the hash of
- other keys in this Collision node. */
- Py_ssize_t key_idx = -1;
- hamt_find_t found;
- PyHamtNode_Collision *new_node;
- Py_ssize_t i;
- /* Let's try to lookup the new 'key', maybe we already have it. */
- found = hamt_node_collision_find_index(self, key, &key_idx);
- switch (found) {
- case F_ERROR:
- /* Exception. */
- return NULL;
- case F_NOT_FOUND:
- /* This is a totally new key. Clone the current node,
- add a new key/value to the cloned node. */
- new_node = (PyHamtNode_Collision *)hamt_node_collision_new(
- self->c_hash, Py_SIZE(self) + 2);
- if (new_node == NULL) {
- return NULL;
- }
- for (i = 0; i < Py_SIZE(self); i++) {
- new_node->c_array[i] = Py_NewRef(self->c_array[i]);
- }
- new_node->c_array[i] = Py_NewRef(key);
- new_node->c_array[i + 1] = Py_NewRef(val);
- *added_leaf = 1;
- return (PyHamtNode *)new_node;
- case F_FOUND:
- /* There's a key which is equal to the key we are adding. */
- assert(key_idx >= 0);
- assert(key_idx < Py_SIZE(self));
- Py_ssize_t val_idx = key_idx + 1;
- if (self->c_array[val_idx] == val) {
- /* We're setting a key/value pair that's already set. */
- return (PyHamtNode *)Py_NewRef(self);
- }
- /* We need to replace old value for the key
- with a new value. Create a new Collision node.*/
- new_node = (PyHamtNode_Collision *)hamt_node_collision_new(
- self->c_hash, Py_SIZE(self));
- if (new_node == NULL) {
- return NULL;
- }
- /* Copy all elements of the old node to the new one. */
- for (i = 0; i < Py_SIZE(self); i++) {
- new_node->c_array[i] = Py_NewRef(self->c_array[i]);
- }
- /* Replace the old value with the new value for the our key. */
- Py_SETREF(new_node->c_array[val_idx], Py_NewRef(val));
- return (PyHamtNode *)new_node;
- default:
- Py_UNREACHABLE();
- }
- }
- else {
- /* The hash of the new key is different from the hash that
- all keys of this Collision node have.
- Create a Bitmap node inplace with two children:
- key/value pair that we're adding, and the Collision node
- we're replacing on this tree level.
- */
- PyHamtNode_Bitmap *new_node;
- PyHamtNode *assoc_res;
- new_node = (PyHamtNode_Bitmap *)hamt_node_bitmap_new(2);
- if (new_node == NULL) {
- return NULL;
- }
- new_node->b_bitmap = hamt_bitpos(self->c_hash, shift);
- new_node->b_array[1] = Py_NewRef(self);
- assoc_res = hamt_node_bitmap_assoc(
- new_node, shift, hash, key, val, added_leaf);
- Py_DECREF(new_node);
- return assoc_res;
- }
- }
- static inline Py_ssize_t
- hamt_node_collision_count(PyHamtNode_Collision *node)
- {
- return Py_SIZE(node) / 2;
- }
- static hamt_without_t
- hamt_node_collision_without(PyHamtNode_Collision *self,
- uint32_t shift, int32_t hash,
- PyObject *key,
- PyHamtNode **new_node)
- {
- if (hash != self->c_hash) {
- return W_NOT_FOUND;
- }
- Py_ssize_t key_idx = -1;
- hamt_find_t found = hamt_node_collision_find_index(self, key, &key_idx);
- switch (found) {
- case F_ERROR:
- return W_ERROR;
- case F_NOT_FOUND:
- return W_NOT_FOUND;
- case F_FOUND:
- assert(key_idx >= 0);
- assert(key_idx < Py_SIZE(self));
- Py_ssize_t new_count = hamt_node_collision_count(self) - 1;
- if (new_count == 0) {
- /* The node has only one key/value pair and it's for the
- key we're trying to delete. So a new node will be empty
- after the removal.
- */
- return W_EMPTY;
- }
- if (new_count == 1) {
- /* The node has two keys, and after deletion the
- new Collision node would have one. Collision nodes
- with one key shouldn't exist, so convert it to a
- Bitmap node.
