/ Check-in [c3e52119]
Login
Overview
Comment::-) (CVS 216)
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA1:c3e521190f02120a34f1e9244fe1ea3a975a6caa
User & Date: drh 2001-05-11 11:02:47
Context
2001-05-15
00:39
:-) (CVS 217) check-in: ee6760fb user: drh tags: trunk
2001-05-11
11:02
:-) (CVS 216) check-in: c3e52119 user: drh tags: trunk
2001-04-29
23:32
:-) (CVS 215) check-in: 624ccbca user: drh tags: trunk
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to notes/notes2.txt.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28


29
30
31
32
33
34
35
36
37
38
39

40
41

42
43



























44
45




46


47
48




























How to do a B*Tree insert:

add_to_page(cursor, data, ptr){
  if( data_fits_on_page ){ add data to page; return; }
  if( page==root ){
     split currentpage+(data+ptr) into lowerpart, center, upperpart
     newpage1 = lowerpart;
     newpage2 = upperpart;
     page = ptr(newpage1) + center + ptr(newpage2);
     return;
  }
  if( move_some_data_left || move_some_data_right ){
    add data to page
    return
  }
  split currentpage+(data+ptr) into lowerpart, center, upperpart
  newpage = upperpart
  currentpage = lowerpart
  pop cursor one level
  add_to_page(cursor, center, ptr(newpage));
}

unlink_entry(cursor, olddata){
  if( !is_a_leaf ){
    n = next_entry()
    if( n fits pageof(cursor) ){
      if( olddata!=nil ) copy dataof(cursor) into olddata
      copy dataof(n) into dataof(cursor)


      unlink_entry(n, nil)
      return
    }
    n = prev_entry()
    if( n fits pageof(cursor) ){
      if( olddata!=nil ) copy dataof(cursor) into olddata
      copy dataof(n) into dataof(cursor)
      unlink_entry(n, nil)
      return
    }
    unlink_entry(n, leafdata)

    move cursor data and ptr into olddata, oldptr
    add_to_page(cursor, leafdata, oldptr)

    return
  }



























  move cursor data into olddata
  if( !underfull(pageof(cursor)) ) return








}






























|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
>
>



|
|
|
<
|


|
>
|
|
>


>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
<
>
>
>
>
|
>
>
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36

37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74

75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
How to do a B*Tree insert:

add_to_page(pageptr, data, pgno){
  pgno.parent = pageptr
  if( data+pgno fits on pageptr ){
    add data+pgno to pageptr
    return
  }
  if( pageptr==root ){
    split pageptr+(data+pgno) into newpage1, center, newpage2
    pageptr = ptr(newpage1) + center + ptr(newpage2);
    return
  }
  if( move_some_data_left || move_some_data_right ){
    add data+pgno to pageptr
    return
  }
  split pageptr+(data+pgno) into pageptr, center, newpage
  add_to_page(parent(pageptr), center, ptr(newpage));
  newpage.parent = parent(pageptr)
}

Cursor: pageptr, idx

unlink_entry(cursor, olddata){
  if( cursor.pageptr is not a leaf page ){
    if( olddata!=nil) copy payload(cursor) into olddata
    n = next_entry(cursor)
    if( payloadsize(n) <= freesize(cursor) + payloadsize(cursor) ){
      copy payload(n) into payload(cursor)
      unlink_entry(n, nil)
      return
    }
    p = prev_entry(cursor)
    if( payloadsize(p) <= freesize(cursor) + payloadsize(cursor) ){
      copy payload(p) into payload(cursor)

      unlink_entry(p, nil)
      return
    }
    unlink(n, leafdata)
    pageptr = cursor.pageptr
    nextpgno = pageptr.aCell[cursor.idx].pgno;
    convert_cursor_to_free_block(cursor)
    add_to_page(pageptr, leafdata, nextpgno)
    return
  }
  pageptr = cursor.pageptr;
  convert_cursor_to_free_block(cursor)
  if( usage(pageptr)<0.65 ){
    consolidate(pageptr)
  }
}

consolidate(pageptr){
  parentpage = parentof(pageptr)
  idx = index_of_page(parentpage, pageptr);
  leftsibling = parentpage.cell[idx].pgno;
  rightsibling = parentpage.cell[idx+1].pgno;
  if( idx>0 ){
    cursor = makecursor(pageptr,idx-1)
    if( try_to_move_down(cursor) ) return
  }
  if( idx<max ){
    cursor = makecursor(pageptr,idx)
    try_to_move_down(cursor)
  }
  return
}

