| OLD | NEW |
| (Empty) |
| 1 /* | |
| 2 ** 2014 August 11 | |
| 3 ** | |
| 4 ** The author disclaims copyright to this source code. In place of | |
| 5 ** a legal notice, here is a blessing: | |
| 6 ** | |
| 7 ** May you do good and not evil. | |
| 8 ** May you find forgiveness for yourself and forgive others. | |
| 9 ** May you share freely, never taking more than you give. | |
| 10 ** | |
| 11 ****************************************************************************** | |
| 12 ** | |
| 13 */ | |
| 14 | |
| 15 | |
| 16 | |
| 17 #include "fts5Int.h" | |
| 18 | |
| 19 typedef struct Fts5HashEntry Fts5HashEntry; | |
| 20 | |
| 21 /* | |
| 22 ** This file contains the implementation of an in-memory hash table used | |
| 23 ** to accumuluate "term -> doclist" content before it is flused to a level-0 | |
| 24 ** segment. | |
| 25 */ | |
| 26 | |
| 27 | |
| 28 struct Fts5Hash { | |
| 29 int *pnByte; /* Pointer to bytes counter */ | |
| 30 int nEntry; /* Number of entries currently in hash */ | |
| 31 int nSlot; /* Size of aSlot[] array */ | |
| 32 Fts5HashEntry *pScan; /* Current ordered scan item */ | |
| 33 Fts5HashEntry **aSlot; /* Array of hash slots */ | |
| 34 }; | |
| 35 | |
| 36 /* | |
| 37 ** Each entry in the hash table is represented by an object of the | |
| 38 ** following type. Each object, its key (zKey[]) and its current data | |
| 39 ** are stored in a single memory allocation. The position list data | |
| 40 ** immediately follows the key data in memory. | |
| 41 ** | |
| 42 ** The data that follows the key is in a similar, but not identical format | |
| 43 ** to the doclist data stored in the database. It is: | |
| 44 ** | |
| 45 ** * Rowid, as a varint | |
| 46 ** * Position list, without 0x00 terminator. | |
| 47 ** * Size of previous position list and rowid, as a 4 byte | |
| 48 ** big-endian integer. | |
| 49 ** | |
| 50 ** iRowidOff: | |
| 51 ** Offset of last rowid written to data area. Relative to first byte of | |
| 52 ** structure. | |
| 53 ** | |
| 54 ** nData: | |
| 55 ** Bytes of data written since iRowidOff. | |
| 56 */ | |
| 57 struct Fts5HashEntry { | |
| 58 Fts5HashEntry *pHashNext; /* Next hash entry with same hash-key */ | |
| 59 Fts5HashEntry *pScanNext; /* Next entry in sorted order */ | |
| 60 | |
| 61 int nAlloc; /* Total size of allocation */ | |
| 62 int iSzPoslist; /* Offset of space for 4-byte poslist size */ | |
| 63 int nData; /* Total bytes of data (incl. structure) */ | |
| 64 u8 bDel; /* Set delete-flag @ iSzPoslist */ | |
| 65 | |
| 66 int iCol; /* Column of last value written */ | |
| 67 int iPos; /* Position of last value written */ | |
| 68 i64 iRowid; /* Rowid of last value written */ | |