- */
- PyHamtNode_Bitmap *node = (PyHamtNode_Bitmap *)
- hamt_node_bitmap_new(2);
- if (node == NULL) {
- return W_ERROR;
- }
- if (key_idx == 0) {
- node->b_array[0] = Py_NewRef(self->c_array[2]);
- node->b_array[1] = Py_NewRef(self->c_array[3]);
- }
- else {
- assert(key_idx == 2);
- node->b_array[0] = Py_NewRef(self->c_array[0]);
- node->b_array[1] = Py_NewRef(self->c_array[1]);
- }
- node->b_bitmap = hamt_bitpos(hash, shift);
- *new_node = (PyHamtNode *)node;
- return W_NEWNODE;
- }
- /* Allocate a new Collision node with capacity for one
- less key/value pair */
- PyHamtNode_Collision *new = (PyHamtNode_Collision *)
- hamt_node_collision_new(
- self->c_hash, Py_SIZE(self) - 2);
- if (new == NULL) {
- return W_ERROR;
- }
- /* Copy all other keys from `self` to `new` */
- Py_ssize_t i;
- for (i = 0; i < key_idx; i++) {
- new->c_array[i] = Py_NewRef(self->c_array[i]);
- }
- for (i = key_idx + 2; i < Py_SIZE(self); i++) {
- new->c_array[i - 2] = Py_NewRef(self->c_array[i]);
- }
- *new_node = (PyHamtNode*)new;
- return W_NEWNODE;
- default:
- Py_UNREACHABLE();
- }
- }
- static hamt_find_t
- hamt_node_collision_find(PyHamtNode_Collision *self,
- uint32_t shift, int32_t hash,
- PyObject *key, PyObject **val)
- {
- /* Lookup `key` in the Collision node `self`. Set the value
- for the found key to 'val'. */
- Py_ssize_t idx = -1;
- hamt_find_t res;
- res = hamt_node_collision_find_index(self, key, &idx);
- if (res == F_ERROR || res == F_NOT_FOUND) {
- return res;
- }
- assert(idx >= 0);
- assert(idx + 1 < Py_SIZE(self));
- *val = self->c_array[idx + 1];
- assert(*val != NULL);
- return F_FOUND;
- }
- static int
- hamt_node_collision_traverse(PyHamtNode_Collision *self,
- visitproc visit, void *arg)
- {
- /* Collision's tp_traverse */
- Py_ssize_t i;
- for (i = Py_SIZE(self); --i >= 0; ) {
- Py_VISIT(self->c_array[i]);
- }
- return 0;
- }
- static void
- hamt_node_collision_dealloc(PyHamtNode_Collision *self)
- {
- /* Collision's tp_dealloc */
- Py_ssize_t len = Py_SIZE(self);
- PyObject_GC_UnTrack(self);
- Py_TRASHCAN_BEGIN(self, hamt_node_collision_dealloc)
- if (len > 0) {
- while (--len >= 0) {
- Py_XDECREF(self->c_array[len]);
- }
- }
- Py_TYPE(self)->tp_free((PyObject *)self);
- Py_TRASHCAN_END
- }
- #ifdef Py_DEBUG
- static int
- hamt_node_collision_dump(PyHamtNode_Collision *node,
- _PyUnicodeWriter *writer, int level)
- {
- /* Debug build: __dump__() method implementation for Collision nodes. */
- Py_ssize_t i;
- if (_hamt_dump_ident(writer, level + 1)) {
- goto error;
- }
- if (_hamt_dump_format(writer, "CollisionNode(size=%zd id=%p):\n",
- Py_SIZE(node), node))
- {
- goto error;
- }
- for (i = 0; i < Py_SIZE(node); i += 2) {
- PyObject *key = node->c_array[i];
- PyObject *val = node->c_array[i + 1];
- if (_hamt_dump_ident(writer, level + 2)) {
- goto error;
- }
- if (_hamt_dump_format(writer, "%R: %R\n", key, val)) {
- goto error;
- }
- }
- return 0;
- error:
- return -1;
- }
- #endif /* Py_DEBUG */
- /////////////////////////////////// Array Node
- static PyHamtNode *
- hamt_node_array_new(Py_ssize_t count)
- {
- Py_ssize_t i;
- PyHamtNode_Array *node = PyObject_GC_New(
- PyHamtNode_Array, &_PyHamt_ArrayNode_Type);
- if (node == NULL) {
- return NULL;
- }
- for (i = 0; i < HAMT_ARRAY_NODE_SIZE; i++) {
- node->a_array[i] = NULL;
- }
- node->a_count = count;
- _PyObject_GC_TRACK(node);
- return (PyHamtNode *)node;
- }
- static PyHamtNode_Array *
- hamt_node_array_clone(PyHamtNode_Array *node)
- {
- PyHamtNode_Array *clone;
- Py_ssize_t i;
- VALIDATE_ARRAY_NODE(node)
- /* Create a new Array node. */
- clone = (PyHamtNode_Array *)hamt_node_array_new(node->a_count);
- if (clone == NULL) {
- return NULL;
- }
- /* Copy all elements from the current Array node to the new one. */
- for (i = 0; i < HAMT_ARRAY_NODE_SIZE; i++) {
- clone->a_array[i] = (PyHamtNode*)Py_XNewRef(node->a_array[i]);
- }
- VALIDATE_ARRAY_NODE(clone)
- return clone;
- }
- static PyHamtNode *
- hamt_node_array_assoc(PyHamtNode_Array *self,
- uint32_t shift, int32_t hash,
- PyObject *key, PyObject *val, int* added_leaf)
- {
- /* Set a new key to this level (currently a Collision node)
- of the tree.
- Array nodes don't store values, they can only point to
- other nodes. They are simple arrays of 32 BaseNode pointers/
- */
- uint32_t idx = hamt_mask(hash, shift);
- PyHamtNode *node = self->a_array[idx];
- PyHamtNode *child_node;
- PyHamtNode_Array *new_node;
- Py_ssize_t i;
- if (node == NULL) {
- /* There's no child node for the given hash. Create a new
- Bitmap node for this key. */
- PyHamtNode_Bitmap *empty = NULL;
- /* Get an empty Bitmap node to work with. */
- empty = (PyHamtNode_Bitmap *)hamt_node_bitmap_new(0);
- if (empty == NULL) {
- return NULL;
- }
- /* Set key/val to the newly created empty Bitmap, thus
- creating a new Bitmap node with our key/value pair. */
- child_node = hamt_node_bitmap_assoc(
- empty,
- shift + 5, hash, key, val, added_leaf);
- Py_DECREF(empty);
- if (child_node == NULL) {
- return NULL;
- }
- /* Create a new Array node. */
- new_node = (PyHamtNode_Array *)hamt_node_array_new(self->a_count + 1);
- if (new_node == NULL) {
- Py_DECREF(child_node);
- return NULL;
- }
- /* Copy all elements from the current Array node to the
- new one. */
- for (i = 0; i < HAMT_ARRAY_NODE_SIZE; i++) {
- new_node->a_array[i] = (PyHamtNode*)Py_XNewRef(self->a_array[i]);
- }
- assert(new_node->a_array[idx] == NULL);
- new_node->a_array[idx] = child_node; /* borrow */
- VALIDATE_ARRAY_NODE(new_node)
- }
- else {
- /* There's a child node for the given hash.
