X-Git-Url: http://deadsoftware.ru/gitweb?a=blobdiff_plain;ds=sidebyside;f=src%2Fshared%2Fhashtable.pas;h=85133c3efbc10d47581f03f132ed6bda137d36b7;hb=2d5c517f0d38a9f31e8e780416269ea787852930;hp=10b67d97f722af171e2882bb4a0be5ae27745888;hpb=9a8ac2323ba1c23f248ff5effe2747135ecc14f6;p=d2df-sdl.git diff --git a/src/shared/hashtable.pas b/src/shared/hashtable.pas index 10b67d9..85133c3 100644 --- a/src/shared/hashtable.pas +++ b/src/shared/hashtable.pas @@ -19,7 +19,7 @@ {.$DEFINE RBHASH_DEBUG_DELETE} {.$DEFINE RBHASH_DEBUG_COMPACT} {$IF DEFINED(D2F_DEBUG)} - {$DEFINE RBHASH_SANITY_CHECKS} + {.$DEFINE RBHASH_SANITY_CHECKS} {$ENDIF} // hash table (robin hood) unit hashtable; @@ -31,30 +31,83 @@ type // WARNING! don't put structures into hash, use ponters or ids! generic THashBase = class(TObject) private - const InitSize = {$IF DEFINED(D2F_DEBUG)}16{$ELSE}512{$ENDIF}; // *MUST* be power of two + const InitSize = {$IF DEFINED(RBHASH_SANITY_CHECKS)}16{$ELSE}256{$ENDIF}; // *MUST* be power of two const LoadFactorPrc = 90; // it is ok for robin hood hashes public - type THashFn = function (constref o: KeyT): LongWord; - type TEquFn = function (constref a, b: KeyT): Boolean; - type TIteratorFn = function (constref k: KeyT; constref v: ValueT): Boolean is nested; // return `true` to stop - - private type PEntry = ^TEntry; TEntry = record + public key: KeyT; value: ValueT; + private hash: LongWord; // key hash or 0 nextFree: PEntry; // next free entry + private + function getEmpty (): Boolean; inline; + public + property empty: Boolean read getEmpty; + property keyhash: LongWord read hash; // cannot be 0 + end; + + type THashFn = function (constref o: KeyT): LongWord; + type TEquFn = function (constref a, b: KeyT): Boolean; + type TFreeKeyFn = procedure (var k: KeyT); // this may free key + type TFreeValueFn = procedure (var v: ValueT); // this may free value + type TIteratorFn = function (constref k: KeyT; constref v: ValueT): Boolean is nested; // return `true` to stop + type TIteratorExFn = function (constref k: KeyT; constref v: ValueT; keyhash: LongWord): Boolean is nested; // return `true` to stop + + private + type + TEntryArray = array of TEntry; + + public + type + TValEnumerator = record + private + mEntries: TEntryArray; + mFirstEntry, mLastEntry, cur: Integer; + public + constructor Create (const aents: TEntryArray; afirst, alast: Integer); + function MoveNext (): Boolean; inline; + function getCurrent (): ValueT; inline; + function GetEnumerator (): TValEnumerator; inline; + property Current: ValueT read getCurrent; + end; + + TKeyEnumerator = record + private + mEntries: TEntryArray; + mFirstEntry, mLastEntry, cur: Integer; + public + constructor Create (const aents: TEntryArray; afirst, alast: Integer); + function MoveNext (): Boolean; inline; + function getCurrent (): KeyT; inline; + function GetEnumerator (): TKeyEnumerator; inline; + property Current: KeyT read getCurrent; + end; + + TKeyValEnumerator = record + private + mEntries: TEntryArray; + mFirstEntry, mLastEntry, cur: Integer; + public + constructor Create (const aents: TEntryArray; afirst, alast: Integer); + function MoveNext (): Boolean; inline; + function getCurrent (): PEntry; inline; + function GetEnumerator (): TKeyValEnumerator; inline; + property Current: PEntry read getCurrent; end; private hashfn: THashFn; equfn: TEquFn; + freekeyfn: TFreeKeyFn; + freevalfn: TFreeValueFn; mBuckets: array of PEntry; // entries, points to mEntries elements mBucketsUsed: Integer; - mEntries: array of TEntry; + mEntries: TEntryArray; {$IFDEF RBHASH_SANITY_CHECKS} mEntriesUsed: Integer; {$ENDIF} @@ -66,14 +119,16 @@ type function allocEntry (): PEntry; procedure releaseEntry (e: PEntry); - //function distToStIdx (idx: LongWord): LongWord; inline; + function distToStIdx (idx: LongWord): LongWord; inline; procedure