try_to_move_down(cursor){
  pageptr = cursor.pageptr
  if( payload(cursor)+sizeof(left)+sizeof(right)<=pagesize ){
    put cursor and content of left into right
    remove cursor from pageptr

    if( pageptr is root ){
      if( cellcount(pageptr)==0 ){
        copy child into pageptr
        update parent field of child
      }
    }else if( usage(pageptr)<0.65 ){
      try_to_move_down(cursor)
    }
  }
}

cursor_move_next(cursor){
  if( cursor.incr_noop ){
    cursor.incr_noop = FALSE;
    return;
  }
  if( is_leaf(cursor.pageptr) ){
    if( cursor.idx==cursor.pageptr.ncell ){
      if( cursor.pageptr==root ){
        nil cursor
        return
      }
      cursor_move_up(cursor)
      cursor_move_next(cursor)
    }else{
      cursor.idx++;
    }
    return
  }
  pgno = next_pgno(cursor)
  loop {
    cursor.pageptr = get(pgno);
    if( is_leaf(cursor.pageptr) ) break;
    pgno = first_pgno(pageptr);
  }
  cursor.idx = 0;
}

Changes to src/btree.c.

17
18
19
20
21
22
23
24
25
26
27
28
29
30
31









32
33
34
35
36
37
38
39
40
41
42
43






44
45
46
47
48
49
50
..
53
54
55
56
57
58
59
















60
61
62
63
64
65
66
..
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
...
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
...
129
130
131
132
133
134
135











136
137
138
139
140
141
142
...
159
160
161
162
163
164
165

166
167



168
169
170
171

172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214

215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240

241
242
243
244
245
246
247
248
249
** Boston, MA  02111-1307, USA.
**
** Author contact information:
**   drh@hwaci.com
**   http://www.hwaci.com/drh/
**
*************************************************************************
** $Id: btree.c,v 1.3 2001/04/29 23:32:56 drh Exp $
*/
#include "sqliteInt.h"
#include "pager.h"
#include "btree.h"
#include <assert.h>

typedef unsigned int u32;











/*
** The first page contains the following additional information:
**
**      MAGIC-1
**      MAGIC-2
**      First free block
*/
#define EXTRA_PAGE_1_CELLS  3
#define MAGIC_1  0x7264dc61
#define MAGIC_2  0x54e55d9e







/*
** Each database page has a header as follows:
**
**      page1_header          Extra numbers found on page 1 only.
**      leftmost_pgno         Page number of the leftmost child
**      first_cell            Index into MemPage.aPage of first cell
................................................................................
** MemPage.pStart always points to the leftmost_pgno.  First_free is
** 0 if there is no free space on this page.  Otherwise it points to
** an area like this:
**
**      nByte                 Number of free bytes in this block
**      next_free             Next free block or 0 if this is the end
*/

















/*
** The maximum number of database entries that can be held in a single
** page of the database.  Each entry has a 16-byte header consisting of
** 4 unsigned 32-bit numbers, as follows:
**
**       nKey       Number of byte in the key
................................................................................
**       next       index in MemPage.aPage[] of the next entry in sorted order
**
** The key and data follow this header.  The key and data are packed together
** and the total rounded up to the next multiple of 4 bytes.  There must
** be at least 4 bytes in the key/data packet, so each entry consumes at
** least 20 bytes of space on the page.
*/
#define MX_CELL ((SQLITE_PAGE_SIZE-12)/20)

/*
** The maximum amount of data (in bytes) that can be stored locally for a
** database entry.  If the entry contains more data than this, the
** extra goes onto overflow pages.
*/
#define MX_LOCAL_PAYLOAD ((SQLITE_PAGE_SIZE-20-4*24)/4)
................................................................................
  unsigned char validUp;                    /* True if MemPage.up is valid */
  unsigned char validLeft;                  /* True if MemPage.left is valid */
  unsigned char validRight;                 /* True if MemPage.right is valid */
  Pgno up;                     /* The parent page.  0 means this is the root */
  Pgno left;                   /* Left sibling page.  0==none */
  Pgno right;                  /* Right sibling page.  0==none */
  int idxStart;                /* Index in aPage[] of real data */
  int nFree;                   /* Number of free elements of aPage[] */
  int nCell;                   /* Number of entries on this page */
  u32 *aCell[MX_CELL];         /* All entires in sorted order */
}
typedef struct MemPage;

/*
** The in-memory image of a disk page has the auxiliary information appended
................................................................................
  Pager *pPager;        /* The page cache */
  BtCursor *pCursor;    /* All open cursors */
  MemPage *page1;       /* First page of the database */
  int inTrans;          /* True if a transaction is current */
};
typedef Btree Bt;













/*
** The maximum depth of a cursor
*/
#define MX_LEVEL 20

/*
................................................................................
  BtCursor *pPrev, *pNext;      /* Linked list of all cursors */
  int valid;                    /* True if the cursor points to something */
  int nLevel;                   /* Number of levels of indexing used */
  BtIdxpt *pLevel;              /* Pointer to aLevel[nLevel] */
  BtIdxpt aLevel[MX_LEVEL];     /* The index levels */
};


/*
** Mark a section of the memory block as in-use.



