【问题标题】:Fixed length columns size regarding the structure of data rows in SQL Server关于 SQL Server 中数据行结构的固定长度列大小
【发布时间】:2016-11-14 12:48:42
【问题描述】:

我正在阅读本书Microsoft SQL Server 2012 Internals 并尝试了解SQL Server 如何在内部存储数据。以下是本书的节选。

The structure of data rows
==========================

A tables data rows have the general structure shown in the figure below (as long as the data
is stored in uncompressed form). This format is called the FixedVar format because the data
for all fixed-length columns is stored first, followed by the data for all variable-length columns. 

+-----------------------------------------------------------------------------------------------
|1 byte  |1 byte  |2 bytes |n bytes |2 bytes |NULL bitmap|2 bytes |Column offset array|n bytes |
+---^--------^-------^----------^-------^----------^----------^------------^---------------^----
    |        |       |          |       |          |          |            |               |
    |        |       |          |       |          |          |            |               +
    |        |       |          |       |          |          |            |      data for
    |        |       |          |       |          |          |            |      variable-length columns
    |        |       |          |       |          |          |            |
    |        |       |          |       |          |          |            |
    |        |       |          |       |          |          |            +
    |        |       |          |       |          |          |    2x # varlength columns
    |        |       |          |       |          |          +
    |        |       |          |       |          |     Number of variable-length
    |        |       |          |       |          |     columns
    |        |       |          |       |          +
    |        |       |          |       |      NULL bit map, one bit
    |        |       |          |       |      for each column
    |        |       |          |       +
    |        |       |          |     Number of fixed-length
    |        |       |          |     columns
    |        |       |          +
    |        |       |       Fixed length data
    |        |       +
    |        |      Length of fixed-length portion of row,
    |        |      not including the 2 bytes for the number
    |        |      of columns and the NULL bitmap
    |        +
    |      Status bits B
    +
 Status bits A

我的测试如下。

USE test;
GO

IF OBJECT_ID(N'dbo.t1', N'U') IS NOT NULL
    DROP TABLE t1;
GO

-- Create a sample table t1
CREATE TABLE t1
(
 id NVARCHAR(20) NOT NULL DEFAULT '',
 c1 CHAR(100) DEFAULT (REPLICATE('X', 100)),
 c2 NVARCHAR(20) NULL DEFAULT NULL,
 c3 CHAR(100) DEFAULT (REPLICATE('Y', 100)),
);
GO

INSERT INTO t1 VALUES(N'AAAA', REPLICATE('C', 100), N'BBBB', REPLICATE('D', 100));
INSERT INTO t1 DEFAULT VALUES;
INSERT INTO t1 VALUES(N'PPPP', REPLICATE('Q', 100), N'RRRR', REPLICATE('S', 100));
GO

-- Check the data
SELECT * FROM t1;

-- How many pages we have in the table
SELECT  t.name AS TableName,
        p.rows AS RowCounts,
        SUM(a.total_pages) AS TotalPages,
        SUM(a.used_pages) AS UsedPages,
        (SUM(a.total_pages) - SUM(a.used_pages)) AS UnusedPages
FROM    sys.tables t
        INNER JOIN sys.indexes i
            ON t.object_id = i.object_id
        INNER JOIN sys.partitions p
            ON i.object_id = p.object_id
               AND i.index_id = p.index_id
        INNER JOIN sys.allocation_units a
            ON p.partition_id = a.container_id
WHERE   t.name NOT LIKE 'dt%'
        AND t.is_ms_shipped = 0
        AND i.object_id > 255
        AND t.name = 't1'
GROUP BY t.name,
        p.rows
ORDER BY t.name;


-- Create a table to store the DBCC IND result
IF EXISTS ( SELECT  *
            FROM    sys.tables
            WHERE   name = 'sp_dbcc_ind' )
    DROP TABLE sp_dbcc_ind;
GO
CREATE TABLE sp_dbcc_ind
(
 PageFID TINYINT,
 PagePID INT,
 IAMFID TINYINT,
 IAMPID INT,
 ObjectID INT,
 IndexID TINYINT,
 PartitionNumber TINYINT,
 PartitionID BIGINT,
 iam_chain_type VARCHAR(30),
 PageType TINYINT,
 IndexLevel TINYINT,
 NextPageFID TINYINT,
 NextPagePID INT,
 PrevPageFID TINYINT,
 PrevPagePID INT,
 PRIMARY KEY (PageFID, PagePID)
);