- Set the key to it./ */
- child_node = hamt_node_assoc(
- node, shift + 5, hash, key, val, added_leaf);
- if (child_node == NULL) {
- return NULL;
- }
- else if (child_node == (PyHamtNode *)self) {
- Py_DECREF(child_node);
- return (PyHamtNode *)self;
- }
- new_node = hamt_node_array_clone(self);
- if (new_node == NULL) {
- Py_DECREF(child_node);
- return NULL;
- }
- Py_SETREF(new_node->a_array[idx], child_node); /* borrow */
- VALIDATE_ARRAY_NODE(new_node)
- }
- return (PyHamtNode *)new_node;
- }
- static hamt_without_t
- hamt_node_array_without(PyHamtNode_Array *self,
- uint32_t shift, int32_t hash,
- PyObject *key,
- PyHamtNode **new_node)
- {
- uint32_t idx = hamt_mask(hash, shift);
- PyHamtNode *node = self->a_array[idx];
- if (node == NULL) {
- return W_NOT_FOUND;
- }
- PyHamtNode *sub_node = NULL;
- hamt_without_t res = hamt_node_without(
- (PyHamtNode *)node,
- shift + 5, hash, key, &sub_node);
- switch (res) {
- case W_NOT_FOUND:
- case W_ERROR:
- assert(sub_node == NULL);
- return res;
- case W_NEWNODE: {
- /* We need to replace a node at the `idx` index.
- Clone this node and replace.
- */
- assert(sub_node != NULL);
- PyHamtNode_Array *clone = hamt_node_array_clone(self);
- if (clone == NULL) {
- Py_DECREF(sub_node);
- return W_ERROR;
- }
- Py_SETREF(clone->a_array[idx], sub_node); /* borrow */
- *new_node = (PyHamtNode*)clone; /* borrow */
- return W_NEWNODE;
- }
- case W_EMPTY: {
- assert(sub_node == NULL);
- /* We need to remove a node at the `idx` index.
- Calculate the size of the replacement Array node.
- */
- Py_ssize_t new_count = self->a_count - 1;
- if (new_count == 0) {
- return W_EMPTY;
- }
- if (new_count >= 16) {
- /* We convert Bitmap nodes to Array nodes, when a
- Bitmap node needs to store more than 15 key/value
- pairs. So we will create a new Array node if we
- the number of key/values after deletion is still
- greater than 15.
- */
- PyHamtNode_Array *new = hamt_node_array_clone(self);
- if (new == NULL) {
- return W_ERROR;
- }
- new->a_count = new_count;
- Py_CLEAR(new->a_array[idx]);
- *new_node = (PyHamtNode*)new; /* borrow */
- return W_NEWNODE;
- }
- /* New Array node would have less than 16 key/value
- pairs. We need to create a replacement Bitmap node. */
- Py_ssize_t bitmap_size = new_count * 2;
- uint32_t bitmap = 0;
- PyHamtNode_Bitmap *new = (PyHamtNode_Bitmap *)
- hamt_node_bitmap_new(bitmap_size);
- if (new == NULL) {
- return W_ERROR;
- }
- Py_ssize_t new_i = 0;
- for (uint32_t i = 0; i < HAMT_ARRAY_NODE_SIZE; i++) {
- if (i == idx) {
- /* Skip the node we are deleting. */
- continue;
- }
- PyHamtNode *node = self->a_array[i];
- if (node == NULL) {
- /* Skip any missing nodes. */
- continue;
- }
- bitmap |= 1U << i;
- if (IS_BITMAP_NODE(node)) {
- PyHamtNode_Bitmap *child = (PyHamtNode_Bitmap *)node;
- if (hamt_node_bitmap_count(child) == 1 &&
- child->b_array[0] != NULL)
- {
- /* node is a Bitmap with one key/value pair, just
- merge it into the new Bitmap node we're building.
- Note that we don't inline Bitmap nodes that
- have a NULL key -- those nodes point to another
- tree level, and we cannot simply move tree levels
- up or down.
- */
- PyObject *key = child->b_array[0];
- PyObject *val = child->b_array[1];
- new->b_array[new_i] = Py_NewRef(key);
- new->b_array[new_i + 1] = Py_NewRef(val);
- }
- else {
- new->b_array[new_i] = NULL;
- new->b_array[new_i + 1] = Py_NewRef(node);
- }
- }
- else {
- #ifdef Py_DEBUG
- if (IS_COLLISION_NODE(node)) {
- Py_ssize_t child_count = hamt_node_collision_count(
- (PyHamtNode_Collision*)node);
- assert(child_count > 1);
- }
- else if (IS_ARRAY_NODE(node)) {
- assert(((PyHamtNode_Array*)node)->a_count >= 16);
- }
- #endif
- /* Just copy the node into our new Bitmap */
- new->b_array[new_i] = NULL;
- new->b_array[new_i + 1] = Py_NewRef(node);
- }
- new_i += 2;
- }
- new->b_bitmap = bitmap;
- *new_node = (PyHamtNode*)new; /* borrow */
- return W_NEWNODE;
- }
- default:
- Py_UNREACHABLE();
- }
- }
- static hamt_find_t
- hamt_node_array_find(PyHamtNode_Array *self,
- uint32_t shift, int32_t hash,
- PyObject *key, PyObject **val)
- {
- /* Lookup `key` in the Array node `self`. Set the value
- for the found key to 'val'. */
- uint32_t idx = hamt_mask(hash, shift);
