tdbBtree.c 18.5 KB
Newer Older
H
Hongze Cheng 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13
/*
 * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
 *
 * This program is free software: you can use, redistribute, and/or modify
 * it under the terms of the GNU Affero General Public License, version 3
 * or later ("AGPL"), as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.
 *
 * You should have received a copy of the GNU Affero General Public License
 * along with this program. If not, see <http://www.gnu.org/licenses/>.
H
refact  
Hongze Cheng 已提交
14 15
 */

H
Hongze Cheng 已提交
16 17
#include "tdbInt.h"

H
Hongze Cheng 已提交
18 19 20 21 22
#define TDB_BTREE_ROOT 0x1
#define TDB_BTREE_LEAF 0x2

#define TDB_BTREE_PAGE_IS_ROOT(flags) TDB_FLAG_HAS(flags, TDB_BTREE_ROOT)
#define TDB_BTREE_PAGE_IS_LEAF(flags) TDB_FLAG_HAS(flags, TDB_BTREE_LEAF)
H
Hongze Cheng 已提交
23 24 25
#define TDB_BTREE_ASSERT_FLAG(flags)                                                 \
  ASSERT(TDB_FLAG_IS(flags, TDB_BTREE_ROOT) || TDB_FLAG_IS(flags, TDB_BTREE_LEAF) || \
         TDB_FLAG_IS(flags, TDB_BTREE_ROOT | TDB_BTREE_LEAF) || TDB_FLAG_IS(flags, 0))
H
Hongze Cheng 已提交
26

H
Hongze Cheng 已提交
27
struct SBTree {
H
more  
Hongze Cheng 已提交
28 29 30
  SPgno          root;
  int            keyLen;
  int            valLen;
H
Hongze Cheng 已提交
31
  SPager        *pPager;
H
more  
Hongze Cheng 已提交
32
  FKeyComparator kcmpr;
H
more  
Hongze Cheng 已提交
33
  u8             fanout;
H
Hongze Cheng 已提交
34
  int            pageSize;
H
more  
Hongze Cheng 已提交
35 36 37 38
  int            maxLocal;
  int            minLocal;
  int            maxLeaf;
  int            minLeaf;
H
Hongze Cheng 已提交
39 40
};

H
Hongze Cheng 已提交
41 42 43
typedef struct __attribute__((__packed__)) {
  SPgno rChild;
} SBtPageHdr;
H
more  
Hongze Cheng 已提交
44

H
Hongze Cheng 已提交
45 46 47 48 49
typedef struct {
  u16     flags;
  SBTree *pBt;
} SBtreeZeroPageArg;

H
Hongze Cheng 已提交
50
static int tdbBtCursorMoveTo(SBtCursor *pCur, const void *pKey, int kLen, int *pCRst);
H
more  
Hongze Cheng 已提交
51
static int tdbEncodeLength(u8 *pBuf, uint32_t len);
H
more  
Hongze Cheng 已提交
52
static int tdbCompareKeyAndCell(const void *pKey, int kLen, const void *pCell);
H
Hongze Cheng 已提交
53
static int tdbDefaultKeyCmprFn(const void *pKey1, int keyLen1, const void *pKey2, int keyLen2);
H
Hongze Cheng 已提交
54
static int tdbBtreeOpenImpl(SBTree *pBt);
H
Hongze Cheng 已提交
55 56
static int tdbBtreeZeroPage(SPage *pPage, void *arg);
static int tdbBtreeInitPage(SPage *pPage, void *arg);
H
Hongze Cheng 已提交
57

H
Hongze Cheng 已提交
58
int tdbBtreeOpen(int keyLen, int valLen, SPager *pPager, FKeyComparator kcmpr, SBTree **ppBt) {
H
Hongze Cheng 已提交
59
  SBTree *pBt;
H
Hongze Cheng 已提交
60
  int     ret;
H
more  
Hongze Cheng 已提交
61

H
Hongze Cheng 已提交
62
  *ppBt = NULL;
H
Hongze Cheng 已提交
63 64 65 66 67 68 69 70 71 72

