生产过程中常常会遇到对表需要在线增加列，尤其是新增带缺省值的非空列，例如：
ALTER TABLE .. ADD COLUMN ( ... NOT NULL DEFAULT ... )
在Oracle 11g以前，执行这条DDL会涉及以下步骤：
- 在数据字典中中增加新列定义：列类型、长度、NOT NULL及DEFAULT等；
- 更新表上所有行，使用新列的缺省值增加新列的值。
对百万级以上的大表，该DDL可能会耗费较长时间及系统资源，同时也会对该表进行长时间加锁，影响业务的正常使用。
Oracle 11g起引入了新的隐藏参数：_ADD_COL_OPTIM_ENABLED。如果数据库参数：COMPATIBLE>=11时，则_ADD_COL_OPTIM_ENABLED缺省为TRUE。
_ADD_COL_OPTIM_ENABLED值为True时，新增带缺省值非空列的DDL会仅仅将新列添加到表定义中，但是不会更新实际的表行记录。 内部会在数据字典中存储一个标志，对于没有列值的行均会返回DEFAULT值。 对于大表，这会节省大量资源和加快执行速度。 在数据库内部，访问该表的任何代码路径都会查询数据字典设置，为该列返回正确的数据。该 参数的设置仅影响新的“ ADD COLUMN”命令。 一旦使用此种方式增加新的列，则数据字典中的标志将指示该特定列已被优化。可以通过以下查询获取已优化的列：
col object_name format a30
set lines 100
select owner, object_name, name
from dba_objects, col$
where bitand(col$.PROPERTY,1073741824)=1073741824
and object_id=obj#;
下面通过几个简单案例来分析下列值在数据块中的存储变化。
准备
测试数据库：11.2.0.4
compatible=11.2.0.4
_add_col_optim_enabled=true
SQL> create table test.t_obj(c1 varchar2(1));
Table created.
SQL> insert into test.t_obj values('1');
1 row created.
SQL> commit;
Commit complete.
SQL> select * from test.t_obj;
C
-
1
SQL> select dbms_rowid.rowid_object(rowid) obj#,
dbms_rowid.rowid_relative_fno(rowid) rfile#,
dbms_rowid.rowid_block_number(rowid) block#,
dbms_rowid.rowid_row_number(rowid) row#
from test.t_obj; 2 3 4 5
OBJ# RFILE# BLOCK# ROW#
---------- ---------- ---------- ----------
70623 4 134 0
SQL> alter system checkpoint;
System altered.
SQL> alter system dump datafile 4 block 134;
System altered.
查看trace文件：
Block header dump: 0x01000086
Object id on Block? Y
seg/obj: 0x113df csc: 0x00.95330 itc: 2 flg: E typ: 1 - DATA
brn: 0 bdba: 0x1000080 ver: 0x01 opc: 0
inc: 0 exflg: 0
Itl Xid Uba Flag Lck Scn/Fsc
0x01 0x000a.011.000010a8 0x00c028a4.00cb.26 --U- 1 fsc 0x0000.00095331
0x02 0x0000.000.00000000 0x00000000.0000.00 ---- 0 fsc 0x0000.00000000
bdba: 0x01000086
data_block_dump,data header at 0x7f96a71b3264
===============
tsiz: 0x1f98
hsiz: 0x14
pbl: 0x7f96a71b3264
76543210
flag=--------
ntab=1
nrow=1
frre=-1
fsbo=0x14
fseo=0x1f93
avsp=0x1f7b
tosp=0x1f7b
0xe:pti[0] nrow=1 offs=0
0x12:pri[0] offs=0x1f93
block_row_dump:
tab 0, row 0, @0x1f93
tl: 5 fb: --H-FL-- lb: 0x1 cc: 1
col 0: [ 1] 31
end_of_block_dump
End dump data blocks tsn: 4 file#: 4 minblk 134 maxblk 134
案例一：新增带缺省值的非空列
SQL> alter table test.t_obj add c2 varchar2(1) default 'a' not null;
Table altered.