- PyHamtNode *node;
- node = self->a_array[idx];
- if (node == NULL) {
- return F_NOT_FOUND;
- }
- /* Dispatch to the generic hamt_node_find */
- return hamt_node_find(node, shift + 5, hash, key, val);
- }
- static int
- hamt_node_array_traverse(PyHamtNode_Array *self,
- visitproc visit, void *arg)
- {
- /* Array's tp_traverse */
- Py_ssize_t i;
- for (i = 0; i < HAMT_ARRAY_NODE_SIZE; i++) {
- Py_VISIT(self->a_array[i]);
- }
- return 0;
- }
- static void
- hamt_node_array_dealloc(PyHamtNode_Array *self)
- {
- /* Array's tp_dealloc */
- Py_ssize_t i;
- PyObject_GC_UnTrack(self);
- Py_TRASHCAN_BEGIN(self, hamt_node_array_dealloc)
- for (i = 0; i < HAMT_ARRAY_NODE_SIZE; i++) {
- Py_XDECREF(self->a_array[i]);
- }
- Py_TYPE(self)->tp_free((PyObject *)self);
- Py_TRASHCAN_END
- }
- #ifdef Py_DEBUG
- static int
- hamt_node_array_dump(PyHamtNode_Array *node,
- _PyUnicodeWriter *writer, int level)
- {
- /* Debug build: __dump__() method implementation for Array nodes. */
- Py_ssize_t i;
- if (_hamt_dump_ident(writer, level + 1)) {
- goto error;
- }
- if (_hamt_dump_format(writer, "ArrayNode(id=%p):\n", node)) {
- goto error;
- }
- for (i = 0; i < HAMT_ARRAY_NODE_SIZE; i++) {
- if (node->a_array[i] == NULL) {
- continue;
- }
- if (_hamt_dump_ident(writer, level + 2)) {
- goto error;
- }
- if (_hamt_dump_format(writer, "%zd::\n", i)) {
- goto error;
- }
- if (hamt_node_dump(node->a_array[i], writer, level + 1)) {
- goto error;
- }
- if (_hamt_dump_format(writer, "\n")) {
- goto error;
- }
- }
- return 0;
- error:
- return -1;
- }
- #endif /* Py_DEBUG */
- /////////////////////////////////// Node Dispatch
- static PyHamtNode *
- hamt_node_assoc(PyHamtNode *node,
- uint32_t shift, int32_t hash,
- PyObject *key, PyObject *val, int* added_leaf)
- {
- /* Set key/value to the 'node' starting with the given shift/hash.
- Return a new node, or the same node if key/value already
- set.
- added_leaf will be set to 1 if key/value wasn't in the
- tree before.
- This method automatically dispatches to the suitable
- hamt_node_{nodetype}_assoc method.
- */
- if (IS_BITMAP_NODE(node)) {
- return hamt_node_bitmap_assoc(
- (PyHamtNode_Bitmap *)node,
- shift, hash, key, val, added_leaf);
- }
- else if (IS_ARRAY_NODE(node)) {
- return hamt_node_array_assoc(
- (PyHamtNode_Array *)node,
- shift, hash, key, val, added_leaf);
- }
- else {
- assert(IS_COLLISION_NODE(node));
- return hamt_node_collision_assoc(
- (PyHamtNode_Collision *)node,
- shift, hash, key, val, added_leaf);
- }
- }
- static hamt_without_t
- hamt_node_without(PyHamtNode *node,
- uint32_t shift, int32_t hash,
- PyObject *key,
- PyHamtNode **new_node)
- {
- if (IS_BITMAP_NODE(node)) {
- return hamt_node_bitmap_without(
- (PyHamtNode_Bitmap *)node,
- shift, hash, key,
- new_node);
- }
- else if (IS_ARRAY_NODE(node)) {
- return hamt_node_array_without(
- (PyHamtNode_Array *)node,
- shift, hash, key,
- new_node);
- }
- else {
- assert(IS_COLLISION_NODE(node));
- return hamt_node_collision_without(
- (PyHamtNode_Collision *)node,
- shift, hash, key,
- new_node);
- }
- }
- static hamt_find_t
- hamt_node_find(PyHamtNode *node,
- uint32_t shift, int32_t hash,
- PyObject *key, PyObject **val)
- {
- /* Find the key in the node starting with the given shift/hash.
- If a value is found, the result will be set to F_FOUND, and
- *val will point to the found value object.
- If a value wasn't found, the result will be set to F_NOT_FOUND.
- If an exception occurs during the call, the result will be F_ERROR.
- This method automatically dispatches to the suitable
- hamt_node_{nodetype}_find method.
- */
- if (IS_BITMAP_NODE(node)) {
- return hamt_node_bitmap_find(
- (PyHamtNode_Bitmap *)node,
- shift, hash, key, val);
- }
- else if (IS_ARRAY_NODE(node)) {
- return hamt_node_array_find(
- (PyHamtNode_Array *)node,
- shift, hash, key, val);
- }
- else {
- assert(IS_COLLISION_NODE(node));
- return hamt_node_collision_find(
- (PyHamtNode_Collision *)node,
- shift, hash, key, val);
- }
- }
- #ifdef Py_DEBUG
- static int
- hamt_node_dump(PyHamtNode *node,
- _PyUnicodeWriter *writer, int level)
- {
- /* Debug build: __dump__() method implementation for a node.
- This method automatically dispatches to the suitable
- hamt_node_{nodetype})_dump method.