  pBt = (SBTree *)calloc(1, sizeof(*pBt));
  if (pBt == NULL) {
    return -1;
  }

  // pBt->keyLen
  pBt->keyLen = keyLen;
  // pBt->valLen
  pBt->valLen = valLen;
H
refact  
Hongze Cheng 已提交
73 74
  // pBt->pPager
  pBt->pPager = pPager;
H
Hongze Cheng 已提交
75 76
  // pBt->kcmpr
  pBt->kcmpr = kcmpr ? kcmpr : tdbDefaultKeyCmprFn;
H
more  
Hongze Cheng 已提交
77 78 79 80 81 82 83
  // pBt->fanout
  if (keyLen == TDB_VARIANT_LEN) {
    pBt->fanout = TDB_DEFAULT_FANOUT;
  } else {
    ASSERT(0);
    // TODO: pBt->fanout = 0;
  }
H
Hongze Cheng 已提交
84 85 86 87
  // pBt->pageSize
  pBt->pageSize = tdbPagerGetPageSize(pPager);
  // pBt->maxLocal
  pBt->maxLocal = (pBt->pageSize - sizeof(SPageHdr)) / pBt->fanout;
H
Hongze Cheng 已提交
88
  // pBt->minLocal: Should not be allowed smaller than 15, which is [nPayload][nKey][nData]
H
Hongze Cheng 已提交
89 90 91 92 93
  pBt->minLocal = (pBt->pageSize - sizeof(SPageHdr)) / pBt->fanout / 2;
  // pBt->maxLeaf
  pBt->maxLeaf = pBt->pageSize - sizeof(SPageHdr);
  // pBt->minLeaf
  pBt->minLeaf = pBt->minLocal;
H
Hongze Cheng 已提交
94

H
Hongze Cheng 已提交
95 96 97 98 99 100 101
  // TODO: pBt->root
  ret = tdbBtreeOpenImpl(pBt);
  if (ret < 0) {
    free(pBt);
    return -1;
  }

H
Hongze Cheng 已提交
102
  *ppBt = pBt;
H
Hongze Cheng 已提交
103 104 105 106 107 108 109 110
  return 0;
}

int tdbBtreeClose(SBTree *pBt) {
  // TODO
  return 0;
}

H
more  
Hongze Cheng 已提交
111
int tdbBtreeCursor(SBtCursor *pCur, SBTree *pBt) {
H
more  
Hongze Cheng 已提交
112 113
  pCur->pBt = pBt;
  pCur->iPage = -1;
H
more  
Hongze Cheng 已提交
114
  pCur->pPage = NULL;
H
Hongze Cheng 已提交
115
  pCur->idx = -1;
H
more  
Hongze Cheng 已提交
116

H
more  
Hongze Cheng 已提交
117 118 119
  return 0;
}

H
Hongze Cheng 已提交
120
int tdbBtCursorInsert(SBtCursor *pCur, const void *pKey, int kLen, const void *pVal, int vLen) {
H
Hongze Cheng 已提交
121
  int     ret;
H
Hongze Cheng 已提交
122
  int     idx;
H
Hongze Cheng 已提交
123
  SPager *pPager;
H
Hongze Cheng 已提交
124 125 126
  SCell  *pCell;
  int     szCell;
  int     cret;
H
Hongze Cheng 已提交
127

H
Hongze Cheng 已提交
128
  ret = tdbBtCursorMoveTo(pCur, pKey, kLen, &cret);
H
Hongze Cheng 已提交
129 130 131 132 133
  if (ret < 0) {
    // TODO: handle error
    return -1;
  }

H
Hongze Cheng 已提交
134
  if (pCur->idx == -1) {
H
Hongze Cheng 已提交
135
    ASSERT(TDB_PAGE_NCELLS(pCur->pPage) == 0);
H
Hongze Cheng 已提交
136 137 138 139 140 141 142 143 144 145 146 147
    idx = 0;
  } else {
    if (cret > 0) {
      // TODO
    } else if (cret < 0) {
      // TODO
    } else {
      /* TODO */
      ASSERT(0);
    }
  }

H
Hongze Cheng 已提交
148
  // Encode the cell
H
Hongze Cheng 已提交
149 150 151 152 153

  // Insert the cell to the index
  ret = tdbPageInsertCell(pCur->pPage, idx, pCell, szCell);
  if (ret < 0) {
    return -1;
H
Hongze Cheng 已提交
154 155
  }

H
more  
Hongze Cheng 已提交
156 157 158
  return 0;
}

H
Hongze Cheng 已提交
159
static int tdbBtCursorMoveTo(SBtCursor *pCur, const void *pKey, int kLen, int *pCRst) {
H
Hongze Cheng 已提交
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179
  int     ret;
  SBTree *pBt;
  SPager *pPager;

  pBt = pCur->pBt;
  pPager = pBt->pPager;

  if (pCur->iPage < 0) {
    ASSERT(pCur->iPage == -1);
    ASSERT(pCur->idx == -1);

    // Move from the root
    ret = tdbPagerFetchPage(pPager, pBt->root, &(pCur->pPage), tdbBtreeInitPage, pBt);
    if (ret < 0) {
      ASSERT(0);
      return -1;
    }

    pCur->iPage = 0;

H
Hongze Cheng 已提交
180
    if (TDB_PAGE_NCELLS(pCur->pPage) == 0) {
H
Hongze Cheng 已提交
181
      // Current page is empty
H
Hongze Cheng 已提交
182
      ASSERT(TDB_FLAG_IS(TDB_PAGE_FLAGS(pCur->pPage), TDB_BTREE_ROOT | TDB_BTREE_LEAF));
H
Hongze Cheng 已提交
183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
      return 0;
    }

