// // Copyright (c) Microsoft Corporation 1991-1993 // // File: Hash.c // // Comments: // This file contains functions that are roughly equivelent to the // kernel atom function. There are two main differences. The first // is that in 32 bit land the tables are maintined in our shared heap, // which makes it shared between all of our apps. The second is that // we can assocate a long pointer with each of the items, which in many // cases allows us to keep from having to do a secondary lookup from // a different table // // History: // 09/08/93 - Created KurtE // ??/??/94 - ported for unicode (anonymous) // 10/26/95 - rearranged hashitems for perf, alignment FrancisH // //--------------------------------------------------------------------------- #include "shellprv.h" #pragma hdrstop #include "fstreex.h" // for SHCF_ICON_INDEX #define DM_PERF 0 // perf stats //-------------------------------------------------------------------------- // First define a data structure to use to maintain the list #define DEF_HASH_BUCKET_COUNT 71 // NOTE a PHASHITEM is defined as a LPCSTR externaly (for old code to work) #undef PHASHITEM typedef struct _HashItem * PHASHITEM; //----------------------------------------------------------------------------- // // Hash item layout: // // [extra data][_HashItem struct][item text] // //----------------------------------------------------------------------------- typedef struct _HashItem { // // this part of the struct is aligned // PHASHITEM phiNext; // WORD wCount; // Usage count WORD cchLen; // Length of name in characters. // // this member is just a placeholder // TCHAR szName[1]; // name } HASHITEM; #pragma warning(disable:4200) // Zero-sized array in struct typedef struct _HashTable { UINT uBuckets; // Number of buckets UINT uItems; // Number of items UINT cbExtra; // Extra bytes per item LPCTSTR pszHTCache; // MRU ptr for last lookup/add/etc. PHASHITEM ahiBuckets[0]; // Set of buckets for the table } HASHTABLE, * PHASHTABLE; #define HIFROMSZ(sz) ((PHASHITEM)((BYTE*)(sz) - FIELD_OFFSET(HASHITEM, szName))) #define HIDATAPTR(pht, sz) ((void *)(((BYTE *)HIFROMSZ(sz)) - (pht? pht->cbExtra : 0))) #define HIDATAARRAY(pht, sz) ((DWORD_PTR *)HIDATAPTR(pht, sz)) #define LOOKUPHASHITEM 0 #define ADDHASHITEM 1 #define DELETEHASHITEM 2 #define PURGEHASHITEM 3 // DANGER: EVIL! static HHASHTABLE g_hHashTable = NULL; HHASHTABLE GetGlobalHashTable(); PHASHTABLE _CreateHashTable(UINT uBuckets, UINT cbExtra); //-------------------------------------------------------------------------- // This function allocs a hashitem. // PHASHITEM _AllocHashItem(PHASHTABLE pht, DWORD cchName) { BYTE *mem; ASSERT(pht); mem = (BYTE *)LocalAlloc(LPTR, SIZEOF(HASHITEM) + (cchName * SIZEOF(TCHAR)) + pht->cbExtra); if (mem) mem += pht->cbExtra; return (PHASHITEM)mem; } //-------------------------------------------------------------------------- // This function frees a hashitem. // __inline void _FreeHashItem(PHASHTABLE pht, PHASHITEM phi) { ASSERT(pht && phi); LocalFree((BYTE *)phi - pht->cbExtra); } // PERF_CACHE //*** c_szHTNil -- 1-element MRU for hashtable // DESCRIPTION // it turns out we have long runs of duplicate lookups (e.g. "Directory" // and ".lnk"). a 1-element MRU is a v. cheap way of speeding things up. // rather than check for the (rare) special case of NULL each time we // check our cache, we pt at at this guy. then iff we think it's a // cache hit, we make sure it's not pting at this special guy. const TCHAR c_szHTNil[] = TEXT(""); // arbitrary value, unique-& #ifdef DEBUG int g_cHTTot, g_cHTHit; int g_cHTMod = 100; #endif // -------------------------------------------------------- // Compute a hash value from an input string of any type, i.e. // the input is just treated as a sequence of bytes. // Based on a hash function originally proposed by J. Zobel. // Author: Paul Larson, 1999, palarson@microsoft.com // -------------------------------------------------------- ULONG _CalculateHashKey(LPCTSTR pszName, WORD *pcch) { // initialize HashKey to a reasonably large constant so very // short keys won't get mapped to small values. Virtually any // large odd constant will do. unsigned int HashKey = 314159269 ; TUCHAR *pC = (TUCHAR *)pszName; for(; *pC; pC++){ HashKey ^= (HashKey<<11) + (HashKey<<5) + (HashKey>>2) + (unsigned int) *pC ; } if (pcch) *pcch = (WORD)(pC - pszName); return (HashKey & 0x7FFFFFFF) ; } void _GrowTable(HHASHTABLE hht) { // hht can't be NULL here PHASHTABLE pht = *hht; PHASHTABLE phtNew = _CreateHashTable((pht->uBuckets * 2) -1, pht->cbExtra); if (phtNew) { int i; for (i=0; i<(int)pht->uBuckets; i++) { PHASHITEM phi; PHASHITEM phiNext; for (phi=pht->ahiBuckets[i]; phi; phi=phiNext) { // We always use case sensitive hash here since the case has already been fixed when adding the key. ULONG uBucket = _CalculateHashKey(phi->szName, NULL) % phtNew->uBuckets; phiNext = phi->phiNext; // And link it in to the right bucket phi->phiNext = phtNew->ahiBuckets[uBucket]; phtNew->ahiBuckets[uBucket] = phi; phtNew->uItems++; // One more item in the table } } ASSERT(phtNew->uItems == pht->uItems); // Now switch the 2 tables LocalFree(pht); *hht = phtNew; } } //-------------------------------------------------------------------------- // This function looks up the name in the hash table and optionally does // things like add it, or delete it. // LPCTSTR LookupItemInHashTable(HHASHTABLE hht, LPCTSTR pszName, int iOp) { // First thing to do is calculate the hash value for the item UINT uBucket; WORD cchName; PHASHITEM phi, phiPrev; PHASHTABLE pht; ENTERCRITICAL; pht = hht ? *hht : NULL; ASSERT(!hht || pht); // If hht is not NULL, then pht can't be NULL either if (pht == NULL) { hht = GetGlobalHashTable(); if (hht) { pht = *hht; } if (pht == NULL) { TraceMsg(TF_WARNING, "LookupItemInHashTable() - Can't get global hash table!"); LEAVECRITICAL; return NULL; } } #ifdef DEBUG if ((g_cHTTot % g_cHTMod) == 0) TraceMsg(DM_PERF, "ht: tot=%d hit=%d", g_cHTTot, g_cHTHit); #endif DBEXEC(TRUE, g_cHTTot++); if (*pszName == *pht->pszHTCache && iOp == LOOKUPHASHITEM) { // StrCmpC is a fast ansi strcmp, good enough for a quick/approx check if (StrCmpC(pszName, pht->pszHTCache) == 0 && pht->pszHTCache != c_szHTNil) { DBEXEC(TRUE, g_cHTHit++); LEAVECRITICAL; // see 'semi-race' comment below return (LPCTSTR)pht->pszHTCache; #if 0 // currently not worth it (very few ADDHASHITEMs of existing) // careful! this is o.k. for ADDHASHITEM but not for (e.g.) // DELETEHASHITEM (which needs phiPrev) phi = HIFROMSZ(pht->pszHTCache); goto Ldoop; // warning: pending ENTERCRITICAL #endif } } uBucket = _CalculateHashKey(pszName, &cchName) % pht->uBuckets; // now search for the item in the buckets. phiPrev = NULL; phi = pht->ahiBuckets[uBucket]; while (phi) { if (phi->cchLen == cchName) { if (!lstrcmp(pszName, phi->szName)) break; // Found match } phiPrev = phi; // Keep the previous item phi = phi->phiNext; } // // Sortof gross, but do the work here // switch (iOp) { case ADDHASHITEM: if (phi) { // Simply increment the reference count DebugMsg(TF_HASH, TEXT("Add Hit on '%s'"), pszName); phi->wCount++; } else { DebugMsg(TF_HASH, TEXT("Add MISS on '%s'"), pszName); // Not Found, try to allocate it out of the heap if ((phi = _AllocHashItem(pht, cchName)) != NULL) { // Initialize it phi->wCount = 1; // One use of it phi->cchLen = cchName; // The length of it; lstrcpy(phi->szName, pszName); // And link it in to the right bucket phi->phiNext = pht->ahiBuckets[uBucket]; pht->ahiBuckets[uBucket] = phi; pht->uItems++; // One more item in the table if (pht->uItems > pht->uBuckets) { _GrowTable(hht); pht = *hht; } TraceMsg(TF_HASH, "Added new hash item %x(phiNext=%x,szName=\"%s\") for hash table %x at bucket %x", phi, phi->phiNext, phi->szName, pht, uBucket); } } break; case PURGEHASHITEM: case DELETEHASHITEM: if (phi && ((iOp == PURGEHASHITEM) || (!