- */
- if (IS_BITMAP_NODE(node)) {
- return hamt_node_bitmap_dump(
- (PyHamtNode_Bitmap *)node, writer, level);
- }
- else if (IS_ARRAY_NODE(node)) {
- return hamt_node_array_dump(
- (PyHamtNode_Array *)node, writer, level);
- }
- else {
- assert(IS_COLLISION_NODE(node));
- return hamt_node_collision_dump(
- (PyHamtNode_Collision *)node, writer, level);
- }
- }
- #endif /* Py_DEBUG */
- /////////////////////////////////// Iterators: Machinery
- static hamt_iter_t
- hamt_iterator_next(PyHamtIteratorState *iter, PyObject **key, PyObject **val);
- static void
- hamt_iterator_init(PyHamtIteratorState *iter, PyHamtNode *root)
- {
- for (uint32_t i = 0; i < _Py_HAMT_MAX_TREE_DEPTH; i++) {
- iter->i_nodes[i] = NULL;
- iter->i_pos[i] = 0;
- }
- iter->i_level = 0;
- /* Note: we don't incref/decref nodes in i_nodes. */
- iter->i_nodes[0] = root;
- }
- static hamt_iter_t
- hamt_iterator_bitmap_next(PyHamtIteratorState *iter,
- PyObject **key, PyObject **val)
- {
- int8_t level = iter->i_level;
- PyHamtNode_Bitmap *node = (PyHamtNode_Bitmap *)(iter->i_nodes[level]);
- Py_ssize_t pos = iter->i_pos[level];
- if (pos + 1 >= Py_SIZE(node)) {
- #ifdef Py_DEBUG
- assert(iter->i_level >= 0);
- iter->i_nodes[iter->i_level] = NULL;
- #endif
- iter->i_level--;
- return hamt_iterator_next(iter, key, val);
- }
- if (node->b_array[pos] == NULL) {
- iter->i_pos[level] = pos + 2;
- int8_t next_level = level + 1;
- assert(next_level < _Py_HAMT_MAX_TREE_DEPTH);
- iter->i_level = next_level;
- iter->i_pos[next_level] = 0;
- iter->i_nodes[next_level] = (PyHamtNode *)
- node->b_array[pos + 1];
- return hamt_iterator_next(iter, key, val);
- }
- *key = node->b_array[pos];
- *val = node->b_array[pos + 1];
- iter->i_pos[level] = pos + 2;
- return I_ITEM;
- }
- static hamt_iter_t
- hamt_iterator_collision_next(PyHamtIteratorState *iter,
- PyObject **key, PyObject **val)
- {
- int8_t level = iter->i_level;
- PyHamtNode_Collision *node = (PyHamtNode_Collision *)(iter->i_nodes[level]);
- Py_ssize_t pos = iter->i_pos[level];
- if (pos + 1 >= Py_SIZE(node)) {
- #ifdef Py_DEBUG
- assert(iter->i_level >= 0);
- iter->i_nodes[iter->i_level] = NULL;
- #endif
- iter->i_level--;
- return hamt_iterator_next(iter, key, val);
- }
- *key = node->c_array[pos];
- *val = node->c_array[pos + 1];
- iter->i_pos[level] = pos + 2;
- return I_ITEM;
- }
- static hamt_iter_t
- hamt_iterator_array_next(PyHamtIteratorState *iter,
- PyObject **key, PyObject **val)
- {
- int8_t level = iter->i_level;
- PyHamtNode_Array *node = (PyHamtNode_Array *)(iter->i_nodes[level]);
- Py_ssize_t pos = iter->i_pos[level];
- if (pos >= HAMT_ARRAY_NODE_SIZE) {
- #ifdef Py_DEBUG
- assert(iter->i_level >= 0);
- iter->i_nodes[iter->i_level] = NULL;
- #endif
- iter->i_level--;
- return hamt_iterator_next(iter, key, val);
- }
- for (Py_ssize_t i = pos; i < HAMT_ARRAY_NODE_SIZE; i++) {
- if (node->a_array[i] != NULL) {
- iter->i_pos[level] = i + 1;
- int8_t next_level = level + 1;
- assert(next_level < _Py_HAMT_MAX_TREE_DEPTH);
- iter->i_pos[next_level] = 0;
- iter->i_nodes[next_level] = node->a_array[i];
- iter->i_level = next_level;
- return hamt_iterator_next(iter, key, val);
- }
- }
- #ifdef Py_DEBUG
- assert(iter->i_level >= 0);
- iter->i_nodes[iter->i_level] = NULL;
- #endif
- iter->i_level--;
- return hamt_iterator_next(iter, key, val);
- }
- static hamt_iter_t
- hamt_iterator_next(PyHamtIteratorState *iter, PyObject **key, PyObject **val)
- {
- if (iter->i_level < 0) {
- return I_END;
- }
- assert(iter->i_level < _Py_HAMT_MAX_TREE_DEPTH);
- PyHamtNode *current = iter->i_nodes[iter->i_level];
- if (IS_BITMAP_NODE(current)) {
- return hamt_iterator_bitmap_next(iter, key, val);
- }
- else if (IS_ARRAY_NODE(current)) {
- return hamt_iterator_array_next(iter, key, val);
- }
- else {
- assert(IS_COLLISION_NODE(current));
- return hamt_iterator_collision_next(iter, key, val);
- }
- }
- /////////////////////////////////// HAMT high-level functions
- PyHamtObject *
- _PyHamt_Assoc(PyHamtObject *o, PyObject *key, PyObject *val)
- {
- int32_t key_hash;
- int added_leaf = 0;
- PyHamtNode *new_root;
- PyHamtObject *new_o;
- key_hash = hamt_hash(key);
- if (key_hash == -1) {
- return NULL;
- }
- new_root = hamt_node_assoc(
- (PyHamtNode *)(o->h_root),
- 0, key_hash, key, val, &added_leaf);
- if (new_root == NULL) {
- return NULL;