| 69 char zKey[8]; /* Nul-terminated entry key */ | |
| 70 }; | |
| 71 | |
| 72 /* | |
| 73 ** Size of Fts5HashEntry without the zKey[] array. | |
| 74 */ | |
| 75 #define FTS5_HASHENTRYSIZE (sizeof(Fts5HashEntry)-8) | |
| 76 | |
| 77 | |
| 78 | |
| 79 /* | |
| 80 ** Allocate a new hash table. | |
| 81 */ | |
| 82 int sqlite3Fts5HashNew(Fts5Hash **ppNew, int *pnByte){ | |
| 83 int rc = SQLITE_OK; | |
| 84 Fts5Hash *pNew; | |
| 85 | |
| 86 *ppNew = pNew = (Fts5Hash*)sqlite3_malloc(sizeof(Fts5Hash)); | |
| 87 if( pNew==0 ){ | |
| 88 rc = SQLITE_NOMEM; | |
| 89 }else{ | |
| 90 int nByte; | |
| 91 memset(pNew, 0, sizeof(Fts5Hash)); | |
| 92 pNew->pnByte = pnByte; | |
| 93 | |
| 94 pNew->nSlot = 1024; | |
| 95 nByte = sizeof(Fts5HashEntry*) * pNew->nSlot; | |
| 96 pNew->aSlot = (Fts5HashEntry**)sqlite3_malloc(nByte); | |
| 97 if( pNew->aSlot==0 ){ | |
| 98 sqlite3_free(pNew); | |
| 99 *ppNew = 0; | |
| 100 rc = SQLITE_NOMEM; | |
| 101 }else{ | |
| 102 memset(pNew->aSlot, 0, nByte); | |
| 103 } | |
| 104 } | |
| 105 return rc; | |
| 106 } | |
| 107 | |
| 108 /* | |
| 109 ** Free a hash table object. | |
| 110 */ | |
| 111 void sqlite3Fts5HashFree(Fts5Hash *pHash){ | |
| 112 if( pHash ){ | |
| 113 sqlite3Fts5HashClear(pHash); | |
| 114 sqlite3_free(pHash->aSlot); | |
| 115 sqlite3_free(pHash); | |
| 116 } | |
| 117 } | |
| 118 | |
| 119 /* | |
| 120 ** Empty (but do not delete) a hash table. | |
| 121 */ | |
| 122 void sqlite3Fts5HashClear(Fts5Hash *pHash){ | |
| 123 int i; | |
| 124 for(i=0; i<pHash->nSlot; i++){ | |
| 125 Fts5HashEntry *pNext; | |
| 126 Fts5HashEntry *pSlot; | |
| 127 for(pSlot=pHash->aSlot[i]; pSlot; pSlot=pNext){ | |
| 128 pNext = pSlot->pHashNext; | |
| 129 sqlite3_free(pSlot); | |
| 130 } | |
| 131 } | |
| 132 memset(pHash->aSlot, 0, pHash->nSlot * sizeof(Fts5HashEntry*)); | |
| 133 pHash->nEntry = 0; | |
| 134 } | |
| 135 | |
| 136 static unsigned int fts5HashKey(int nSlot, const u8 *p, int n){ | |
| 137 int i; | |
| 138 unsigned int h = 13; | |
| 139 for(i=n-1; i>=0; i--){ | |
| 140 h = (h << 3) ^ h ^ p[i]; | |
| 141 } | |
| 142 return (h % nSlot); | |
| 143 } | |
| 144 | |
| 145 static unsigned int fts5HashKey2(int nSlot, u8 b, const u8 *p, int n){ | |
| 146 int i; | |
| 147 unsigned int h = 13; | |
| 148 for(i=n-1; i>=0; i--){ | |
| 149 h = (h << 3) ^ h ^ p[i]; | |
| 150 } | |
| 151 h = (h << 3) ^ h ^ b; | |
| 152 return (h % nSlot); | |
| 153 } | |
| 154 | |
| 155 /* | |
| 156 ** Resize the hash table by doubling the number of slots. | |
| 157 */ | |
| 158 static int fts5HashResize(Fts5Hash *pHash){ | |
| 159 int nNew = pHash->nSlot*2; | |
| 160 int i; | |
| 161 Fts5HashEntry **apNew; | |