INSERT  INTO sp_dbcc_ind
EXECUTE ('DBCC IND(''test'', ''test.dbo.t1'', 1)');

-- Generate DBCC PAGE command
SELECT  'DBCC PAGE ([test], ' + CAST(PageFID AS NVARCHAR(MAX)) + ', ' +
        CAST (PagePID AS NVARCHAR(MAX)) + ', 3) --WITH TABLERESULTS;' DBCC_PAGE,
        *
FROM    sp_dbcc_ind
WHERE PageType = 1;

-- Enable output
DBCC TRACEON(3604);
GO

-- Pasted from the first column of the last query
DBCC PAGE ([test], 1, 552, 3) WITH TABLERESULTS;

-- DBCC PAGE output    
/*
PAGE: (1:552)


BUFFER:


BUF @0x00000045BEE87500

bpage = 0x00000045929D6000          bhash = 0x0000000000000000          bpageno = (1:552)
bdbid = 8                           breferences = 0                     bcputicks = 380
bsampleCount = 1                    bUse1 = 38513                       bstat = 0x10b
blog = 0x212121cc                   bnext = 0x0000000000000000          bDirtyContext = 0x00000045941E04D0
bstat2 = 0x0                        

PAGE HEADER:


Page @0x00000045929D6000

m_pageId = (1:552)                  m_headerVersion = 1                 m_type = 1
m_typeFlagBits = 0x0                m_level = 0                         m_flagBits = 0x8000
m_objId (AllocUnitId.idObj) = 173   m_indexId (AllocUnitId.idInd) = 256 
Metadata: AllocUnitId = 72057594049265664                                
Metadata: PartitionId = 72057594042974208                                Metadata: IndexId = 0
Metadata: ObjectId = 1909581841     m_prevPage = (0:0)                  m_nextPage = (0:0)
pminlen = 204                       m_slotCnt = 3                       m_freeCnt = 7425
m_freeData = 761                    m_reservedCnt = 0                   m_lsn = (34:2220:2)
m_xactReserved = 0                  m_xdesId = (0:0)                    m_ghostRecCnt = 0
m_tornBits = 0                      DB Frag ID = 1                      

Allocation Status

GAM (1:2) = ALLOCATED               SGAM (1:3) = NOT ALLOCATED          PFS (1:1) = 0x41 ALLOCATED  50_PCT_FULL
DIFF (1:6) = CHANGED                ML (1:7) = NOT MIN_LOGGED           

Slot 0 Offset 0x60 Length 229

Record Type = PRIMARY_RECORD        Record Attributes =  NULL_BITMAP VARIABLE_COLUMNS
Record Size = 229                   
Memory Dump @0x00000047E94BA060

0000000000000000:   3000cc00 43434343 43434343 43434343 43434343  0.Ì.CCCCCCCCCCCCCCCC
0000000000000014:   43434343 43434343 43434343 43434343 43434343  CCCCCCCCCCCCCCCCCCCC
0000000000000028:   43434343 43434343 43434343 43434343 43434343  CCCCCCCCCCCCCCCCCCCC
000000000000003C:   43434343 43434343 43434343 43434343 43434343  CCCCCCCCCCCCCCCCCCCC
0000000000000050:   43434343 43434343 43434343 43434343 43434343  CCCCCCCCCCCCCCCCCCCC
0000000000000064:   43434343 44444444 44444444 44444444 44444444  CCCCDDDDDDDDDDDDDDDD
0000000000000078:   44444444 44444444 44444444 44444444 44444444  DDDDDDDDDDDDDDDDDDDD
000000000000008C:   44444444 44444444 44444444 44444444 44444444  DDDDDDDDDDDDDDDDDDDD
00000000000000A0:   44444444 44444444 44444444 44444444 44444444  DDDDDDDDDDDDDDDDDDDD
00000000000000B4:   44444444 44444444 44444444 44444444 44444444  DDDDDDDDDDDDDDDDDDDD
00000000000000C8:   44444444 04000002 00dd00e5 00410041 00410041  DDDD.....Ý.å.A.A.A.A
00000000000000DC:   00420042 00420042 00                          .B.B.B.B.