- }
- if (new_root == o->h_root) {
- Py_DECREF(new_root);
- return (PyHamtObject*)Py_NewRef(o);
- }
- new_o = hamt_alloc();
- if (new_o == NULL) {
- Py_DECREF(new_root);
- return NULL;
- }
- new_o->h_root = new_root; /* borrow */
- new_o->h_count = added_leaf ? o->h_count + 1 : o->h_count;
- return new_o;
- }
- PyHamtObject *
- _PyHamt_Without(PyHamtObject *o, PyObject *key)
- {
- int32_t key_hash = hamt_hash(key);
- if (key_hash == -1) {
- return NULL;
- }
- PyHamtNode *new_root = NULL;
- hamt_without_t res = hamt_node_without(
- (PyHamtNode *)(o->h_root),
- 0, key_hash, key,
- &new_root);
- switch (res) {
- case W_ERROR:
- return NULL;
- case W_EMPTY:
- return _PyHamt_New();
- case W_NOT_FOUND:
- return (PyHamtObject*)Py_NewRef(o);
- case W_NEWNODE: {
- assert(new_root != NULL);
- PyHamtObject *new_o = hamt_alloc();
- if (new_o == NULL) {
- Py_DECREF(new_root);
- return NULL;
- }
- new_o->h_root = new_root; /* borrow */
- new_o->h_count = o->h_count - 1;
- assert(new_o->h_count >= 0);
- return new_o;
- }
- default:
- Py_UNREACHABLE();
- }
- }
- static hamt_find_t
- hamt_find(PyHamtObject *o, PyObject *key, PyObject **val)
- {
- if (o->h_count == 0) {
- return F_NOT_FOUND;
- }
- int32_t key_hash = hamt_hash(key);
- if (key_hash == -1) {
- return F_ERROR;
- }
- return hamt_node_find(o->h_root, 0, key_hash, key, val);
- }
- int
- _PyHamt_Find(PyHamtObject *o, PyObject *key, PyObject **val)
- {
- hamt_find_t res = hamt_find(o, key, val);
- switch (res) {
- case F_ERROR:
- return -1;
- case F_NOT_FOUND:
- return 0;
- case F_FOUND:
- return 1;
- default:
- Py_UNREACHABLE();
- }
- }
- int
- _PyHamt_Eq(PyHamtObject *v, PyHamtObject *w)
- {
- if (v == w) {
- return 1;
- }
- if (v->h_count != w->h_count) {
- return 0;
- }
- PyHamtIteratorState iter;
- hamt_iter_t iter_res;
- hamt_find_t find_res;
- PyObject *v_key;
- PyObject *v_val;
- PyObject *w_val;
- hamt_iterator_init(&iter, v->h_root);
- do {
- iter_res = hamt_iterator_next(&iter, &v_key, &v_val);
- if (iter_res == I_ITEM) {
- find_res = hamt_find(w, v_key, &w_val);
- switch (find_res) {
- case F_ERROR:
- return -1;
- case F_NOT_FOUND:
- return 0;
- case F_FOUND: {
- int cmp = PyObject_RichCompareBool(v_val, w_val, Py_EQ);
- if (cmp < 0) {
- return -1;
- }
- if (cmp == 0) {
- return 0;
- }
- }
- }
- }
- } while (iter_res != I_END);
- return 1;
- }
- Py_ssize_t
- _PyHamt_Len(PyHamtObject *o)
- {
- return o->h_count;
- }
- static PyHamtObject *
- hamt_alloc(void)
- {
- PyHamtObject *o;
- o = PyObject_GC_New(PyHamtObject, &_PyHamt_Type);
- if (o == NULL) {
- return NULL;
- }
- o->h_count = 0;
- o->h_root = NULL;
- o->h_weakreflist = NULL;
- PyObject_GC_Track(o);
- return o;
- }
- #define _empty_hamt \
- (&_Py_INTERP_SINGLETON(_PyInterpreterState_Get(), hamt_empty))
- PyHamtObject *
- _PyHamt_New(void)
- {
- /* HAMT is an immutable object so we can easily cache an
- empty instance. */
- return (PyHamtObject*)Py_NewRef(_empty_hamt);
- }
- #ifdef Py_DEBUG
- static PyObject *
- hamt_dump(PyHamtObject *self)
- {
- _PyUnicodeWriter writer;
- _PyUnicodeWriter_Init(&writer);
- if (_hamt_dump_format(&writer, "HAMT(len=%zd):\n", self->h_count)) {
- goto error;
- }
- if (hamt_node_dump(self->h_root, &writer, 0)) {
- goto error;
- }
- return _PyUnicodeWriter_Finish(&writer);
- error:
- _PyUnicodeWriter_Dealloc(&writer);
- return NULL;
- }
- #endif /* Py_DEBUG */
- /////////////////////////////////// Iterators: Shared Iterator Implementation
- static int
- hamt_baseiter_tp_clear(PyHamtIterator *it)
- {
- Py_CLEAR(it->hi_obj);
- return 0;
- }
- static void
- hamt_baseiter_tp_dealloc(PyHamtIterator *it)
- {
- PyObject_GC_UnTrack(it);
- (void)hamt_baseiter_tp_clear(it);
- PyObject_GC_Del(it);
- }
- static int
- hamt_baseiter_tp_traverse(PyHamtIterator *it, visitproc visit, void *arg)
- {
- Py_VISIT(it->hi_obj);
- return 0;
- }
- static PyObject *
- hamt_baseiter_tp_iternext(PyHamtIterator *it)
- {
- PyObject *key;
- PyObject *val;
- hamt_iter_t res = hamt_iterator_next(&it->hi_iter, &key, &val);
- switch (res) {
- case I_END:
- PyErr_SetNone(PyExc_StopIteration);
- return NULL;
- case I_ITEM: {
- return (*(it->hi_yield))(key, val);
- }
- default: {
- Py_UNREACHABLE();
- }
- }
- }
- static Py_ssize_t
- hamt_baseiter_tp_len(PyHamtIterator *it)
- {
- return it->hi_obj->h_count;
- }