putEntryInternal (swpe: PEntry); function getCapacity (): Integer; inline; + procedure freeEntries (); + public - constructor Create (ahashfn: THashFn; aequfn: TEquFn); + constructor Create (ahashfn: THashFn; aequfn: TEquFn; afreekeyfn: TFreeKeyFn=nil; afreevalfn: TFreeValueFn=nil); destructor Destroy (); override; procedure clear (); @@ -82,19 +137,29 @@ type procedure rehash (); procedure compact (); // call this instead of `rehash()` after alot of deletions - function get (constref akey: KeyT; out rval: ValueT): Boolean; // `true`: found - function put (constref akey: KeyT; constref aval: ValueT): Boolean; // `true`: replaced - function has (constref akey: KeyT): Boolean; // `true`: found - function del (constref akey: KeyT): Boolean; // `true`: deleted + // you may pass `keyhash` to bypass hash calculation + function get (constref akey: KeyT; out rval: ValueT; keyhashin: PLongWord=nil): Boolean; // `true`: found + // the function may return calculated value hash in `keyhash` + function put (constref akey: KeyT; constref aval: ValueT; keyhashout: PLongWord=nil): Boolean; // `true`: replaced + // you may pass `keyhash` to bypass hash calculation + function has (constref akey: KeyT; keyhashin: PLongWord=nil): Boolean; // `true`: found + // you may pass `keyhash` to bypass hash calculation + function del (constref akey: KeyT; keyhashin: PLongWord=nil): Boolean; // `true`: deleted //WARNING! don't modify table in iterator (queries are ok, though) - function forEach (it: TIteratorFn): Boolean; + function forEach (it: TIteratorFn): Boolean; overload; + function forEach (it: TIteratorExFn): Boolean; overload; + + // default `for ... in` enums values + function GetEnumerator (): TValEnumerator; + function byKey (): TKeyEnumerator; + function byValue (): TValEnumerator; + function byKeyValue (): TKeyValEnumerator; // PEntry property count: Integer read mBucketsUsed; property capacity: Integer read getCapacity; end; - type TJoaatHasher = record private @@ -107,7 +172,7 @@ type procedure reset (); inline; overload; procedure reset (aseed: LongWord); inline; overload; - procedure put (const buf; len: LongWord); + procedure put (constref buf; len: LongWord); // current hash value // you can continue putting data, as this is not destructive @@ -117,17 +182,32 @@ type type THashIntInt = specialize THashBase; + THashStrInt = specialize THashBase; + THashIntStr = specialize THashBase; + THashStrStr = specialize THashBase; + function hashNewIntInt (): THashIntInt; +function hashNewStrInt (): THashStrInt; +function hashNewIntStr (): THashIntStr; +function hashNewStrStr (): THashStrStr; function u32Hash (a: LongWord): LongWord; inline; -function fnvHash (const buf; len: LongWord): LongWord; -function joaatHash (const buf; len: LongWord): LongWord; +function fnvHash (constref buf; len: LongWord): LongWord; +function joaatHash (constref buf; len: LongWord): LongWord; function nextPOT (x: LongWord): LongWord; inline; +// for integer keys +function hashIntEqu (constref a, b: Integer): Boolean; +function hashIntHash (constref k: Integer): LongWord; +function hashStrEqu (constref a, b: AnsiString): Boolean; +function hashStrHash (constref k: AnsiString): LongWord; +procedure hashStrFree (var s: AnsiString); + + implementation uses @@ -152,13 +232,15 @@ end; // ////////////////////////////////////////////////////////////////////////// // -function hiiequ (constref a, b: Integer): Boolean; begin result := (a = b); end; +function hashIntEqu (constref a, b: Integer): Boolean; begin result := (a = b); end; +function hashStrEqu (constref a, b: AnsiString): Boolean; begin result := (a = b); end; +procedure hashStrFree (var s: AnsiString); begin s := ''; end; {$PUSH} {$RANGECHECKS OFF} -function hiihash (constref k: Integer): LongWord; +function hashIntHash (constref k: Integer): LongWord; begin - result := k; + result := LongWord(k); result -= (result shl 6); result := result xor (result shr 17); result -= (result shl 9); @@ -167,12 +249,35 @@ begin result := result xor (result shl 10); result := result xor (result shr 15); end; + +function hashStrHash (constref k: AnsiString): LongWord; +begin + if (Length(k) > 0) then result := fnvHash(PAnsiChar(k)^, Length(k)) else result := 0; +end; {$POP} function hashNewIntInt (): THashIntInt; begin - result := THashIntInt.Create(hiihash, hiiequ); + result := THashIntInt.Create(hashIntHash, hashIntEqu); +end; + + +function hashNewStrInt (): THashStrInt; +begin + result := THashStrInt.Create(hashStrHash, hashStrEqu, hashStrFree); +end; + + +function hashNewIntStr (): THashIntStr; +begin + result := THashIntStr.Create(hashIntHash, hashIntEqu, nil, hashStrFree); +end; + + +function hashNewStrStr (): THashStrStr; +begin + result := THashStrStr.Create(hashStrHash, hashStrEqu, hashStrFree, hashStrFree); end; @@ -198,7 +303,7 @@ begin end; -procedure TJoaatHasher.put (const buf; len: LongWord); +procedure TJoaatHasher.put (constref buf; len: LongWord); var bytes: PByte; h: LongWord; @@ -228,12 +333,12 @@ end; {$POP} -function joaatHash (const buf; len: LongWord): LongWord; +function joaatHash (constref buf; len: LongWord): LongWord; var h: TJoaatHasher; begin h := TJoaatHasher.Create(0); - h.put(buf, len); + h.put(PByte(@buf)^, len); result := h.value; end; @@ -242,7 +347,7 @@ end; {$PUSH} {$RANGECHECKS OFF} // fnv-1a: http://www.isthe.com/chongo/tech/comp/fnv/ -function fnvHash (const buf; len: LongWord): LongWord; +function fnvHash (constref buf; len: LongWord): LongWord; var b: PByte; begin @@ -276,15 +381,26 @@ end; // ////////////////////////////////////////////////////////////////////////// // -constructor THashBase.Create (ahashfn: THashFn; aequfn: TEquFn); +function THashBase.TEntry.getEmpty (): Boolean; inline; begin result := (hash = 0); end; + + +// ////////////////////////////////////////////////////////////////////////// // +function THashBase.getCapacity (): Integer; inline; begin result := Length(mBuckets); end; + + +constructor THashBase.Create (ahashfn: THashFn; aequfn: TEquFn; afreekeyfn: TFreeKeyFn=nil; afreevalfn: TFreeValueFn=nil); begin if not assigned(ahashfn) then raise Exception.Create('cannot create hash without hash function'); if not assigned(aequfn) then raise Exception.Create('cannot create hash without equality function'); hashfn := ahashfn; equfn := aequfn; + freekeyfn := afreekeyfn; + freevalfn := afreevalfn; mSeed := u32Hash($29a); + mFirstEntry := -1; + mLastEntry := -1; clear(); end; @@ -297,68 +413,88 @@ begin end; -procedure THashBase.clear (); +procedure THashBase.freeEntries (); var - idx: Integer; + f: Integer; + e: PEntry; begin - SetLength(mBuckets, InitSize); - for idx := 0 to High(mBuckets) do mBuckets[idx] := nil; - - SetLength(mEntries, Length(mBuckets)); - for idx := 0 to High(mEntries)-1 do + if (mFirstEntry >= 0) then + begin + for f := mFirstEntry to mLastEntry do + begin + e := @mEntries[f]; + if not e.empty then + begin + if assigned(freekeyfn) then freekeyfn(e.key); + if assigned(freevalfn) then freevalfn(e.value); + e.key := Default(KeyT); + e.value := Default(ValueT); + e.hash := 0; + end; + end; + end + else if (Length(mEntries) > 0) then begin - mEntries[idx].hash := 0; - mEntries[idx].nextFree := @mEntries[idx+1]; //idx+1; + FillChar(mEntries[0], Length(mEntries)*sizeof(mEntries[0]), 0); end; - mEntries[High(mEntries)].hash := 0; - mEntries[High(mEntries)].nextFree := nil; - - mBucketsUsed := 0; + mFreeEntryHead := nil; + mFirstEntry := -1; + mLastEntry := -1; {$IFDEF RBHASH_SANITY_CHECKS} mEntriesUsed := 0; {$ENDIF} - mFreeEntryHead := @mEntries[0]; - mFirstEntry := -1; - mLastEntry := -1; +end; + + +procedure THashBase.clear (); +//var idx: Integer; +begin + freeEntries(); + SetLength(mBuckets, InitSize); + FillChar(mBuckets[0], InitSize*sizeof(mBuckets[0]), 0); + //for idx := 0 to High(mBuckets) do mBuckets[idx] := nil; + SetLength(mEntries, InitSize); + FillChar(mEntries[0], InitSize*sizeof(mEntries[0]), 0); + mBucketsUsed := 0; end; procedure THashBase.reset (); -var - idx: Integer; +//var idx: Integer; begin + freeEntries(); if (mBucketsUsed > 0) then begin - for idx := 0 to High(mBuckets) do mBuckets[idx] := nil; - for idx := 0 to High(mEntries)-1 do - begin - mEntries[idx].hash := 0; - mEntries[idx].nextFree := @mEntries[idx+1]; //idx+1; - end; - mEntries[High(mEntries)].hash := 0; - mEntries[High(mEntries)].nextFree := nil; - + //for idx := 0 to High(mBuckets) do mBuckets[idx] := nil; + FillChar(mBuckets[0], Length(mBuckets)*sizeof(mBuckets[0]), 0); mBucketsUsed := 0; - {$IFDEF RBHASH_SANITY_CHECKS} - mEntriesUsed := 0; - {$ENDIF} - mFreeEntryHead := @mEntries[0]; - mFirstEntry := -1; - mLastEntry := -1; end; end; -function THashBase.getCapacity (): Integer; inline; begin result := Length(mBuckets); end; - - function THashBase.allocEntry (): PEntry; var idx: Integer; begin + if (mFreeEntryHead = nil) then + begin + if (mLastEntry = High(mEntries)) then raise Exception.Create('internal error in hash entry allocator (0.0)'); + Inc(mLastEntry); + if (mFirstEntry = -1) then + begin + if (mLastEntry <> 0) then raise Exception.Create('internal error in hash entry allocator (0.1)'); + mFirstEntry := 0; + end; + result := @mEntries[mLastEntry]; + result.nextFree := nil; // just in case + {$IFDEF RBHASH_SANITY_CHECKS} + Inc(mEntriesUsed); + {$ENDIF} + exit; + end; {$IFDEF RBHASH_SANITY_CHECKS} if (mFreeEntryHead = nil) then raise Exception.Create('internal error in hash entry allocator (0)'); - if (mFreeEntryHead.hash <> 0) then raise Exception.Create('internal error in hash entry allocator (1)'); + if (not mFreeEntryHead.empty) then raise Exception.Create('internal error in hash entry allocator (1)'); {$ENDIF} result := mFreeEntryHead; mFreeEntryHead := result.nextFree; @@ -385,19 +521,23 @@ begin if (mEntriesUsed = 0) then raise Exception.Create('internal error in hash entry allocator'); if (mEntriesUsed <> mBucketsUsed) then raise Exception.Create('internal error in hash entry allocator (entry/bucket count mismatch)'); if (e = nil) then raise Exception.Create('internal error in hash entry allocator (trying to release nil entry)'); - if (e.hash = 0) then raise Exception.Create('internal error in hash entry allocator (trying to release unallocated entry)'); + if (e.empty) then raise Exception.Create('internal error in hash entry allocator (trying to release unallocated entry)'); {$ENDIF} idx := Integer((PtrUInt(e)-PtrUInt(@mEntries[0])) div sizeof(mEntries[0])); {$IFDEF RBHASH_SANITY_CHECKS} if (idx < 0) or (idx > High(mEntries)) then raise Exception.Create('internal error in hash entry allocator (invalid entry address)'); if (e <> @mEntries[idx]) then raise Exception.Create('internal error in hash entry allocator (wtf?!)'); {$ENDIF} - e.hash := 0; - e.nextFree := mFreeEntryHead; - mFreeEntryHead := e; //idx; + if assigned(freekeyfn) then freekeyfn(e.key); + if assigned(freevalfn) then freevalfn(e.value); {$IFDEF RBHASH_SANITY_CHECKS} Dec(mEntriesUsed); {$ENDIF} + e.key := Default(KeyT); + e.value := Default(ValueT); + e.hash := 0; + e.nextFree := mFreeEntryHead; + mFreeEntryHead := e; // fix mFirstEntry and mLastEntry {$IFDEF RBHASH_SANITY_CHECKS} if (mFirstEntry < 0) or (mLastEntry < 0) then raise Exception.Create('internal error in hash entry allocator (invalid first/last range; 0)'); @@ -407,6 +547,7 @@ begin {$IFDEF RBHASH_SANITY_CHECKS} if (mEntriesUsed <> 0) then raise Exception.Create('internal error in hash entry allocator (invalid first/last range; 1)'); {$ENDIF} + mFreeEntryHead := nil; mFirstEntry := -1; mLastEntry := -1; end @@ -419,7 +560,7 @@ begin if (idx = mFirstEntry) then begin cidx := idx+1; - while (mEntries[cidx].hash = 0) do Inc(cidx); + while (mEntries[cidx].empty) do Inc(cidx); {$IFDEF RBHASH_SANITY_CHECKS} if (cidx > High(mEntries)) then raise Exception.Create('internal error in hash entry allocator (invalid first/last range; 3)'); {$ENDIF} @@ -429,7 +570,7 @@ begin if (idx = mLastEntry) then begin cidx := idx-1; - while (mEntries[cidx].hash = 0) do Dec(cidx); + while (mEntries[cidx].empty) do Dec(cidx); {$IFDEF RBHASH_SANITY_CHECKS} if (cidx < 0) then raise Exception.Create('internal error in hash entry allocator (invalid first/last range; 3)'); {$ENDIF} @@ -439,40 +580,47 @@ begin end; -(* function THashBase.distToStIdx (idx: LongWord): LongWord; inline; begin {$IFDEF RBHASH_SANITY_CHECKS} assert(idx < Length(mBuckets)); assert(mBuckets[idx] <> nil); {$ENDIF} - result := mBuckets[idx].hash and High(mBuckets); + result := (mBuckets[idx].hash xor mSeed) and High(mBuckets); if (result <= idx) then result := idx-result else result := idx+(Length(mBuckets)-result); end; -*) -function THashBase.has (constref akey: KeyT): Boolean; +function THashBase.has (constref akey: KeyT; keyhashin: PLongWord=nil): Boolean; var khash, idx: LongWord; dist, pdist: LongWord; - bhigh: LongWord; + bhigh, xseed: LongWord; begin result := false; if (mBucketsUsed = 0) then exit; bhigh := High(mBuckets); - khash := hashfn(akey) xor mSeed; if (khash = 0) then khash := $29a; - idx := khash and bhigh; + xseed := mSeed; + + if (keyhashin <> nil) then + begin + khash := keyhashin^; + if (khash = 0) then khash := hashfn(akey); + end + else + begin + khash := hashfn(akey); + end; + if (khash = 0) then khash := $29a; + + idx := (khash xor xseed) and bhigh; if (mBuckets[idx] = nil) then exit; for dist := 0 to bhigh do begin if (mBuckets[idx] = nil) then break; - //pdist := distToStIdx(idx); - pdist := mBuckets[idx].hash and bhigh; - if (pdist <= idx) then pdist := idx-pdist else pdist := idx+((bhigh+1)-pdist); - // + pdist := distToStIdx(idx); if (dist > pdist) then break; result := (mBuckets[idx].hash = khash) and equfn(mBuckets[idx].key, akey); if result then break; @@ -481,34 +629,38 @@ begin end; -function THashBase.get (constref akey: KeyT; out rval: ValueT): Boolean; +function THashBase.get (constref akey: KeyT; out rval: ValueT; keyhashin: PLongWord=nil): Boolean; var khash, idx: LongWord; dist, pdist: LongWord; - bhigh: LongWord; + bhigh, xseed: LongWord; begin result := false; if (mBucketsUsed = 0) then begin rval := Default(ValueT); exit; end; bhigh := High(mBuckets); - khash := hashfn(akey) xor mSeed; if (khash = 0) then khash := $29a; - idx := khash and bhigh; - if (mBuckets[idx] = nil) then begin rval := Default(ValueT); exit; end; + xseed := mSeed; + + if (keyhashin <> nil) then + begin + khash := keyhashin^; + if (khash = 0) then khash := hashfn(akey); + end + else + begin + khash := hashfn(akey); + end; + if (khash = 0) then khash := $29a; + + idx := (khash xor xseed) and bhigh; for dist := 0 to bhigh do begin if (mBuckets[idx] = nil) then break; - //pdist := distToStIdx(idx); - pdist := mBuckets[idx].hash and bhigh; - if (pdist <= idx) then pdist := idx-pdist else pdist := idx+((bhigh+1)-pdist); - // + pdist := distToStIdx(idx); if (dist > pdist) then break; result := (mBuckets[idx].hash = khash) and equfn(mBuckets[idx].key, akey); - if result then - begin - rval := mBuckets[idx].value; - break; - end; + if result then begin rval := mBuckets[idx].value; break; end; idx := (idx+1) and bhigh; end; @@ -519,11 +671,12 @@ end; procedure THashBase.putEntryInternal (swpe: PEntry); var idx, dist, pcur, pdist: LongWord; - tmpe: PEntry; // current entry to swap (or nothing) - bhigh: LongWord; + tmpe: PEntry; + bhigh, xseed: LongWord; begin bhigh := High(mBuckets); - idx := swpe.hash and bhigh; + xseed := mSeed; + idx := (swpe.hash xor xseed) and bhigh; {$IFDEF RBHASH_DEBUG_INSERT}writeln('inserting key ', swpe.key, '; value=', swpe.value, '; wantidx=', idx, '; bhigh=', bhigh);{$ENDIF} pcur := 0; for dist := 0 to bhigh do @@ -536,10 +689,7 @@ begin Inc(mBucketsUsed); break; end; - //pdist := distToStIdx(idx); - pdist := mBuckets[idx].hash and bhigh; - if (pdist <= idx) then pdist := idx-pdist else pdist := idx+((bhigh+1)-pdist); - // + pdist := distToStIdx(idx); if (pcur > pdist) then begin // swapping the current bucket with the one to insert @@ -554,18 +704,21 @@ begin end; -function THashBase.put (constref akey: KeyT; constref aval: ValueT): Boolean; +function THashBase.put (constref akey: KeyT; constref aval: ValueT; keyhashout: PLongWord=nil): Boolean; var khash, idx, dist, pdist: LongWord; swpe: PEntry = nil; // current entry to swap (or nothing) - bhigh: LongWord; + bhigh, xseed: LongWord; newsz, eidx: Integer; begin result := false; bhigh := High(mBuckets); - khash := hashfn(akey) xor mSeed; if (khash = 0) then khash := $29a; - idx := khash and bhigh; + xseed := mSeed; + khash := hashfn(akey); + if (khash = 0) then khash := $29a; + if (keyhashout <> nil) then keyhashout^ := khash; + idx := (khash xor xseed) and bhigh; // check if we already have this key if (mBucketsUsed <> 0) and (mBuckets[idx] <> nil) then @@ -573,16 +726,15 @@ begin for dist := 0 to bhigh do begin if (mBuckets[idx] = nil) then break; - //pdist := distToStIdx(idx); - pdist := mBuckets[idx].hash and bhigh; - if (pdist <= idx) then pdist := idx-pdist else pdist := idx+((bhigh+1)-pdist); - // + pdist := distToStIdx(idx); if (dist > pdist) then break; result := (mBuckets[idx].hash = khash) and equfn(mBuckets[idx].key, akey); if result then begin // replace element - //mBuckets[idx].key := akey; + if assigned(freekeyfn) then freekeyfn(mBuckets[idx].key); + if assigned(freevalfn) then freevalfn(mBuckets[idx].value); + mBuckets[idx].key := akey; mBuckets[idx].value := aval; exit; end; @@ -607,8 +759,6 @@ begin // mFreeEntryHead will be fixed in `rehash()` // reinsert entries rehash(); - // as seed was changed, recalc hash - khash := hashfn(akey) xor mSeed; if (khash = 0) then khash := $29a; end; // create new entry @@ -622,27 +772,36 @@ end; // see http://codecapsule.com/2013/11/17/robin-hood-hashing-backward-shift-deletion/ -function THashBase.del (constref akey: KeyT): Boolean; +function THashBase.del (constref akey: KeyT; keyhashin: PLongWord=nil): Boolean; var khash, idx, idxnext, pdist, dist: LongWord; - bhigh: LongWord; + bhigh, xseed: LongWord; begin result := false; if (mBucketsUsed = 0) then exit; bhigh := High(mBuckets); - khash := hashfn(akey) xor mSeed; if (khash = 0) then khash := $29a; - idx := khash and bhigh; + xseed := mSeed; + + if (keyhashin <> nil) then + begin + khash := keyhashin^; + if (khash = 0) then khash := hashfn(akey); + end + else + begin + khash := hashfn(akey); + end; + if (khash = 0) then khash := $29a; + + idx := (khash xor xseed) and bhigh; // find key if (mBuckets[idx] = nil) then exit; // no key for dist := 0 to bhigh do begin if (mBuckets[idx] = nil) then break; - //pdist := distToStIdx(idxcur); - pdist := mBuckets[idx].hash and bhigh; - if (pdist <= idx) then pdist := idx-pdist else pdist := idx+((bhigh+1)-pdist); - // + pdist := distToStIdx(idx); if (dist > pdist) then break; result := (mBuckets[idx].hash = khash) and equfn(mBuckets[idx].key, akey); if result then break; @@ -670,10 +829,7 @@ begin writeln(' dist=', dist, '; idx=', idx, '; idxnext=', idxnext, '; ce=', (mBuckets[idx] <> nil), '; ne=', (mBuckets[idxnext] <> nil)); {$ENDIF} if (mBuckets[idxnext] = nil) then begin {$IFDEF RBHASH_DEBUG_DELETE}writeln(' idxnext nil');{$ENDIF} mBuckets[idx] := nil; break; end; - //pdist := distToStIdx(idxnext); - pdist := mBuckets[idxnext].hash and bhigh; - if (pdist <= idxnext) then pdist := idxnext-pdist else pdist := idxnext+((bhigh+1)-pdist); - // + pdist := distToStIdx(idxnext); if (pdist = 0) then begin {$IFDEF RBHASH_DEBUG_DELETE}writeln(' pdist is zero');{$ENDIF} mBuckets[idx] := nil; break; end; {$IFDEF RBHASH_DEBUG_DELETE}writeln(' pdist=', pdist);{$ENDIF} mBuckets[idx] := mBuckets[idxnext]; @@ -695,10 +851,12 @@ var {$ENDIF} begin // change seed, to minimize pathological cases + //TODO: use prng to generate new hash if (mSeed = 0) then mSeed := $29a; mSeed := u32Hash(mSeed); // clear buckets - for idx := 0 to High(mBuckets) do mBuckets[idx] := nil; + //for idx := 0 to High(mBuckets) do mBuckets[idx] := nil; + FillChar(mBuckets[0], Length(mBuckets)*sizeof(mBuckets[0]), 0); mBucketsUsed := 0; // reinsert entries mFreeEntryHead := nil; @@ -706,7 +864,7 @@ begin for idx := 0 to High(mEntries) do begin e := @mEntries[idx]; - if (e.hash <> 0) then + if (not e.empty) then begin {$IFDEF RBHASH_SANITY_CHECKS} if (e.nextFree <> nil) then raise Exception.Create('internal error in rehash: inconsistent'); @@ -714,7 +872,8 @@ begin Inc(cnt); if (cnt = mBucketsUsed) and (idx <> mLastEntry) then raise Exception.Create('internal error in rehash: inconsistent (2)'); {$ENDIF} - e.hash := hashfn(e.key) xor mSeed; if (e.hash = 0) then e.hash := $29a; + // no need to recalculate hash + //e.hash := hashfn(e.key) xor mSeed; if (e.hash = 0) then e.hash := $29a; putEntryInternal(e); end else @@ -751,12 +910,12 @@ begin if (mBucketsUsed < 1) then raise Exception.Create('internal error in hash table (invalid bucket count; 0)'); {$ENDIF} didx := 0; - while (didx < Length(mEntries)) do if (mEntries[didx].hash <> 0) then Inc(didx) else break; + while (didx < Length(mEntries)) do if (not mEntries[didx].empty) then Inc(didx) else break; f := didx+1; // copy entries while true do begin - if (mEntries[f].hash <> 0) then + if (not mEntries[f].empty) then begin {$IFDEF RBHASH_SANITY_CHECKS} if (didx >= f) then raise Exception.Create('internal error in hash: inconsistent'); @@ -765,7 +924,7 @@ begin mEntries[f].hash := 0; Inc(didx); if (f = mLastEntry) then break; - while (didx < Length(mEntries)) do if (mEntries[didx].hash <> 0) then Inc(didx) else break; + while (didx < Length(mEntries)) do if (not mEntries[didx].empty) then Inc(didx) else break; end; Inc(f); end; @@ -778,14 +937,14 @@ begin cnt := 0; for f := mFirstEntry to mLastEntry do begin - if (mEntries[f].hash = 0) then raise Exception.Create('internal error in hash table (invalid first/last range; 2)'); + if (mEntries[f].empty) then raise Exception.Create('internal error in hash table (invalid first/last range; 2)'); Inc(cnt); end; if (cnt <> mBucketsUsed) then raise Exception.Create('internal error in hash table (invalid first/last range; 3)'); if (cnt <> mEntriesUsed) then raise Exception.Create('internal error in hash table (invalid first/last range; 4)'); for f := mLastEntry+1 to High(mEntries) do begin - if (mEntries[f].hash <> 0) then raise Exception.Create('internal error in hash table (invalid first/last range; 5)'); + if (not mEntries[f].empty) then raise Exception.Create('internal error in hash table (invalid first/last range; 5)'); end; {$ENDIF} end @@ -804,24 +963,143 @@ begin end; -function THashBase.forEach (it: TIteratorFn): Boolean; +function THashBase.forEach (it: TIteratorFn): Boolean; overload; var - i: Integer; + f: Integer; begin result := false; - if not assigned(it) then exit; - i := mFirstEntry; - if (i < 0) then exit; - while (i <= mLastEntry) do + if not assigned(it) or (mFirstEntry < 0) then exit; + for f := mFirstEntry to mLastEntry do begin - if (mEntries[i].hash <> 0) then + if (not mEntries[f].empty) then begin - result := it(mEntries[i].key, mEntries[i].value); + result := it(mEntries[f].key, mEntries[f].value); if result then exit; end; - Inc(i); end; end; +function