SQL> col binarydefval format a10
select ec.* from sys.ecol$ ec,sys.obj$ o where ec.tabobj#=o.dataobj#;
SQL>
TABOBJ# COLNUM BINARYDEFV
---------- ---------- ----------
70623 2 61
SQL> select c1,c2,dbms_rowid.rowid_object(rowid) obj#,
dbms_rowid.rowid_relative_fno(rowid) rfile#,
dbms_rowid.rowid_block_number(rowid) block#,
dbms_rowid.rowid_row_number(rowid) row#
from test.t_obj; 2 3 4 5
C C OBJ# RFILE# BLOCK# ROW#
- - ---------- ---------- ---------- ----------
1 a 70623 4 134 0
SQL> col object_name format a30
set lines 100
select owner, object_name, name
from dba_objects, col$
where bitand(col$.PROPERTY,1073741824)=1073741824
and object_id=obj#;
OWNER OBJECT_NAME NAME
------------------------------ ------------------------------ ------------------------------
TEST T_OBJ C2
SQL> alter system checkpoint;
System altered.
SQL> alter system dump datafile 4 block 134;
System altered.
查看trace文件：
Block header dump: 0x01000086
Object id on Block? Y
seg/obj: 0x113df csc: 0x00.95330 itc: 2 flg: E typ: 1 - DATA
brn: 0 bdba: 0x1000080 ver: 0x01 opc: 0
inc: 0 exflg: 0
Itl Xid Uba Flag Lck Scn/Fsc
0x01 0x000a.011.000010a8 0x00c028a4.00cb.26 --U- 1 fsc 0x0000.00095331
0x02 0x0000.000.00000000 0x00000000.0000.00 ---- 0 fsc 0x0000.00000000
bdba: 0x01000086
data_block_dump,data header at 0x7f96a71b3264
===============
tsiz: 0x1f98
hsiz: 0x14
pbl: 0x7f96a71b3264
76543210
flag=--------
ntab=1
nrow=1
frre=-1
fsbo=0x14
fseo=0x1f93
avsp=0x1f7b
tosp=0x1f7b
0xe:pti[0] nrow=1 offs=0
0x12:pri[0] offs=0x1f93
block_row_dump:
tab 0, row 0, @0x1f93
tl: 5 fb: --H-FL-- lb: 0x1 cc: 1
col 0: [ 1] 31
end_of_block_dump
查询返回了期望的列值，但在数据块实际没有存储（根据Oracle内部机制，因为是NULL，所以未在数据块中存储）。
案例二：修改非空列的default
在案例一的基础上继续
SQL> alter table test modify (c2 default 'b');
Table altered.
SQL> select ec.* from sys.ecol$ ec,sys.obj$ o where ec.tabobj#=o.dataobj#;
TABOBJ# COLNUM BINARYDEFV
---------- ---------- ----------
70623 2 61
SQL> select column_name,data_default from dba_tab_columns where owner='TEST';
COLUMN_NAME DATA_DEFAULT
------------------------------ ---------------
C1
C2 'b'
SQL> insert into test.t_obj(c1) values('2');
1 row created.
SQL> commit;
Commit complete.
SQL> select * from test.t_obj;
C C
- -
1 a
2 b
SQL> select c1,c2,dbms_rowid.rowid_object(rowid) obj#,
dbms_rowid.rowid_relative_fno(rowid) rfile#,
dbms_rowid.rowid_block_number(rowid) block#,
dbms_rowid.rowid_row_number(rowid) row#
from test.t_obj; 2 3 4 5
C C OBJ# RFILE# BLOCK# ROW#
- - ---------- ---------- ---------- ----------
1 a 70623 4 134 0
2 b 70623 4 134 1
SQL> alter system checkpoint;
System altered.
SQL> alter system dump datafile 4 block 134;
System altered.