- static PyMappingMethods PyHamtIterator_as_mapping = {
- (lenfunc)hamt_baseiter_tp_len,
- };
- static PyObject *
- hamt_baseiter_new(PyTypeObject *type, binaryfunc yield, PyHamtObject *o)
- {
- PyHamtIterator *it = PyObject_GC_New(PyHamtIterator, type);
- if (it == NULL) {
- return NULL;
- }
- it->hi_obj = (PyHamtObject*)Py_NewRef(o);
- it->hi_yield = yield;
- hamt_iterator_init(&it->hi_iter, o->h_root);
- return (PyObject*)it;
- }
- #define ITERATOR_TYPE_SHARED_SLOTS \
- .tp_basicsize = sizeof(PyHamtIterator), \
- .tp_itemsize = 0, \
- .tp_as_mapping = &PyHamtIterator_as_mapping, \
- .tp_dealloc = (destructor)hamt_baseiter_tp_dealloc, \
- .tp_getattro = PyObject_GenericGetAttr, \
- .tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC, \
- .tp_traverse = (traverseproc)hamt_baseiter_tp_traverse, \
- .tp_clear = (inquiry)hamt_baseiter_tp_clear, \
- .tp_iter = PyObject_SelfIter, \
- .tp_iternext = (iternextfunc)hamt_baseiter_tp_iternext,
- /////////////////////////////////// _PyHamtItems_Type
- PyTypeObject _PyHamtItems_Type = {
- PyVarObject_HEAD_INIT(NULL, 0)
- "items",
- ITERATOR_TYPE_SHARED_SLOTS
- };
- static PyObject *
- hamt_iter_yield_items(PyObject *key, PyObject *val)
- {
- return PyTuple_Pack(2, key, val);
- }
- PyObject *
- _PyHamt_NewIterItems(PyHamtObject *o)
- {
- return hamt_baseiter_new(
- &_PyHamtItems_Type, hamt_iter_yield_items, o);
- }
- /////////////////////////////////// _PyHamtKeys_Type
- PyTypeObject _PyHamtKeys_Type = {
- PyVarObject_HEAD_INIT(NULL, 0)
- "keys",
- ITERATOR_TYPE_SHARED_SLOTS
- };
- static PyObject *
- hamt_iter_yield_keys(PyObject *key, PyObject *val)
- {
- return Py_NewRef(key);
- }
- PyObject *
- _PyHamt_NewIterKeys(PyHamtObject *o)
- {
- return hamt_baseiter_new(
- &_PyHamtKeys_Type, hamt_iter_yield_keys, o);
- }
- /////////////////////////////////// _PyHamtValues_Type
- PyTypeObject _PyHamtValues_Type = {
- PyVarObject_HEAD_INIT(NULL, 0)
- "values",
- ITERATOR_TYPE_SHARED_SLOTS
- };
- static PyObject *
- hamt_iter_yield_values(PyObject *key, PyObject *val)
- {
- return Py_NewRef(val);
- }
- PyObject *
- _PyHamt_NewIterValues(PyHamtObject *o)
- {
- return hamt_baseiter_new(
- &_PyHamtValues_Type, hamt_iter_yield_values, o);
- }
- /////////////////////////////////// _PyHamt_Type
- #ifdef Py_DEBUG
- static PyObject *
- hamt_dump(PyHamtObject *self);
- #endif
- static PyObject *
- hamt_tp_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
- {
- return (PyObject*)_PyHamt_New();
- }
- static int
- hamt_tp_clear(PyHamtObject *self)
- {
- Py_CLEAR(self->h_root);
- return 0;
- }
- static int
- hamt_tp_traverse(PyHamtObject *self, visitproc visit, void *arg)
- {
- Py_VISIT(self->h_root);
- return 0;
- }
- static void
- hamt_tp_dealloc(PyHamtObject *self)
- {
- if (self == _empty_hamt) {
- /* The empty one is statically allocated. */
- #ifdef Py_DEBUG
- _Py_FatalRefcountError("deallocating the empty hamt singleton");
- #else
- return;
- #endif
- }
- PyObject_GC_UnTrack(self);
- if (self->h_weakreflist != NULL) {
- PyObject_ClearWeakRefs((PyObject*)self);
- }
- (void)hamt_tp_clear(self);
- Py_TYPE(self)->tp_free(self);
- }
- static PyObject *
- hamt_tp_richcompare(PyObject *v, PyObject *w, int op)
- {
- if (!PyHamt_Check(v) || !PyHamt_Check(w) || (op != Py_EQ && op != Py_NE)) {
- Py_RETURN_NOTIMPLEMENTED;
- }
- int res = _PyHamt_Eq((PyHamtObject *)v, (PyHamtObject *)w);
- if (res < 0) {
- return NULL;
- }
- if (op == Py_NE) {
- res = !res;
- }
- if (res) {
- Py_RETURN_TRUE;
- }
- else {
- Py_RETURN_FALSE;
- }
- }
- static int
- hamt_tp_contains(PyHamtObject *self, PyObject *key)
- {
- PyObject *val;
- return _PyHamt_Find(self, key, &val);
- }
- static PyObject *
- hamt_tp_subscript(PyHamtObject *self, PyObject *key)
- {
- PyObject *val;
- hamt_find_t res = hamt_find(self, key, &val);
- switch (res) {
- case F_ERROR:
- return NULL;
- case F_FOUND:
- return Py_NewRef(val);
- case F_NOT_FOUND:
- PyErr_SetObject(PyExc_KeyError, key);
- return NULL;
- default:
- Py_UNREACHABLE();
- }
- }
- static Py_ssize_t
- hamt_tp_len(PyHamtObject *self)
- {
- return _PyHamt_Len(self);
- }
- static PyObject *
- hamt_tp_iter(PyHamtObject *self)
- {
- return _PyHamt_NewIterKeys(self);
- }
- static PyObject *
- hamt_py_set(PyHamtObject *self, PyObject *args)
- {
- PyObject *key;
- PyObject *val;
- if (!PyArg_UnpackTuple(args, "set", 2, 2, &key, &val)) {