--phi->wCount))) { // Useage count went to zero so unlink it and delete it if (phiPrev != NULL) phiPrev->phiNext = phi->phiNext; else pht->ahiBuckets[uBucket] = phi->phiNext; // And delete it TraceMsg(TF_HASH, "Free hash item %x(szName=\"%s\") from hash table %x at bucket %x", phi, phi->szName, pht, uBucket); _FreeHashItem(pht, phi); phi = NULL; pht->uItems--; // One less item in the table } } // kill cache if this was a PURGE/DELETEHASHITEM, o.w. cache it. // note that there's a semi-race on pht->pszHTCache ops, viz. that // we LEAVECRITICAL but then return a ptr into our table. however // it's 'no worse' than the existing races. so i guess the caller // is supposed to avoid a concurrent lookup/delete. pht->pszHTCache = phi ? phi->szName : c_szHTNil; LEAVECRITICAL; // If find was passed in simply return it. if (phi) return (LPCTSTR)phi->szName; else return NULL; } //-------------------------------------------------------------------------- LPCTSTR WINAPI FindHashItem(HHASHTABLE hht, LPCTSTR lpszStr) { return LookupItemInHashTable(hht, lpszStr, LOOKUPHASHITEM); } //-------------------------------------------------------------------------- LPCTSTR WINAPI AddHashItem(HHASHTABLE hht, LPCTSTR lpszStr) { return LookupItemInHashTable(hht, lpszStr, ADDHASHITEM); } //-------------------------------------------------------------------------- LPCTSTR WINAPI DeleteHashItem(HHASHTABLE hht, LPCTSTR lpszStr) { return LookupItemInHashTable(hht, lpszStr, DELETEHASHITEM); } //-------------------------------------------------------------------------- LPCTSTR WINAPI PurgeHashItem(HHASHTABLE hht, LPCTSTR lpszStr) { return LookupItemInHashTable(hht, lpszStr, PURGEHASHITEM); } //-------------------------------------------------------------------------- // this sets the extra data in an HashItem void WINAPI SetHashItemData(HHASHTABLE hht, LPCTSTR sz, int n, DWORD_PTR dwData) { PHASHTABLE pht; ENTERCRITICAL; pht = hht ? *hht : NULL; ASSERT(!hht || pht); // If hht is not NULL, then pht can't be NULL either // string must be from the hash table ASSERT(FindHashItem(hht, sz) == sz); // the default hash table does not have extra data! if ((pht != NULL) && (n >= 0) && (n < (int)(pht->cbExtra/SIZEOF(DWORD_PTR)))) HIDATAARRAY(pht, sz)[n] = dwData; LEAVECRITICAL; } //====================================================================== // this is like SetHashItemData, except it gets the HashItem data... DWORD_PTR WINAPI GetHashItemData(HHASHTABLE hht, LPCTSTR sz, int n) { DWORD_PTR dwpRet; PHASHTABLE pht; ENTERCRITICAL; pht = hht ? *hht : NULL; ASSERT(!hht || pht); // If hht is not NULL, then pht can't be NULL either // string must be from the hash table ASSERT(FindHashItem(hht, sz) == sz); // the default hash table does not have extra data! if (pht != NULL && n <= (int)(pht->cbExtra/SIZEOF(DWORD_PTR))) dwpRet = HIDATAARRAY(pht, sz)[n]; else dwpRet = 0; LEAVECRITICAL; return dwpRet; } //====================================================================== // like GetHashItemData, except it just gets a pointer to the buffer... void * WINAPI GetHashItemDataPtr(HHASHTABLE hht, LPCTSTR sz) { void *pvRet; PHASHTABLE pht; ENTERCRITICAL; pht = hht ? *hht : NULL; ASSERT(!hht || pht); // If hht is not NULL, then pht can't be NULL either // string must be from the hash table ASSERT(FindHashItem(hht, sz) == sz); // the default hash table does not have extra data! pvRet = (pht? HIDATAPTR(pht, sz) : NULL); LEAVECRITICAL; return pvRet; } //====================================================================== PHASHTABLE _CreateHashTable(UINT uBuckets, UINT cbExtra) { PHASHTABLE pht; if (uBuckets == 0) uBuckets = DEF_HASH_BUCKET_COUNT; pht = (PHASHTABLE)LocalAlloc(LPTR, SIZEOF(HASHTABLE) + uBuckets * SIZEOF(PHASHITEM)); if (pht) { pht->uBuckets = uBuckets; pht->cbExtra = (cbExtra + sizeof(DWORD_PTR) - 1) & ~(sizeof(DWORD_PTR)-1); // rounding to the next DWORD_PTR size pht->pszHTCache = c_szHTNil; } return pht; } HHASHTABLE WINAPI CreateHashItemTable(UINT uBuckets, UINT cbExtra) { PHASHTABLE *hht = NULL; PHASHTABLE pht; pht = _CreateHashTable(uBuckets, cbExtra); if (pht) { hht = (PHASHTABLE *)LocalAlloc(LPTR, sizeof(PHASHTABLE)); if (hht) { *hht = pht; } else { LocalFree(pht); } } TraceMsg(TF_HASH, "Created hash table %x(uBuckets=%x, cbExtra=%x)", pht, pht->uBuckets, pht->cbExtra); return hht; } //====================================================================== void WINAPI EnumHashItems(HHASHTABLE hht, HASHITEMCALLBACK callback, DWORD_PTR dwParam) { PHASHTABLE pht; ENTERCRITICAL; pht = hht ? *hht : NULL; ASSERT(!hht || pht); // If hht is not NULL, then pht can't be NULL either if (!pht && g_hHashTable) pht = *g_hHashTable; if (pht) { int i; PHASHITEM phi; PHASHITEM phiNext; #ifdef DEBUG ULONG uCount = 0; #endif for (i=0; i<(int)pht->uBuckets; i++) { for (phi=pht->ahiBuckets[i]; phi; phi=phiNext) { phiNext = phi->phiNext; (*callback)(hht, phi->szName, phi->wCount, dwParam); #ifdef DEBUG uCount++; #endif } } ASSERT(uCount == pht->uItems); } LEAVECRITICAL; } //====================================================================== void _DeleteHashItem(HHASHTABLE hht, LPCTSTR sz, UINT usage, DWORD_PTR param) { PHASHTABLE pht; ENTERCRITICAL; pht = hht ? *hht : NULL; _FreeHashItem(pht, HIFROMSZ(sz)); LEAVECRITICAL; } //====================================================================== void WINAPI DestroyHashItemTable(HHASHTABLE hht) { PHASHTABLE pht; ENTERCRITICAL; pht = hht ? *hht : NULL; ASSERT(!hht || pht); // If hht is not NULL, then pht can't be NULL either TraceMsg(TF_HASH, "DestroyHashItemTable(pht=%x)", pht); if (pht == NULL) { if (g_hHashTable) { pht = *g_hHashTable; hht = g_hHashTable; g_hHashTable = NULL; } } if (pht) { EnumHashItems(hht, _DeleteHashItem, 0); LocalFree(pht); } if (hht) { LocalFree(hht); } LEAVECRITICAL; } //====================================================================== HHASHTABLE GetGlobalHashTable() { if (!g_hHashTable) { ENTERCRITICAL; g_hHashTable = CreateHashItemTable(0, 0); LEAVECRITICAL; } return g_hHashTable; } //====================================================================== #ifdef DEBUG static int TotalBytes; void CALLBACK _DumpHashItem(HHASHTABLE hht, LPCTSTR sz, UINT usage, DWORD_PTR param) { PHASHTABLE pht; ENTERCRITICAL; pht = hht ? *hht : NULL; DebugMsg(TF_ALWAYS, TEXT(" %08x %5ld \"%s\""), HIFROMSZ(sz), usage, sz); TotalBytes += (HIFROMSZ(sz)->cchLen * SIZEOF(TCHAR)) + SIZEOF(HASHITEM); LEAVECRITICAL; } void CALLBACK _DumpHashItemWithData(HHASHTABLE hht, LPCTSTR sz, UINT usage, DWORD_PTR param) { PHASHTABLE pht; ENTERCRITICAL; pht = hht ? *hht : NULL; DebugMsg(TF_ALWAYS, TEXT(" %08x %5ld %08x \"%s\""), HIFROMSZ(sz), usage, HIDATAARRAY(pht, sz)[0], sz); TotalBytes += (HIFROMSZ(sz)->cchLen * SIZEOF(TCHAR)) + SIZEOF(HASHITEM) + (pht? pht->cbExtra : 0); LEAVECRITICAL; } void WINAPI DumpHashItemTable(HHASHTABLE hht) { PHASHTABLE pht; ENTERCRITICAL; pht = hht ? *hht : NULL; TotalBytes = 0; if (IsFlagSet(g_dwDumpFlags, DF_HASH)) { DebugMsg(TF_ALWAYS, TEXT("Hash Table: %08x"), pht); if (pht && (pht->cbExtra > 0)) { DebugMsg(TF_ALWAYS, TEXT(" Hash Usage dwEx[0] String")); DebugMsg(TF_ALWAYS, TEXT(" -------- ----- -------- ------------------------------")); EnumHashItems(hht, _DumpHashItemWithData, 0); } else { DebugMsg(TF_ALWAYS, TEXT(" Hash Usage String")); DebugMsg(TF_ALWAYS, TEXT(" -------- ----- --------------------------------")); EnumHashItems(hht, _DumpHashItem, 0); } DebugMsg(TF_ALWAYS, TEXT("Total Bytes: %d"), TotalBytes); } LEAVECRITICAL; } #endif