Slot 0 Column 1 Offset 0xd5 Length 8 Length (physical) 8

id = AAAA                           

Slot 0 Column 2 Offset 0x4 Length 100 Length (physical) 100

c1 = CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC

Slot 0 Column 3 Offset 0xdd Length 8 Length (physical) 8

c2 = BBBB                           

Slot 0 Column 4 Offset 0x68 Length 100 Length (physical) 100

c3 = DDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD

Slot 1 Offset 0x145 Length 207

Record Type = PRIMARY_RECORD        Record Attributes =  NULL_BITMAP    Record Size = 207

Memory Dump @0x00000047E94BA145

0000000000000000:   1000cc00 58585858 58585858 58585858 58585858  ..Ì.XXXXXXXXXXXXXXXX
0000000000000014:   58585858 58585858 58585858 58585858 58585858  XXXXXXXXXXXXXXXXXXXX
0000000000000028:   58585858 58585858 58585858 58585858 58585858  XXXXXXXXXXXXXXXXXXXX
000000000000003C:   58585858 58585858 58585858 58585858 58585858  XXXXXXXXXXXXXXXXXXXX
0000000000000050:   58585858 58585858 58585858 58585858 58585858  XXXXXXXXXXXXXXXXXXXX
0000000000000064:   58585858 59595959 59595959 59595959 59595959  XXXXYYYYYYYYYYYYYYYY
0000000000000078:   59595959 59595959 59595959 59595959 59595959  YYYYYYYYYYYYYYYYYYYY
000000000000008C:   59595959 59595959 59595959 59595959 59595959  YYYYYYYYYYYYYYYYYYYY
00000000000000A0:   59595959 59595959 59595959 59595959 59595959  YYYYYYYYYYYYYYYYYYYY
00000000000000B4:   59595959 59595959 59595959 59595959 59595959  YYYYYYYYYYYYYYYYYYYY
00000000000000C8:   59595959 040004                               YYYY...

Slot 1 Column 1 Offset 0x0 Length 0 Length (physical) 0

id =                                

Slot 1 Column 2 Offset 0x4 Length 100 Length (physical) 100

c1 = XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

Slot 1 Column 3 Offset 0x0 Length 0 Length (physical) 0

c2 = [NULL]                         

Slot 1 Column 4 Offset 0x68 Length 100 Length (physical) 100

c3 = YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY

Slot 2 Offset 0x214 Length 229

Record Type = PRIMARY_RECORD        Record Attributes =  NULL_BITMAP VARIABLE_COLUMNS
Record Size = 229                   
Memory Dump @0x00000047E94BA214

0000000000000000:   3000cc00 51515151 51515151 51515151 51515151  0.Ì.QQQQQQQQQQQQQQQQ
0000000000000014:   51515151 51515151 51515151 51515151 51515151  QQQQQQQQQQQQQQQQQQQQ
0000000000000028:   51515151 51515151 51515151 51515151 51515151  QQQQQQQQQQQQQQQQQQQQ
000000000000003C:   51515151 51515151 51515151 51515151 51515151  QQQQQQQQQQQQQQQQQQQQ
0000000000000050:   51515151 51515151 51515151 51515151 51515151  QQQQQQQQQQQQQQQQQQQQ
0000000000000064:   51515151 53535353 53535353 53535353 53535353  QQQQSSSSSSSSSSSSSSSS
0000000000000078:   53535353 53535353 53535353 53535353 53535353  SSSSSSSSSSSSSSSSSSSS
000000000000008C:   53535353 53535353 53535353 53535353 53535353  SSSSSSSSSSSSSSSSSSSS
00000000000000A0:   53535353 53535353 53535353 53535353 53535353  SSSSSSSSSSSSSSSSSSSS
00000000000000B4:   53535353 53535353 53535353 53535353 53535353  SSSSSSSSSSSSSSSSSSSS
00000000000000C8:   53535353 04000002 00dd00e5 00500050 00500050  SSSS.....Ý.å.P.P.P.P
00000000000000DC:   00520052 00520052 00                          .R.R.R.R.