    // Search from root page down to leaf
    {
      // TODO
      ASSERT(0);
      // ret = tdbBtCursorMoveToRoot(pCur);
      // if (ret < 0) {
      //   return -1;
      // }

      // if (pCur->pPage->pHdr->nCells == 0) {
      //   // Tree is empty
      // } else {
      //   for (;;) {
      //     int lidx, ridx, midx, c;

      //     pBtPage = pCur->pPage;
      //     lidx = 0;
      //     ridx = pBtPage->pHdr->nCells - 1;
      //     while (lidx <= ridx) {
      //       midx = (lidx + ridx) >> 1;
H
Hongze Cheng 已提交
206
      //       pCell = (void *)(pBtPage->aData + pBtPage->pCellIdx[midx]);
H
Hongze Cheng 已提交
207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225

      //       c = tdbCompareKeyAndCell(pKey, kLen, pCell);
      //       if (c == 0) {
      //         break;
      //       } else if (c < 0) {
      //         lidx = lidx + 1;
      //       } else {
      //         ridx = ridx - 1;
      //       }
      //     }
      //   }

      //   /* code */
      // }
    }

  } else {
    // TODO: Move the cursor from a some position instead of a clear state
  }
H
more  
Hongze Cheng 已提交
226

H
Hongze Cheng 已提交
227 228
  return 0;
}
H
more  
Hongze Cheng 已提交
229

H
more  
Hongze Cheng 已提交
230
static int tdbEncodeLength(u8 *pBuf, uint32_t len) {
H
Hongze Cheng 已提交
231 232 233 234 235 236 237 238 239 240
  int iCount = 0;

  while (len > 127) {
    pBuf[iCount++] = (u8)((len & 0xff) | 128);
    len >>= 7;
  }

  pBuf[iCount++] = (u8)len;

  return iCount;
H
more  
Hongze Cheng 已提交
241 242 243
}

static int tdbBtCursorMoveToRoot(SBtCursor *pCur) {
H
refact  
Hongze Cheng 已提交
244 245
  SBTree *pBt;
  SPager *pPager;
H
Hongze Cheng 已提交
246
  SPage  *pPage;
H
refact  
Hongze Cheng 已提交
247
  int     ret;
H
more  
Hongze Cheng 已提交
248 249

  pBt = pCur->pBt;
H
refact  
Hongze Cheng 已提交
250
  pPager = pBt->pPager;
H
more  
Hongze Cheng 已提交
251

H
Hongze Cheng 已提交
252 253 254 255
  // pPage = tdbPagerGet(pPager, pBt->root, true);
  // if (pPage == NULL) {
  //   // TODO: handle error
  // }
H
more  
Hongze Cheng 已提交
256

H
Hongze Cheng 已提交
257 258 259 260 261
  // ret = tdbInitBtPage(pPage, &pBtPage);
  // if (ret < 0) {
  //   // TODO
  //   return 0;
  // }
H
more  
Hongze Cheng 已提交
262

H
Hongze Cheng 已提交
263 264
  // pCur->pPage = pBtPage;
  // pCur->iPage = 0;
H
more  
Hongze Cheng 已提交
265 266 267 268

  return 0;
}

H
more  
Hongze Cheng 已提交
269 270
static int tdbCompareKeyAndCell(const void *pKey, int kLen, const void *pCell) {
  /* TODO */
H
more  
Hongze Cheng 已提交
271
  return 0;
H
Hongze Cheng 已提交
272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291
}

static int tdbDefaultKeyCmprFn(const void *pKey1, int keyLen1, const void *pKey2, int keyLen2) {
  int mlen;
  int cret;

  ASSERT(keyLen1 > 0 && keyLen2 > 0 && pKey1 != NULL && pKey2 != NULL);

  mlen = keyLen1 < keyLen2 ? keyLen1 : keyLen2;
  cret = memcmp(pKey1, pKey2, mlen);
  if (cret == 0) {
    if (keyLen1 < keyLen2) {
      cret = -1;
    } else if (keyLen1 > keyLen2) {
      cret = 1;
    } else {
      cret = 0;
    }
  }
  return cret;
H
Hongze Cheng 已提交
292 293 294 295 296 297 298 299 300 301
}

static int tdbBtreeOpenImpl(SBTree *pBt) {
  // Try to get the root page of the an existing btree

  SPgno  pgno;
  SPage *pPage;
  int    ret;