Block header dump: 0x01000086
Object id on Block? Y
seg/obj: 0x113df csc: 0x00.95330 itc: 2 flg: E typ: 1 - DATA
brn: 0 bdba: 0x1000080 ver: 0x01 opc: 0
inc: 0 exflg: 0
Itl Xid Uba Flag Lck Scn/Fsc
0x01 0x000a.011.000010a8 0x00c028a4.00cb.26 --U- 1 fsc 0x0000.00095331
0x02 0x000a.01b.0000109b 0x00c028aa.00cb.03 --U- 1 fsc 0x0000.0009558d
bdba: 0x01000086
data_block_dump,data header at 0x7f96a71b3264
===============
tsiz: 0x1f98
hsiz: 0x16
pbl: 0x7f96a71b3264
76543210
flag=--------
ntab=1
nrow=2
frre=-1
fsbo=0x16
fseo=0x1f8c
avsp=0x1f70
tosp=0x1f70
0xe:pti[0] nrow=2 offs=0
0x12:pri[0] offs=0x1f93
0x14:pri[1] offs=0x1f8c
block_row_dump:
tab 0, row 0, @0x1f93
tl: 5 fb: --H-FL-- lb: 0x1 cc: 1
col 0: [ 1] 31
tab 0, row 1, @0x1f8c
tl: 7 fb: --H-FL-- lb: 0x2 cc: 2
col 0: [ 1] 32
col 1: [ 1] 62
end_of_block_dump
新插入的行，非空列使用了修改后的default值，已有行继续使用原有的default值。
案例三：将非空列去除非空属性
在案例二的基础上继续
SQL> alter table test.t_obj modify c2 null;
Table altered.
SQL> select ec.* from sys.ecol$ ec,sys.obj$ o where ec.tabobj#=o.dataobj#;
no rows selected
SQL> select column_name,data_default from dba_tab_columns where owner='TEST';
COLUMN_NAME DATA_DEFAULT
------------------------------ ---------------
C1
C2 'b'
SQL> select c1,c2,dbms_rowid.rowid_object(rowid) obj#,
dbms_rowid.rowid_relative_fno(rowid) rfile#,
dbms_rowid.rowid_block_number(rowid) block#,
dbms_rowid.rowid_row_number(rowid) row#
from test.t_obj; 2 3 4 5
C C OBJ# RFILE# BLOCK# ROW#
- - ---------- ---------- ---------- ----------
1 a 70623 4 134 0
2 b 70623 4 134 1
SQL> alter system checkpoint;
System altered.
SQL> alter system dump datafile 4 block 134;
System altered.
Block header dump: 0x01000086
Object id on Block? Y
seg/obj: 0x113df csc: 0x00.95643 itc: 2 flg: E typ: 1 - DATA
brn: 0 bdba: 0x1000080 ver: 0x01 opc: 0
inc: 0 exflg: 0
Itl Xid Uba Flag Lck Scn/Fsc
0x01 0x000a.004.000010a3 0x00c028ab.00cb.04 --U- 2 fsc 0x0000.0009564a
0x02 0x000a.01b.0000109b 0x00c028aa.00cb.03 C--- 0 scn 0x0000.0009558d
bdba: 0x01000086
data_block_dump,data header at 0x7f96a71b3264
===============
tsiz: 0x1f98
hsiz: 0x16
pbl: 0x7f96a71b3264
76543210
flag=--------
ntab=1
nrow=2
frre=-1
fsbo=0x16
fseo=0x1f85
avsp=0x1f70
tosp=0x1f70
0xe:pti[0] nrow=2 offs=0
0x12:pri[0] offs=0x1f85
0x14:pri[1] offs=0x1f8c
block_row_dump:
tab 0, row 0, @0x1f85
tl: 7 fb: --H-FL-- lb: 0x1 cc: 2
col 0: [ 1] 31
col 1: [ 1] 61
tab 0, row 1, @0x1f8c
tl: 7 fb: --H-FL-- lb: 0x1 cc: 2
col 0: [ 1] 32
col 1: [ 1] 62
end_of_block_dump
此时可以看到数据块中的列值发生了变化，原有的NULL更新为具体值。对于大表来说，这个操作将较为耗时。
案例四：继续增加带缺省值的非空列和不带缺省值的空列
在案例三的基础上继续
SQL> alter table test.t_obj add c3 varchar2(1) default 'c' not null;
Table altered.