THashBase.forEach (it: TIteratorExFn): Boolean; overload; +var + f: Integer; +begin + result := false; + if not assigned(it) or (mFirstEntry < 0) then exit; + for f := mFirstEntry to mLastEntry do + begin + if (not mEntries[f].empty) then + begin + result := it(mEntries[f].key, mEntries[f].value, mEntries[f].hash); + if result then exit; + end; + end; +end; + + +// enumerators +function THashBase.GetEnumerator (): TValEnumerator; +begin + if (Length(mEntries) > 0) then result := TValEnumerator.Create(mEntries, mFirstEntry, mLastEntry) + else result := TValEnumerator.Create(nil, -1, -1); +end; + +function THashBase.byKey (): TKeyEnumerator; +begin + if (Length(mEntries) > 0) then result := TKeyEnumerator.Create(mEntries, mFirstEntry, mLastEntry) + else result := TKeyEnumerator.Create(nil, -1, -1); +end; + +function THashBase.byValue (): TValEnumerator; +begin + if (Length(mEntries) > 0) then result := TValEnumerator.Create(mEntries, mFirstEntry, mLastEntry) + else result := TValEnumerator.Create(nil, -1, -1); +end; + +function THashBase.byKeyValue (): TKeyValEnumerator; // PEntry +begin + if (Length(mEntries) > 0) then result := TKeyValEnumerator.Create(mEntries, mFirstEntry, mLastEntry) + else result := TKeyValEnumerator.Create(nil, -1, -1); +end; + + +function THashBase.TValEnumerator.GetEnumerator (): TValEnumerator; inline; begin result.mEntries := self.mEntries; result.mFirstEntry := self.mFirstEntry; result.mLastEntry := self.mLastEntry; result.cur := self.cur; end; +function THashBase.TKeyEnumerator.GetEnumerator (): TKeyEnumerator; inline; begin result.mEntries := self.mEntries; result.mFirstEntry := self.mFirstEntry; result.mLastEntry := self.mLastEntry; result.cur := self.cur; end; +function THashBase.TKeyValEnumerator.GetEnumerator (): TKeyValEnumerator; inline; begin result.mEntries := self.mEntries; result.mFirstEntry := self.mFirstEntry; result.mLastEntry := self.mLastEntry; result.cur := self.cur; end; + + +// ////////////////////////////////////////////////////////////////////////// // +constructor THashBase.TValEnumerator.Create (const aents: TEntryArray; afirst, alast: Integer); +begin + mEntries := aents; + mFirstEntry := afirst; + mLastEntry := alast; + cur := mFirstEntry-1; +end; + +function THashBase.TValEnumerator.MoveNext (): Boolean; inline; +begin + Inc(cur); + while (cur <= mLastEntry) do + begin + if (not mEntries[cur].empty) then begin result := true; exit; end; + end; + result := false; +end; + +function THashBase.TValEnumerator.getCurrent (): ValueT; inline; +begin + result := mEntries[cur].value; +end; + + +// ////////////////////////////////////////////////////////////////////////// // +constructor THashBase.TKeyEnumerator.Create (const aents: TEntryArray; afirst, alast: Integer); +begin + mEntries := aents; + mFirstEntry := afirst; + mLastEntry := alast; + cur := mFirstEntry-1; +end; + +function THashBase.TKeyEnumerator.MoveNext (): Boolean; inline; +begin + Inc(cur); + while (cur <= mLastEntry) do + begin + if (not mEntries[cur].empty) then begin result := true; exit; end; + end; + result := false; +end; + +function THashBase.TKeyEnumerator.getCurrent (): KeyT; inline; +begin + result := mEntries[cur].key; +end; + + +// ////////////////////////////////////////////////////////////////////////// // +constructor THashBase.TKeyValEnumerator.Create (const aents: TEntryArray; afirst, alast: Integer); +begin + mEntries := aents; + mFirstEntry := afirst; + mLastEntry := alast; + cur := mFirstEntry-1; +end; + +function THashBase.TKeyValEnumerator.MoveNext (): Boolean; inline; +begin + Inc(cur); + while (cur <= mLastEntry) do + begin + if (not mEntries[cur].empty) then begin result := true; exit; end; + end; + result := false; +end; + +function THashBase.TKeyValEnumerator.getCurrent (): PEntry; inline; +begin + result := @mEntries[cur]; +end; + end.