| 162 Fts5HashEntry **apOld = pHash->aSlot; | |
| 163 | |
| 164 apNew = (Fts5HashEntry**)sqlite3_malloc(nNew*sizeof(Fts5HashEntry*)); | |
| 165 if( !apNew ) return SQLITE_NOMEM; | |
| 166 memset(apNew, 0, nNew*sizeof(Fts5HashEntry*)); | |
| 167 | |
| 168 for(i=0; i<pHash->nSlot; i++){ | |
| 169 while( apOld[i] ){ | |
| 170 int iHash; | |
| 171 Fts5HashEntry *p = apOld[i]; | |
| 172 apOld[i] = p->pHashNext; | |
| 173 iHash = fts5HashKey(nNew, (u8*)p->zKey, (int)strlen(p->zKey)); | |
| 174 p->pHashNext = apNew[iHash]; | |
| 175 apNew[iHash] = p; | |
| 176 } | |
| 177 } | |
| 178 | |
| 179 sqlite3_free(apOld); | |
| 180 pHash->nSlot = nNew; | |
| 181 pHash->aSlot = apNew; | |
| 182 return SQLITE_OK; | |
| 183 } | |
| 184 | |
| 185 static void fts5HashAddPoslistSize(Fts5HashEntry *p){ | |
| 186 if( p->iSzPoslist ){ | |
| 187 u8 *pPtr = (u8*)p; | |
| 188 int nSz = (p->nData - p->iSzPoslist - 1); /* Size in bytes */ | |
| 189 int nPos = nSz*2 + p->bDel; /* Value of nPos field */ | |
| 190 | |
| 191 assert( p->bDel==0 || p->bDel==1 ); | |
| 192 if( nPos<=127 ){ | |
| 193 pPtr[p->iSzPoslist] = (u8)nPos; | |
| 194 }else{ | |
| 195 int nByte = sqlite3Fts5GetVarintLen((u32)nPos); | |
| 196 memmove(&pPtr[p->iSzPoslist + nByte], &pPtr[p->iSzPoslist + 1], nSz); | |
| 197 sqlite3Fts5PutVarint(&pPtr[p->iSzPoslist], nPos); | |
| 198 p->nData += (nByte-1); | |
| 199 } | |
| 200 p->bDel = 0; | |
| 201 p->iSzPoslist = 0; | |
| 202 } | |
| 203 } | |
| 204 | |
| 205 int sqlite3Fts5HashWrite( | |
| 206 Fts5Hash *pHash, | |
| 207 i64 iRowid, /* Rowid for this entry */ | |
| 208 int iCol, /* Column token appears in (-ve -> delete) */ | |
| 209 int iPos, /* Position of token within column */ | |
| 210 char bByte, /* First byte of token */ | |
| 211 const char *pToken, int nToken /* Token to add or remove to or from index */ | |
| 212 ){ | |
| 213 unsigned int iHash; | |
| 214 Fts5HashEntry *p; | |
| 215 u8 *pPtr; | |
| 216 int nIncr = 0; /* Amount to increment (*pHash->pnByte) by */ | |
| 217 | |
| 218 /* Attempt to locate an existing hash entry */ | |
| 219 iHash = fts5HashKey2(pHash->nSlot, (u8)bByte, (const u8*)pToken, nToken); | |
| 220 for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){ | |
| 221 if( p->zKey[0]==bByte | |
| 222 && memcmp(&p->zKey[1], pToken, nToken)==0 | |
| 223 && p->zKey[nToken+1]==0 | |
| 224 ){ | |
| 225 break; | |
| 226 } | |
| 227 } | |
| 228 | |
| 229 /* If an existing hash entry cannot be found, create a new one. */ | |
| 230 if( p==0 ){ | |
| 231 int nByte = FTS5_HASHENTRYSIZE + (nToken+1) + 1 + 64; | |
| 232 if( nByte<128 ) nByte = 128; | |
| 233 | |
| 234 if( (pHash->nEntry*2)>=pHash->nSlot ){ | |