  {
H
Hongze Cheng 已提交
302
    // 1. TODO: Search the main DB to check if the DB exists
H
Hongze Cheng 已提交
303 304 305 306 307 308 309 310 311
    pgno = 0;
  }

  if (pgno != 0) {
    pBt->root = pgno;
    return 0;
  }

  // Try to create a new database
H
Hongze Cheng 已提交
312 313 314 315 316 317 318
  SBtreeZeroPageArg zArg = {.flags = TDB_BTREE_ROOT | TDB_BTREE_LEAF, .pBt = pBt};
  ret = tdbPagerNewPage(pBt->pPager, &pgno, &pPage, tdbBtreeZeroPage, &zArg);
  if (ret < 0) {
    return -1;
  }

  // TODO: Unref the page
H
Hongze Cheng 已提交
319

H
Hongze Cheng 已提交
320 321
  ASSERT(pgno != 0);
  pBt->root = pgno;
H
Hongze Cheng 已提交
322

H
Hongze Cheng 已提交
323 324 325
  return 0;
}

H
Hongze Cheng 已提交
326
static int tdbBtreeInitPage(SPage *pPage, void *arg) {
H
Hongze Cheng 已提交
327
  SBTree *pBt;
H
Hongze Cheng 已提交
328
  u16     flags;
H
Hongze Cheng 已提交
329

H
Hongze Cheng 已提交
330
  pBt = (SBTree *)arg;
H
Hongze Cheng 已提交
331

H
Hongze Cheng 已提交
332 333 334 335 336 337
  flags = TDB_PAGE_FLAGS(pPage);
  if (TDB_BTREE_PAGE_IS_LEAF(flags)) {
    pPage->szAmHdr = 0;
  } else {
    pPage->szAmHdr = sizeof(SBtPageHdr);
  }
H
Hongze Cheng 已提交
338
  pPage->pPageHdr = pPage->pData;
H
Hongze Cheng 已提交
339 340 341 342 343
  pPage->pAmHdr = pPage->pPageHdr + pPage->szPageHdr;
  pPage->pCellIdx = pPage->pAmHdr + pPage->szAmHdr;
  pPage->pFreeStart = pPage->pCellIdx + pPage->szOffset * TDB_PAGE_NCELLS(pPage);
  pPage->pFreeEnd = pPage->pData + TDB_PAGE_CCELLS(pPage);
  pPage->pPageFtr = (SPageFtr *)(pPage->pData + pPage->pageSize - sizeof(SPageFtr));
H
Hongze Cheng 已提交
344

H
Hongze Cheng 已提交
345 346
  TDB_BTREE_ASSERT_FLAG(flags);

H
Hongze Cheng 已提交
347
  // Init other fields
H
Hongze Cheng 已提交
348
  if (TDB_BTREE_PAGE_IS_LEAF(flags)) {
H
Hongze Cheng 已提交
349 350 351 352 353 354 355 356 357 358
    pPage->kLen = pBt->keyLen;
    pPage->vLen = pBt->valLen;
    pPage->maxLocal = pBt->maxLeaf;
    pPage->minLocal = pBt->minLeaf;
  } else {
    pPage->kLen = pBt->keyLen;
    pPage->vLen = sizeof(SPgno);
    pPage->maxLocal = pBt->maxLocal;
    pPage->minLocal = pBt->minLocal;
  }
H
Hongze Cheng 已提交
359 360 361 362

  return 0;
}

H
Hongze Cheng 已提交
363
static int tdbBtreeZeroPage(SPage *pPage, void *arg) {
H
Hongze Cheng 已提交
364
  u16     flags;
H
Hongze Cheng 已提交
365
  SBTree *pBt;
H
Hongze Cheng 已提交
366

H
Hongze Cheng 已提交
367 368
  flags = ((SBtreeZeroPageArg *)arg)->flags;
  pBt = ((SBtreeZeroPageArg *)arg)->pBt;
H
Hongze Cheng 已提交
369

H
Hongze Cheng 已提交
370
  pPage->pPageHdr = pPage->pData;
H
Hongze Cheng 已提交
371

H
Hongze Cheng 已提交
372 373 374 375 376 377
  // Init the page header
  TDB_PAGE_FLAGS_SET(pPage, flags);
  TDB_PAGE_NCELLS_SET(pPage, 0);
  TDB_PAGE_CCELLS_SET(pPage, pBt->pageSize - sizeof(SPageFtr));
  TDB_PAGE_FCELL_SET(pPage, 0);
  TDB_PAGE_NFREE_SET(pPage, 0);
H
Hongze Cheng 已提交
378

H
Hongze Cheng 已提交
379
  tdbBtreeInitPage(pPage, (void *)pBt);
H
Hongze Cheng 已提交
380