*/
static void useSpace(MemPage *pPage, int start, int size){
  int i;
  FreeBlk *p;


  /* Some basic sanity checking */
  assert( pPage && pPage->isInit );
  assert( pPage->nFree>0 && pPage->nFree<=MX_FREE );
  assert( pPage->nFreeSlot >= size );
  assert( start > pPage->idxStart );
  assert( size>0 );
  assert( start + size < SQLITE_PAGE_SIZE/sizeof(pPage->aPage[0]) );

  /* Search for the freeblock that describes the space to be used */
  for(i=0; i<pPage->nFree; i++){
    p = &pPage->aFree[i]
    if( p->idx<=start && p->idx+p->size>start ) break;
  }

  /* The freeblock must contain all the space that is to be used */
  assert( i<pPage->nFree );
  assert( p->idx+p->size >= start+size );

  /* Remove the used space from the freeblock */
  if( p->idx==start ){
    /* The space is at the beginning of the block
    p->size -= size;
    if( p->size==0 ){
      *p = pPage->aFree[pPage->nFree-1];
      pPage->nFree--;
    }
  }else if( p->idx+p->size==start+size ){
    /* Space at the end of the block */
    p->size -= size;
  }else{
    /* Space in the middle of the freeblock. */
    FreeBlk *pNew;
    assert( p->nFreeSlot < MX_FREE );
    pNew->idx = start+size;
    pNew->size = p->idx+p->size - pNew->idx;
    p->size = start - p->idx;
  }
  pPage->nFreeSlot -= size;
}

/*
** Return a section of the MemPage.aPage[] to the freelist.

*/
static void freeSpace(MemPage *pPage, int start, int size){
  int end = start+size;
  int i;
  FreeBlk *pMatch = 0;
  FreeBlk *
  for(i=0; i<pPage->nFreeSlot; i++){
    FreeBlk *p = &pPage->aFree[i];
    if( p->idx==end+1 ){
      if( pMatch ){
        
      }else{
        p->idx = start;
        p->size += size;
        pMatch = p;
      }
    }
    if( p->idx+p->size+1==start ){
      p->size += size;
      break;
    }
  }
}

/*
** Defragment the freespace

*/
static void defragmentSpace(MemPage *pPage){
}

/*
** Initialize the auxiliary information for a disk block.
*/
static int initPage(MemPage *pPage, Pgno pgnoThis, Pgno pgnoParent){
  u32 idx;







|







>
>
>
>
>
>
>
>
>












>
>
>
>
>
>







 







>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







 







|







 







|







 







>
>
>
>
>
>
>
>
>
>
>







 







>

<
>
>
>

<
<
<
>
|
<
<
<
<
<
<
<

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<

<
>

|
<
<
<
<
<
<
<
<
|
<
<
<
<
|
<
<
<
<
<
<
<
<

<
>

|







17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
..
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
...
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
...
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
...
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
...
201
202
203
204
205
206
207
208
209

210
211
212
213



214
215







216
































217

218
219
220








221




222








223

224
225
226
227
228
229
230
231
232
233
** Boston, MA  02111-1307, USA.
**
** Author contact information:
**   drh@hwaci.com
**   http://www.hwaci.com/drh/
**
*************************************************************************
** $Id: btree.c,v 1.4 2001/05/11 11:02:47 drh Exp $
*/
#include "sqliteInt.h"
#include "pager.h"
#include "btree.h"
#include <assert.h>

typedef unsigned int u32;
typedef unsigned short int u16;

/*
** Forward declarations of structures used only in this file.
*/
typedef struct Page1Header Page1Header;
typedef struct PageHdr PageHdr;
typedef struct Cell Cell;
typedef struct FreeBlk FreeBlk;


/*
** The first page contains the following additional information:
**
**      MAGIC-1
**      MAGIC-2
**      First free block
*/
#define EXTRA_PAGE_1_CELLS  3
#define MAGIC_1  0x7264dc61
#define MAGIC_2  0x54e55d9e

struct Page1Header {
  u32 magic1;
  u32 magic2;
  Pgno firstList;
};