| 235 int rc = fts5HashResize(pHash); | |
| 236 if( rc!=SQLITE_OK ) return rc; | |
| 237 iHash = fts5HashKey2(pHash->nSlot, (u8)bByte, (const u8*)pToken, nToken); | |
| 238 } | |
| 239 | |
| 240 p = (Fts5HashEntry*)sqlite3_malloc(nByte); | |
| 241 if( !p ) return SQLITE_NOMEM; | |
| 242 memset(p, 0, FTS5_HASHENTRYSIZE); | |
| 243 p->nAlloc = nByte; | |
| 244 p->zKey[0] = bByte; | |
| 245 memcpy(&p->zKey[1], pToken, nToken); | |
| 246 assert( iHash==fts5HashKey(pHash->nSlot, (u8*)p->zKey, nToken+1) ); | |
| 247 p->zKey[nToken+1] = '\0'; | |
| 248 p->nData = nToken+1 + 1 + FTS5_HASHENTRYSIZE; | |
| 249 p->nData += sqlite3Fts5PutVarint(&((u8*)p)[p->nData], iRowid); | |
| 250 p->iSzPoslist = p->nData; | |
| 251 p->nData += 1; | |
| 252 p->iRowid = iRowid; | |
| 253 p->pHashNext = pHash->aSlot[iHash]; | |
| 254 pHash->aSlot[iHash] = p; | |
| 255 pHash->nEntry++; | |
| 256 nIncr += p->nData; | |
| 257 } | |
| 258 | |
| 259 /* Check there is enough space to append a new entry. Worst case scenario | |
| 260 ** is: | |
| 261 ** | |
| 262 ** + 9 bytes for a new rowid, | |
| 263 ** + 4 byte reserved for the "poslist size" varint. | |
| 264 ** + 1 byte for a "new column" byte, | |
| 265 ** + 3 bytes for a new column number (16-bit max) as a varint, | |
| 266 ** + 5 bytes for the new position offset (32-bit max). | |
| 267 */ | |
| 268 if( (p->nAlloc - p->nData) < (9 + 4 + 1 + 3 + 5) ){ | |
| 269 int nNew = p->nAlloc * 2; | |
| 270 Fts5HashEntry *pNew; | |
| 271 Fts5HashEntry **pp; | |
| 272 pNew = (Fts5HashEntry*)sqlite3_realloc(p, nNew); | |
| 273 if( pNew==0 ) return SQLITE_NOMEM; | |
| 274 pNew->nAlloc = nNew; | |
| 275 for(pp=&pHash->aSlot[iHash]; *pp!=p; pp=&(*pp)->pHashNext); | |
| 276 *pp = pNew; | |
| 277 p = pNew; | |
| 278 } | |
| 279 pPtr = (u8*)p; | |
| 280 nIncr -= p->nData; | |
| 281 | |
| 282 /* If this is a new rowid, append the 4-byte size field for the previous | |
| 283 ** entry, and the new rowid for this entry. */ | |
| 284 if( iRowid!=p->iRowid ){ | |
| 285 fts5HashAddPoslistSize(p); | |
| 286 p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iRowid - p->iRowid); | |
| 287 p->iSzPoslist = p->nData; | |
| 288 p->nData += 1; | |
| 289 p->iCol = 0; | |
| 290 p->iPos = 0; | |
| 291 p->iRowid = iRowid; | |
| 292 } | |
| 293 | |
| 294 if( iCol>=0 ){ | |
| 295 /* Append a new column value, if necessary */ | |
| 296 assert( iCol>=p->iCol ); | |
| 297 if( iCol!=p->iCol ){ | |
| 298 pPtr[p->nData++] = 0x01; | |
| 299 p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iCol); | |
| 300 p->iCol = iCol; | |
| 301 p->iPos = 0; | |
| 302 } | |
| 303 | |
| 304 /* Append the new position offset */ | |