H
Hongze Cheng 已提交
381
  return 0;
H
Hongze Cheng 已提交
382 383 384
}

#ifndef TDB_BTREE_BALANCE
H
Hongze Cheng 已提交
385 386 387 388 389 390 391 392 393 394
typedef struct {
  SBTree *pBt;
  SPage  *pParent;
  int     idx;
  i8      nOldPages;
  SPage  *pOldPages[3];
  i8      nNewPages;
  SPage  *pNewPages[5];
} SBtreeBalanceHelper;

H
Hongze Cheng 已提交
395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435
static int tdbBtreeCopyPageContent(SPage *pFrom, SPage *pTo) {
  /* TODO */

  return 0;
}

static int tdbBtreeBalanceDeeper(SBTree *pBt, SPage *pRoot, SPage **ppChild) {
  SPager           *pPager;
  SPage            *pChild;
  SPgno             pgnoChild;
  int               ret;
  SBtreeZeroPageArg zArg;

  pPager = pRoot->pPager;

  // Allocate a new child page
  zArg.flags = TDB_BTREE_LEAF;
  zArg.pBt = pBt;
  ret = tdbPagerNewPage(pPager, &pgnoChild, &pChild, tdbBtreeZeroPage, &zArg);
  if (ret < 0) {
    return -1;
  }

  // Copy the root page content to the child page
  ret = tdbBtreeCopyPageContent(pRoot, pChild);
  if (ret < 0) {
    return -1;
  }

  {
    // TODO: Copy the over flow part of the page
  }

  // Reinitialize the root page
  zArg.flags = TDB_BTREE_ROOT;
  zArg.pBt = pBt;
  ret = tdbBtreeZeroPage(pRoot, &zArg);
  if (ret < 0) {
    return -1;
  }

H
Hongze Cheng 已提交
436 437 438 439
  *ppChild = pChild;
  return 0;
}

H
Hongze Cheng 已提交
440
static int tdbBtreeBalanceStep1(SBtreeBalanceHelper *pBlh) {
H
Hongze Cheng 已提交
441
#if 0
H
Hongze Cheng 已提交
442 443 444 445 446 447 448 449
  int    i;
  SPage *pParent;
  int    nDiv;
  SPgno  pgno;
  SPage *pPage;
  int    ret;

  pParent = pBlh->pParent;
H
Hongze Cheng 已提交
450
  i = pParent->pPageHdr->nCells + pParent->nOverflow;
H
Hongze Cheng 已提交
451

H
Hongze Cheng 已提交
452 453 454
  if (i < 1) {
    nDiv = 0;
  } else {
H
Hongze Cheng 已提交
455
    if (pBlh->idx == 0) {
H
Hongze Cheng 已提交
456
      nDiv = 0;
H
Hongze Cheng 已提交
457
    } else if (pBlh->idx == i) {
H
Hongze Cheng 已提交
458
      nDiv = i - 2;
H
Hongze Cheng 已提交
459
    } else {
H
Hongze Cheng 已提交
460
      nDiv = pBlh->idx - 1;
H
Hongze Cheng 已提交
461
    }
H
Hongze Cheng 已提交
462 463
    i = 2;
  }
H
Hongze Cheng 已提交
464
  pBlh->nOldPages = i + 1;
H
Hongze Cheng 已提交
465

H
Hongze Cheng 已提交
466
  if (i + nDiv - pParent->nOverflow == pParent->pPageHdr->nCells) {
H
Hongze Cheng 已提交
467
    // pgno = pParent->pPageHdr->rChild;
H
Hongze Cheng 已提交
468 469 470 471 472 473
  } else {
    ASSERT(0);
    // TODO
    pgno = 0;
  }
  for (;;) {
H
Hongze Cheng 已提交
474
    ret = tdbPagerFetchPage(pBlh->pBt->pPager, pgno, &pPage, tdbBtreeInitPage, pBlh->pBt);
H
Hongze Cheng 已提交
475
    if (ret < 0) {
H
Hongze Cheng 已提交
476
      ASSERT(0);
H
Hongze Cheng 已提交
477
      return -1;
H
Hongze Cheng 已提交
478
    }
H
Hongze Cheng 已提交
479

H
Hongze Cheng 已提交
480
    pBlh->pOldPages[i] = pPage;
H
Hongze Cheng 已提交
481 482 483 484

    if ((i--) == 0) break;

    if (pParent->nOverflow && i + nDiv == pParent->aiOvfl[0]) {
H
Hongze Cheng 已提交
485
      // pCellDiv[i] = pParent->apOvfl[0];
H
Hongze Cheng 已提交
486 487 488 489 490 491 492 493 494 495 496 497 498
      // pgno = 0;
      // szNew[i] = tdbPageCellSize(pPage, pCell);
      pParent->nOverflow = 0;
    } else {
      // pCellDiv[i] = TDB_PAGE_CELL_AT(pPage, i + nDiv - pParent->nOverflow);
      // pgno = 0;
      // szNew[i] = tdbPageCellSize(pPage, pCell);