/*
** Each database page has a header as follows:
**
**      page1_header          Extra numbers found on page 1 only.
**      leftmost_pgno         Page number of the leftmost child
**      first_cell            Index into MemPage.aPage of first cell
................................................................................
** MemPage.pStart always points to the leftmost_pgno.  First_free is
** 0 if there is no free space on this page.  Otherwise it points to
** an area like this:
**
**      nByte                 Number of free bytes in this block
**      next_free             Next free block or 0 if this is the end
*/
struct PageHdr {
  Pgno pgno;      /* Child page that comes after all cells on this page */
  u16 firstCell;  /* Index in MemPage.aPage[] of the first cell */
  u16 firstFree;  /* Index in MemPage.aPage[] of the first free block */
};
struct Cell {
  Pgno pgno;      /* Child page that comes before this cell */
  u16 nKey;       /* Number of bytes in the key */
  u16 iNext;      /* Index in MemPage.aPage[] of next cell in sorted order */
  u32 nData;      /* Number of bytes of data */
  char aData[4];  /* Key and data */
};
struct FreeBlk {
  u16 iSize;      /* Number of u32-sized slots in the block of free space */
  u16 iNext;      /* Index in MemPage.aPage[] of the next free block */
};

/*
** The maximum number of database entries that can be held in a single
** page of the database.  Each entry has a 16-byte header consisting of
** 4 unsigned 32-bit numbers, as follows:
**
**       nKey       Number of byte in the key
................................................................................
**       next       index in MemPage.aPage[] of the next entry in sorted order
**
** The key and data follow this header.  The key and data are packed together
** and the total rounded up to the next multiple of 4 bytes.  There must
** be at least 4 bytes in the key/data packet, so each entry consumes at
** least 20 bytes of space on the page.
*/
#define MX_CELL ((SQLITE_PAGE_SIZE-sizeof(PageHdr))/sizeof(Cell))

/*
** The maximum amount of data (in bytes) that can be stored locally for a
** database entry.  If the entry contains more data than this, the
** extra goes onto overflow pages.
*/
#define MX_LOCAL_PAYLOAD ((SQLITE_PAGE_SIZE-20-4*24)/4)
................................................................................
  unsigned char validUp;                    /* True if MemPage.up is valid */
  unsigned char validLeft;                  /* True if MemPage.left is valid */
  unsigned char validRight;                 /* True if MemPage.right is valid */
  Pgno up;                     /* The parent page.  0 means this is the root */
  Pgno left;                   /* Left sibling page.  0==none */
  Pgno right;                  /* Right sibling page.  0==none */
  int idxStart;                /* Index in aPage[] of real data */
  int nFree;                   /* Number of free slots of aPage[] */
  int nCell;                   /* Number of entries on this page */
  u32 *aCell[MX_CELL];         /* All entires in sorted order */
}
typedef struct MemPage;

/*
** The in-memory image of a disk page has the auxiliary information appended
................................................................................
  Pager *pPager;        /* The page cache */
  BtCursor *pCursor;    /* All open cursors */
  MemPage *page1;       /* First page of the database */
  int inTrans;          /* True if a transaction is current */
};
typedef Btree Bt;

/*
** A cursor is a pointer to a particular entry in the BTree.
** The entry is identified by its MemPage and the index in
** MemPage.aCell[] of the entry.
*/
struct Cursor {
  Btree *pBt;           /* The pointer back to the BTree */
  MemPage *pPage;       /* Page that contains the entry */
  int idx;              /* Index of the entry in pPage->aCell[] */
  int skip_incr;        /* */
};

/*
** The maximum depth of a cursor
*/
#define MX_LEVEL 20

/*
................................................................................
  BtCursor *pPrev, *pNext;      /* Linked list of all cursors */
  int valid;                    /* True if the cursor points to something */
  int nLevel;                   /* Number of levels of indexing used */
  BtIdxpt *pLevel;              /* Pointer to aLevel[nLevel] */
  BtIdxpt aLevel[MX_LEVEL];     /* The index levels */
};


/*

** Defragment the page given.  All of the free space
** is collected into one big block at the end of the
** page.
*/



static void defragmentPage(MemPage *pPage){
}








































/*

** Mark a section of the memory block as in-use.
*/
static void useSpace(MemPage *pPage, int start, int size){








}













/*

** Return a section of the MemPage.aPage[] to the freelist.
*/
static void freeSpace(MemPage *pPage, int start, int size){
}

/*
** Initialize the auxiliary information for a disk block.
*/
static int initPage(MemPage *pPage, Pgno pgnoThis, Pgno pgnoParent){
  u32 idx;