| 305 p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iPos - p->iPos + 2); | |
| 306 p->iPos = iPos; | |
| 307 }else{ | |
| 308 /* This is a delete. Set the delete flag. */ | |
| 309 p->bDel = 1; | |
| 310 } | |
| 311 nIncr += p->nData; | |
| 312 | |
| 313 *pHash->pnByte += nIncr; | |
| 314 return SQLITE_OK; | |
| 315 } | |
| 316 | |
| 317 | |
| 318 /* | |
| 319 ** Arguments pLeft and pRight point to linked-lists of hash-entry objects, | |
| 320 ** each sorted in key order. This function merges the two lists into a | |
| 321 ** single list and returns a pointer to its first element. | |
| 322 */ | |
| 323 static Fts5HashEntry *fts5HashEntryMerge( | |
| 324 Fts5HashEntry *pLeft, | |
| 325 Fts5HashEntry *pRight | |
| 326 ){ | |
| 327 Fts5HashEntry *p1 = pLeft; | |
| 328 Fts5HashEntry *p2 = pRight; | |
| 329 Fts5HashEntry *pRet = 0; | |
| 330 Fts5HashEntry **ppOut = &pRet; | |
| 331 | |
| 332 while( p1 || p2 ){ | |
| 333 if( p1==0 ){ | |
| 334 *ppOut = p2; | |
| 335 p2 = 0; | |
| 336 }else if( p2==0 ){ | |
| 337 *ppOut = p1; | |
| 338 p1 = 0; | |
| 339 }else{ | |
| 340 int i = 0; | |
| 341 while( p1->zKey[i]==p2->zKey[i] ) i++; | |
| 342 | |
| 343 if( ((u8)p1->zKey[i])>((u8)p2->zKey[i]) ){ | |
| 344 /* p2 is smaller */ | |
| 345 *ppOut = p2; | |
| 346 ppOut = &p2->pScanNext; | |
| 347 p2 = p2->pScanNext; | |
| 348 }else{ | |
| 349 /* p1 is smaller */ | |
| 350 *ppOut = p1; | |
| 351 ppOut = &p1->pScanNext; | |
| 352 p1 = p1->pScanNext; | |
| 353 } | |
| 354 *ppOut = 0; | |
| 355 } | |
| 356 } | |
| 357 | |
| 358 return pRet; | |
| 359 } | |
| 360 | |
| 361 /* | |
| 362 ** Extract all tokens from hash table iHash and link them into a list | |
| 363 ** in sorted order. The hash table is cleared before returning. It is | |
| 364 ** the responsibility of the caller to free the elements of the returned | |
| 365 ** list. | |
| 366 */ | |
| 367 static int fts5HashEntrySort( | |
| 368 Fts5Hash *pHash, | |
| 369 const char *pTerm, int nTerm, /* Query prefix, if any */ | |
| 370 Fts5HashEntry **ppSorted | |
| 371 ){ | |
| 372 const int nMergeSlot = 32; | |
| 373 Fts5HashEntry **ap; | |
| 374 Fts5HashEntry *pList; | |
| 375 int iSlot; | |
| 376 int i; | |
| 377 | |
| 378 *ppSorted = 0; | |
| 379 ap = sqlite3_malloc(sizeof(Fts5HashEntry*) * nMergeSlot); | |
| 380 if( !ap ) return SQLITE_NOMEM; | |
| 381 memset(ap, 0, sizeof(Fts5HashEntry*) * nMergeSlot); | |
| 382 | |
| 383 for(iSlot=0; iSlot<pHash->nSlot; iSlot++){ | |
| 384 Fts5HashEntry *pIter; | |
| 385 for(pIter=pHash->aSlot[iSlot]; pIter; pIter=pIter->pHashNext){ | |
| 386 if( pTerm==0 || 0==memcmp(pIter->zKey, pTerm, nTerm) ){ | |
| 387 Fts5HashEntry *pEntry = pIter; | |
| 388 pEntry->pScanNext = 0; | |