      // Drop the cell from the page
      // ret = tdbPageDropCell(pPage, i + nDiv - pParent->nOverflow, szNew[i]);
      // if (ret < 0) {
      //   return -1;
      // }
H
Hongze Cheng 已提交
499 500 501
    }
  }

H
Hongze Cheng 已提交
502
#endif
H
Hongze Cheng 已提交
503 504 505 506
  return 0;
}

static int tdbBtreeBalanceStep2(SBtreeBalanceHelper *pBlh) {
H
Hongze Cheng 已提交
507
#if 0
H
Hongze Cheng 已提交
508
  SPage *pPage;
H
Hongze Cheng 已提交
509 510 511 512
  int    oidx;
  int    cidx;
  int    limit;
  SCell *pCell;
H
Hongze Cheng 已提交
513 514 515

  for (int i = 0; i < pBlh->nOldPages; i++) {
    pPage = pBlh->pOldPages[i];
H
Hongze Cheng 已提交
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544
    oidx = 0;
    cidx = 0;

    if (oidx < pPage->nOverflow) {
      limit = pPage->aiOvfl[oidx];
    } else {
      limit = pPage->pPageHdr->nCells;
    }

    // Loop to copy each cell pointer out
    for (;;) {
      if (oidx >= pPage->nOverflow && cidx >= pPage->pPageHdr->nCells) break;

      if (cidx < limit) {
        // Get local cells
        pCell = TDB_PAGE_CELL_AT(pPage, cidx);
      } else if (cidx == limit) {
        // Get overflow cells
        pCell = pPage->apOvfl[oidx++];

        if (oidx < pPage->nOverflow) {
          limit = pPage->aiOvfl[oidx];
        } else {
          limit = pPage->pPageHdr->nCells;
        }
      } else {
        ASSERT(0);
      }
    }
H
Hongze Cheng 已提交
545 546 547 548

    {
      // TODO: Copy divider cells here
    }
H
Hongze Cheng 已提交
549 550 551 552
  }

  /* TODO */

H
Hongze Cheng 已提交
553
#endif
H
Hongze Cheng 已提交
554 555 556 557
  return 0;
}

static int tdbBtreeBalanceStep3(SBtreeBalanceHelper *pBlh) {
H
Hongze Cheng 已提交
558 559 560 561
  for (int i = 0; i < pBlh->nOldPages; i++) {
    /* code */
  }

H
Hongze Cheng 已提交
562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592
  return 0;
}

static int tdbBtreeBalanceStep4(SBtreeBalanceHelper *pBlh) {
  // TODO
  return 0;
}

static int tdbBtreeBalanceStep5(SBtreeBalanceHelper *pBlh) {
  // TODO
  return 0;
}

static int tdbBtreeBalanceStep6(SBtreeBalanceHelper *pBlh) {
  // TODO
  return 0;
}

static int tdbBtreeBalanceNonRoot(SBTree *pBt, SPage *pParent, int idx) {
  int                 ret;
  SBtreeBalanceHelper blh;

  blh.pBt = pBt;
  blh.pParent = pParent;
  blh.idx = idx;

  // Step 1: find two sibling pages and get engough info about the old pages
  ret = tdbBtreeBalanceStep1(&blh);
  if (ret < 0) {
    ASSERT(0);
    return -1;
H
Hongze Cheng 已提交
593 594
  }

H
Hongze Cheng 已提交
595 596 597 598 599
  // Step 2: Load all cells on the old page and the divider cells
  ret = tdbBtreeBalanceStep2(&blh);
  if (ret < 0) {
    ASSERT(0);
    return -1;
H
Hongze Cheng 已提交
600 601
  }

H
Hongze Cheng 已提交
602 603 604 605 606
  // Step 3: Get the number of pages needed to hold all cells
  ret = tdbBtreeBalanceStep3(&blh);
  if (ret < 0) {
    ASSERT(0);
    return -1;
H
Hongze Cheng 已提交
607 608
  }

H
Hongze Cheng 已提交
609 610 611 612 613
  // Step 4: Allocate enough new pages. Reuse old pages as much as possible
  ret = tdbBtreeBalanceStep4(&blh);
  if (ret < 0) {
    ASSERT(0);
    return -1;
H
Hongze Cheng 已提交
614 615
  }

H
Hongze Cheng 已提交
616 617 618 619 620 621 622 623 624 625 626 627
  // Step 5: Insert new divider cells into pParent
  ret = tdbBtreeBalanceStep5(&blh);
  if (ret < 0) {
    ASSERT(0);
    return -1;
  }

  // Step 6: Update the sibling pages
  ret = tdbBtreeBalanceStep6(&blh);
  if (ret < 0) {
    ASSERT(0);
    return -1;
H
Hongze Cheng 已提交
628 629 630 631 632 633 634 635 636
  }