- return NULL;
- }
- return (PyObject *)_PyHamt_Assoc(self, key, val);
- }
- static PyObject *
- hamt_py_get(PyHamtObject *self, PyObject *args)
- {
- PyObject *key;
- PyObject *def = NULL;
- if (!PyArg_UnpackTuple(args, "get", 1, 2, &key, &def)) {
- return NULL;
- }
- PyObject *val = NULL;
- hamt_find_t res = hamt_find(self, key, &val);
- switch (res) {
- case F_ERROR:
- return NULL;
- case F_FOUND:
- return Py_NewRef(val);
- case F_NOT_FOUND:
- if (def == NULL) {
- Py_RETURN_NONE;
- }
- return Py_NewRef(def);
- default:
- Py_UNREACHABLE();
- }
- }
- static PyObject *
- hamt_py_delete(PyHamtObject *self, PyObject *key)
- {
- return (PyObject *)_PyHamt_Without(self, key);
- }
- static PyObject *
- hamt_py_items(PyHamtObject *self, PyObject *args)
- {
- return _PyHamt_NewIterItems(self);
- }
- static PyObject *
- hamt_py_values(PyHamtObject *self, PyObject *args)
- {
- return _PyHamt_NewIterValues(self);
- }
- static PyObject *
- hamt_py_keys(PyHamtObject *self, PyObject *Py_UNUSED(args))
- {
- return _PyHamt_NewIterKeys(self);
- }
- #ifdef Py_DEBUG
- static PyObject *
- hamt_py_dump(PyHamtObject *self, PyObject *Py_UNUSED(args))
- {
- return hamt_dump(self);
- }
- #endif
- static PyMethodDef PyHamt_methods[] = {
- {"set", _PyCFunction_CAST(hamt_py_set), METH_VARARGS, NULL},
- {"get", _PyCFunction_CAST(hamt_py_get), METH_VARARGS, NULL},
- {"delete", _PyCFunction_CAST(hamt_py_delete), METH_O, NULL},
- {"items", _PyCFunction_CAST(hamt_py_items), METH_NOARGS, NULL},
- {"keys", _PyCFunction_CAST(hamt_py_keys), METH_NOARGS, NULL},
- {"values", _PyCFunction_CAST(hamt_py_values), METH_NOARGS, NULL},
- #ifdef Py_DEBUG
- {"__dump__", _PyCFunction_CAST(hamt_py_dump), METH_NOARGS, NULL},
- #endif
- {NULL, NULL}
- };
- static PySequenceMethods PyHamt_as_sequence = {
- 0, /* sq_length */
- 0, /* sq_concat */
- 0, /* sq_repeat */
- 0, /* sq_item */
- 0, /* sq_slice */
- 0, /* sq_ass_item */
- 0, /* sq_ass_slice */
- (objobjproc)hamt_tp_contains, /* sq_contains */
- 0, /* sq_inplace_concat */
- 0, /* sq_inplace_repeat */
- };
- static PyMappingMethods PyHamt_as_mapping = {
- (lenfunc)hamt_tp_len, /* mp_length */
- (binaryfunc)hamt_tp_subscript, /* mp_subscript */
- };
- PyTypeObject _PyHamt_Type = {
- PyVarObject_HEAD_INIT(&PyType_Type, 0)
- "hamt",
- sizeof(PyHamtObject),
- .tp_methods = PyHamt_methods,
- .tp_as_mapping = &PyHamt_as_mapping,
- .tp_as_sequence = &PyHamt_as_sequence,
- .tp_iter = (getiterfunc)hamt_tp_iter,
- .tp_dealloc = (destructor)hamt_tp_dealloc,
- .tp_getattro = PyObject_GenericGetAttr,
- .tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC,
- .tp_richcompare = hamt_tp_richcompare,
- .tp_traverse = (traverseproc)hamt_tp_traverse,
- .tp_clear = (inquiry)hamt_tp_clear,
- .tp_new = hamt_tp_new,
- .tp_weaklistoffset = offsetof(PyHamtObject, h_weakreflist),
- .tp_hash = PyObject_HashNotImplemented,
- };
- /////////////////////////////////// Tree Node Types
- PyTypeObject _PyHamt_ArrayNode_Type = {
- PyVarObject_HEAD_INIT(&PyType_Type, 0)
- "hamt_array_node",
- sizeof(PyHamtNode_Array),
- 0,
- .tp_dealloc = (destructor)hamt_node_array_dealloc,
- .tp_getattro = PyObject_GenericGetAttr,
- .tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC,
- .tp_traverse = (traverseproc)hamt_node_array_traverse,
- .tp_free = PyObject_GC_Del,
- .tp_hash = PyObject_HashNotImplemented,
- };
- PyTypeObject _PyHamt_BitmapNode_Type = {
- PyVarObject_HEAD_INIT(&PyType_Type, 0)
- "hamt_bitmap_node",
- sizeof(PyHamtNode_Bitmap) - sizeof(PyObject *),
- sizeof(PyObject *),
- .tp_dealloc = (destructor)hamt_node_bitmap_dealloc,
- .tp_getattro = PyObject_GenericGetAttr,
- .tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC,
- .tp_traverse = (traverseproc)hamt_node_bitmap_traverse,
- .tp_free = PyObject_GC_Del,
- .tp_hash = PyObject_HashNotImplemented,
- };
- PyTypeObject _PyHamt_CollisionNode_Type = {
- PyVarObject_HEAD_INIT(&PyType_Type, 0)
- "hamt_collision_node",
- sizeof(PyHamtNode_Collision) - sizeof(PyObject *),
- sizeof(PyObject *),
- .tp_dealloc = (destructor)hamt_node_collision_dealloc,
- .tp_getattro = PyObject_GenericGetAttr,
- .tp_flags = Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC,
- .tp_traverse = (traverseproc)hamt_node_collision_traverse,
- .tp_free = PyObject_GC_Del,
- .tp_hash = PyObject_HashNotImplemented,
- };
|