| 389 for(i=0; ap[i]; i++){ | |
| 390 pEntry = fts5HashEntryMerge(pEntry, ap[i]); | |
| 391 ap[i] = 0; | |
| 392 } | |
| 393 ap[i] = pEntry; | |
| 394 } | |
| 395 } | |
| 396 } | |
| 397 | |
| 398 pList = 0; | |
| 399 for(i=0; i<nMergeSlot; i++){ | |
| 400 pList = fts5HashEntryMerge(pList, ap[i]); | |
| 401 } | |
| 402 | |
| 403 pHash->nEntry = 0; | |
| 404 sqlite3_free(ap); | |
| 405 *ppSorted = pList; | |
| 406 return SQLITE_OK; | |
| 407 } | |
| 408 | |
| 409 /* | |
| 410 ** Query the hash table for a doclist associated with term pTerm/nTerm. | |
| 411 */ | |
| 412 int sqlite3Fts5HashQuery( | |
| 413 Fts5Hash *pHash, /* Hash table to query */ | |
| 414 const char *pTerm, int nTerm, /* Query term */ | |
| 415 const u8 **ppDoclist, /* OUT: Pointer to doclist for pTerm */ | |
| 416 int *pnDoclist /* OUT: Size of doclist in bytes */ | |
| 417 ){ | |
| 418 unsigned int iHash = fts5HashKey(pHash->nSlot, (const u8*)pTerm, nTerm); | |
| 419 Fts5HashEntry *p; | |
| 420 | |
| 421 for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){ | |
| 422 if( memcmp(p->zKey, pTerm, nTerm)==0 && p->zKey[nTerm]==0 ) break; | |
| 423 } | |
| 424 | |
| 425 if( p ){ | |
| 426 fts5HashAddPoslistSize(p); | |
| 427 *ppDoclist = (const u8*)&p->zKey[nTerm+1]; | |
| 428 *pnDoclist = p->nData - (FTS5_HASHENTRYSIZE + nTerm + 1); | |
| 429 }else{ | |
| 430 *ppDoclist = 0; | |
| 431 *pnDoclist = 0; | |
| 432 } | |
| 433 | |
| 434 return SQLITE_OK; | |
| 435 } | |
| 436 | |
| 437 int sqlite3Fts5HashScanInit( | |
| 438 Fts5Hash *p, /* Hash table to query */ | |
| 439 const char *pTerm, int nTerm /* Query prefix */ | |
| 440 ){ | |
| 441 return fts5HashEntrySort(p, pTerm, nTerm, &p->pScan); | |
| 442 } | |
| 443 | |
| 444 void sqlite3Fts5HashScanNext(Fts5Hash *p){ | |
| 445 assert( !sqlite3Fts5HashScanEof(p) ); | |
| 446 p->pScan = p->pScan->pScanNext; | |
| 447 } | |
| 448 | |
| 449 int sqlite3Fts5HashScanEof(Fts5Hash *p){ | |
| 450 return (p->pScan==0); | |
| 451 } | |
| 452 | |
| 453 void sqlite3Fts5HashScanEntry( | |
| 454 Fts5Hash *pHash, | |
| 455 const char **pzTerm, /* OUT: term (nul-terminated) */ | |
| 456 const u8 **ppDoclist, /* OUT: pointer to doclist */ | |
| 457 int *pnDoclist /* OUT: size of doclist in bytes */ | |
| 458 ){ | |
| 459 Fts5HashEntry *p; | |
| 460 if( (p = pHash->pScan) ){ | |
| 461 int nTerm = (int)strlen(p->zKey); | |
| 462 fts5HashAddPoslistSize(p); | |
| 463 *pzTerm = p->zKey; | |
| 464 *ppDoclist = (const u8*)&p->zKey[nTerm+1]; | |
| 465 *pnDoclist = p->nData - (FTS5_HASHENTRYSIZE + nTerm + 1); | |
| 466 }else{ | |
| 467 *pzTerm = 0; | |
| 468 *ppDoclist = 0; | |
| 469 *pnDoclist = 0; | |
| 470 } | |
| 471 } | |
| 472 | |
| OLD | NEW |