  {
      // TODO: Reset states
  }

  {
    // TODO: Clear resources
  }
H
Hongze Cheng 已提交
637

H
Hongze Cheng 已提交
638 639 640 641 642
  return 0;
}

static int tdbBtreeBalance(SBtCursor *pCur) {
  int    iPage;
H
Hongze Cheng 已提交
643 644
  SPage *pParent;
  int    ret;
H
Hongze Cheng 已提交
645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661

  // Main loop to balance the BTree
  for (;;) {
    iPage = pCur->iPage;

    // TODO: Get the page free space if not get yet
    // if (pPage->nFree < 0) {
    //   if (tdbBtreeComputeFreeSpace(pPage) < 0) {
    //     return -1;
    //   }
    // }

    if (0 /*TODO: balance is over*/) {
      break;
    }

    if (iPage == 0) {
H
Hongze Cheng 已提交
662 663
      // Balance the root page by copy the root page content to
      // a child page and set the root page as empty first
H
Hongze Cheng 已提交
664
      // ASSERT(TDB_BTREE_PAGE_IS_ROOT(pCur->pPage->pPageHdr->flags));
H
Hongze Cheng 已提交
665 666 667 668 669 670 671 672 673 674 675

      ret = tdbBtreeBalanceDeeper(pCur->pBt, pCur->pPage, &(pCur->pgStack[1]));
      if (ret < 0) {
        return -1;
      }

      pCur->idx = 0;
      pCur->idxStack[0] = 0;
      pCur->pgStack[0] = pCur->pPage;
      pCur->iPage = 1;
      pCur->pPage = pCur->pgStack[1];
H
Hongze Cheng 已提交
676 677

    } else {
H
Hongze Cheng 已提交
678 679 680
      // Generalized balance step
      pParent = pCur->pgStack[pCur->iPage - 1];

H
Hongze Cheng 已提交
681
      ret = tdbBtreeBalanceNonRoot(pCur->pBt, pParent, pCur->idxStack[pCur->iPage - 1]);
H
Hongze Cheng 已提交
682 683 684 685 686 687
      if (ret < 0) {
        return -1;
      }

      pCur->iPage--;
      pCur->pPage = pCur->pgStack[pCur->iPage];
H
Hongze Cheng 已提交
688 689 690 691 692
    }
  }

  return 0;
}
H
Hongze Cheng 已提交
693 694
#endif

H
Hongze Cheng 已提交
695
#ifndef TDB_BTREE_CELL  // =========================================================
H
Hongze Cheng 已提交
696
typedef struct {
H
Hongze Cheng 已提交
697 698 699 700 701 702
  int   kLen;
  u8   *pKey;
  int   vLen;
  u8   *pVal;
  SPgno pgno;
  u8   *pTmpSpace;
H
Hongze Cheng 已提交
703 704
} SCellDecoder;

H
Hongze Cheng 已提交
705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735
static int tdbBtreeEncodePayload(SPage *pPage, u8 *pPayload, void *pKey, int kLen, void *pVal, int vLen,
                                 int *szPayload) {
  int nPayload;

  ASSERT(pKey != NULL);

  if (pVal == NULL) {
    vLen = 0;
  }

  nPayload = kLen + vLen;
  if (nPayload <= pPage->maxLocal) {
    // General case without overflow
    memcpy(pPayload, pKey, kLen);
    if (pVal) {
      memcpy(pPayload + kLen, pVal, vLen);
    }

    *szPayload = nPayload;
    return 0;
  }

  {
    // TODO: handle overflow case
    ASSERT(0);
  }

  return 0;
}

static int tdbBtreeEncodeCell(SPage *pPage, void *pKey, int kLen, void *pVal, int vLen, SCell *pCell, int *szCell) {
H
Hongze Cheng 已提交
736
  u16 flags;
H
Hongze Cheng 已提交
737 738 739 740
  u8  leaf;
  int nHeader;
  int nPayload;
  int ret;
H
Hongze Cheng 已提交
741 742 743 744

  ASSERT(pPage->kLen == TDB_VARIANT_LEN || pPage->kLen == kLen);
  ASSERT(pPage->vLen == TDB_VARIANT_LEN || pPage->vLen == vLen);

H
Hongze Cheng 已提交
745 746 747 748
  nPayload = 0;
  nHeader = 0;
  flags = TDB_PAGE_FLAGS(pPage);
  leaf = TDB_BTREE_PAGE_IS_LEAF(flags);
H
Hongze Cheng 已提交
749