Slot 2 Column 1 Offset 0xd5 Length 8 Length (physical) 8

id = PPPP                           

Slot 2 Column 2 Offset 0x4 Length 100 Length (physical) 100

c1 = QQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQ

Slot 2 Column 3 Offset 0xdd Length 8 Length (physical) 8

c2 = RRRR                           

Slot 2 Column 4 Offset 0x68 Length 100 Length (physical) 100

c3 = SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS


DBCC execution completed. If DBCC printed error messages, contact your system administrator.

*/

以第一条记录为例。它转储的原始数据是:

0000000000000000:   3000cc00 43434343 43434343 43434343 43434343  0.Ì.CCCCCCCCCCCCCCCC
0000000000000014:   43434343 43434343 43434343 43434343 43434343  CCCCCCCCCCCCCCCCCCCC
0000000000000028:   43434343 43434343 43434343 43434343 43434343  CCCCCCCCCCCCCCCCCCCC
000000000000003C:   43434343 43434343 43434343 43434343 43434343  CCCCCCCCCCCCCCCCCCCC
0000000000000050:   43434343 43434343 43434343 43434343 43434343  CCCCCCCCCCCCCCCCCCCC
0000000000000064:   43434343 44444444 44444444 44444444 44444444  CCCCDDDDDDDDDDDDDDDD
0000000000000078:   44444444 44444444 44444444 44444444 44444444  DDDDDDDDDDDDDDDDDDDD
000000000000008C:   44444444 44444444 44444444 44444444 44444444  DDDDDDDDDDDDDDDDDDDD
00000000000000A0:   44444444 44444444 44444444 44444444 44444444  DDDDDDDDDDDDDDDDDDDD
00000000000000B4:   44444444 44444444 44444444 44444444 44444444  DDDDDDDDDDDDDDDDDDDD
00000000000000C8:   44444444 04000002 00dd00e5 00410041 00410041  DDDD.....Ý.å.A.A.A.A
00000000000000DC:   00420042 00420042 00   

                   .B.B.B.B.

表 t1 有 4 列,两个固定长度(c1 100 字节,c3 100 字节)和两个可变长度(id 和 c2)。根据书上的理论,行结构会先存储定长的列。所以第一个字节是30,是一个状态位,转换成二进制的时候是00110000。第一个 1 表示我们在这一行中有可变长度的列。第二个 1 表示我们在这一行中有 NULL 位图(实际上这始终是 1)。第二个字节是00,是另一个状态位,这里没有特殊含义。接下来的两个字节是cc00,转换成十进制是204。根据本书,这意味着固定长度列的长度。但正如我们所知,有两列,它们都是 char(100) 类型。所以它们的实际长度是200而不是204。根据我的测试,“行定长部分的长度”报告的数据总是比实际的定长数据大小大4个字节。是因为uniquifier吗?所有堆行都有唯一性吗?

【问题讨论】:

  • 能否请您发短信而不是屏幕截图?图片可能在某些域中被阻止
  • @TheGameiswar,完成。
  • 请给出一个重现问题的示例表以及您用来达到这个 4 字节差异的工作逻辑。
  • @MartinSmith OK Martin,我正在添加示例。
  • @MartinSmith 添加了一个示例。

标签: sql-server row storage internal


【解决方案1】:

不,不涉及唯一性。

行的固定长度部分的长度,不包括 2 个字节 列数和 NULL 位图。

但这确实包括您问题|1 byte |1 byte |2 bytes中的前三个元素。

0x3000CC00 + 两个 100 字节的固定长度列构成 204 字节。

查看此值的另一种方法是,它为您提供从行首开始计数的固定长度列计数部分的偏移量。 SQL Server internals viewer 使用如下术语。或者,您可以将其视为从行中当前点到空位图的偏移量 (as Paul Randal does here)

【讨论】:

  • 哦,这就解释了。谢谢马丁。您在快照中使用的工具是什么?
  • @OgrishMan - SQL Server 内部查看器。不幸的是,它仅适用于旧版本的 SSMS。
  • 谢谢!我听说过,但从未尝试过。看起来不错,我试试看。
猜你喜欢
  • 1970-01-01
  • 1970-01-01
  • 1970-01-01
  • 2011-12-19
  • 2011-12-07
  • 2017-07-20
  • 1970-01-01
  • 1970-01-01
  • 1970-01-01
相关资源
最近更新 更多