H
Hongze Cheng 已提交
750 751
  // 1. Encode Header part
  /* Encode kLen if need */
H
Hongze Cheng 已提交
752
  if (pPage->kLen == TDB_VARIANT_LEN) {
H
Hongze Cheng 已提交
753
    nHeader += tdbPutVarInt(pCell + nHeader, kLen);
H
Hongze Cheng 已提交
754 755
  }

H
Hongze Cheng 已提交
756
  /* Encode vLen if need */
H
Hongze Cheng 已提交
757
  if (pPage->vLen == TDB_VARIANT_LEN) {
H
Hongze Cheng 已提交
758
    nHeader += tdbPutVarInt(pCell + nHeader, vLen);
H
Hongze Cheng 已提交
759 760
  }

H
Hongze Cheng 已提交
761 762
  /* Encode SPgno if interior page */
  if (!leaf) {
H
Hongze Cheng 已提交
763 764
    ASSERT(pPage->vLen == sizeof(SPgno));

H
Hongze Cheng 已提交
765 766 767 768 769 770 771 772 773 774 775 776 777
    ((SPgno *)(pCell + nHeader))[0] = ((SPgno *)pVal)[0];
    nHeader = nHeader + sizeof(SPgno);
  }

  // 2. Encode payload part
  if (leaf) {
    ret = tdbBtreeEncodePayload(pPage, pCell + nHeader, pKey, kLen, pVal, vLen, &nPayload);
  } else {
    ret = tdbBtreeEncodePayload(pPage, pCell + nHeader, pKey, kLen, NULL, 0, &nPayload);
  }
  if (ret < 0) {
    // TODO: handle error
    return -1;
H
Hongze Cheng 已提交
778 779
  }

H
Hongze Cheng 已提交
780
  *szCell = nHeader + nPayload;
H
Hongze Cheng 已提交
781 782 783
  return 0;
}

H
Hongze Cheng 已提交
784 785 786 787
static int tdbBtreeDecodePayload(SPage *pPage, const u8 *pPayload, SCellDecoder *pDecoder) {
  int nPayload;

  ASSERT(pDecoder->pKey == NULL);
H
Hongze Cheng 已提交
788

H
Hongze Cheng 已提交
789 790
  if (pDecoder->pVal) {
    nPayload = pDecoder->kLen + pDecoder->vLen;
H
Hongze Cheng 已提交
791
  } else {
H
Hongze Cheng 已提交
792
    nPayload = pDecoder->kLen;
H
Hongze Cheng 已提交
793 794
  }

H
Hongze Cheng 已提交
795 796 797 798 799 800 801 802 803 804 805
  if (nPayload <= pPage->maxLocal) {
    // General case without overflow
    pDecoder->pKey = (void *)pPayload;
    if (pDecoder->pVal) {
      pDecoder->pVal = (void *)(pPayload + pDecoder->vLen);
    }
  }

  {
    // TODO: handle overflow case
    ASSERT(0);
H
Hongze Cheng 已提交
806 807
  }

H
Hongze Cheng 已提交
808 809 810 811 812 813 814 815 816 817
  return 0;
}

static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pDecoder) {
  u16 flags;
  u8  leaf;
  int nHeader;
  int ret;

  nHeader = 0;
H
Hongze Cheng 已提交
818
  flags = TDB_PAGE_FLAGS(pPage);
H
Hongze Cheng 已提交
819 820 821 822 823 824 825 826 827 828 829
  leaf = TDB_BTREE_PAGE_IS_LEAF(flags);

  // 1. Decode header part
  if (pPage->kLen == TDB_VARIANT_LEN) {
    nHeader += tdbGetVarInt(pCell + nHeader, &(pDecoder->kLen));
  } else {
    pDecoder->kLen = pPage->kLen;
  }

  if (pPage->vLen == TDB_VARIANT_LEN) {
    nHeader += tdbGetVarInt(pCell + nHeader, &(pDecoder->vLen));
H
Hongze Cheng 已提交
830
  } else {
H
Hongze Cheng 已提交
831 832
    pDecoder->vLen = pPage->vLen;
  }
H
Hongze Cheng 已提交
833

H
Hongze Cheng 已提交
834 835
  if (!leaf) {
    ASSERT(pPage->vLen == sizeof(SPgno));
H
Hongze Cheng 已提交
836

H
Hongze Cheng 已提交
837 838 839 840 841 842 843 844 845
    pDecoder->pgno = ((SPgno *)(pCell + nHeader))[0];
    pDecoder->pVal = (u8 *)(&(pDecoder->pgno));
    nHeader = nHeader + sizeof(SPgno);
  }

  // 2. Decode payload part
  ret = tdbBtreeDecodePayload(pPage, pCell + nHeader, pDecoder);
  if (ret < 0) {
    return -1;
H
Hongze Cheng 已提交
846 847 848 849 850
  }

  return 0;
}

H
Hongze Cheng 已提交
851
#endif