SQL> select c1,c2,c3,dbms_rowid.rowid_object(rowid) obj#,
dbms_rowid.rowid_relative_fno(rowid) rfile#,
dbms_rowid.rowid_block_number(rowid) block#,
dbms_rowid.rowid_row_number(rowid) row#
from test.t_obj; 2 3 4 5
C C C OBJ# RFILE# BLOCK# ROW#
- - - ---------- ---------- ---------- ----------
1 a c 70623 4 134 0
2 b c 70623 4 134 1
SQL> select ec.* from sys.ecol$ ec,sys.obj$ o where ec.tabobj#=o.dataobj#;
TABOBJ# COLNUM BINARYDEFV
---------- ---------- ----------
70623 3 63
SQL> alter system checkpoint;
System altered.
SQL> alter table test.t_obj add c4 varchar2(1);
Table altered.
SQL> insert into test.t_obj(c1,c4) values('3','3');
1 row created.
SQL> commit;
Commit complete.
SQL> select c1,c2,c3,c4,dbms_rowid.rowid_object(rowid) obj#,
dbms_rowid.rowid_relative_fno(rowid) rfile#,
dbms_rowid.rowid_block_number(rowid) block#,
dbms_rowid.rowid_row_number(rowid) row#
from test.t_obj; 2 3 4 5
C C C C OBJ# RFILE# BLOCK# ROW#
- - - - ---------- ---------- ---------- ----------
1 a c 70623 4 134 0
2 b c 70623 4 134 1
3 b c 3 70623 4 134 2
Block header dump: 0x01000086
Object id on Block? Y
seg/obj: 0x113df csc: 0x00.95643 itc: 2 flg: E typ: 1 - DATA
brn: 0 bdba: 0x1000080 ver: 0x01 opc: 0
inc: 0 exflg: 0
Itl Xid Uba Flag Lck Scn/Fsc
0x01 0x000a.004.000010a3 0x00c028ab.00cb.04 --U- 2 fsc 0x0000.0009564a
0x02 0x000a.01a.000010a8 0x00c028ac.00cb.1a --U- 1 fsc 0x0000.000957eb
bdba: 0x01000086
data_block_dump,data header at 0x7f96a71b3264
===============
tsiz: 0x1f98
hsiz: 0x18
pbl: 0x7f96a71b3264
76543210
flag=--------
ntab=1
nrow=3
frre=-1
fsbo=0x18
fseo=0x1f7a
avsp=0x1f63
tosp=0x1f63
0xe:pti[0] nrow=3 offs=0
0x12:pri[0] offs=0x1f85
0x14:pri[1] offs=0x1f8c
0x16:pri[2] offs=0x1f7a
block_row_dump:
tab 0, row 0, @0x1f85
tl: 7 fb: --H-FL-- lb: 0x1 cc: 2
col 0: [ 1] 31
col 1: [ 1] 61
tab 0, row 1, @0x1f8c
tl: 7 fb: --H-FL-- lb: 0x1 cc: 2
col 0: [ 1] 32
col 1: [ 1] 62
tab 0, row 2, @0x1f7a
tl: 11 fb: --H-FL-- lb: 0x2 cc: 4
col 0: [ 1] 33
col 1: [ 1] 62
col 2: [ 1] 63
col 3: [ 1] 33
end_of_block_dump
案例五：alter table move的影响
在案例四基础上继续：
SQL> alter table test.t_obj move;
Table altered.
SQL> alter system checkpoint;
System altered.
SQL> select c1,c2,c3,c4,dbms_rowid.rowid_object(rowid) obj#,
dbms_rowid.rowid_relative_fno(rowid) rfile#,
dbms_rowid.rowid_block_number(rowid) block#,
dbms_rowid.rowid_row_number(rowid) row#
from test.t_obj; 2 3 4 5
C C C C OBJ# RFILE# BLOCK# ROW#
- - - - ---------- ---------- ---------- ----------
1 a c 70624 4 139 0
2 b c 70624 4 139 1
3 b c 3 70624 4 139 2
SQL> alter system dump datafile 4 block 139;
System altered.
Block header dump: 0x0100008b
Object id on Block? Y
seg/obj: 0x113e0 csc: 0x00.9594b itc: 3 flg: E typ: 1 - DATA
brn: 0 bdba: 0x1000088 ver: 0x01 opc: 0
inc: 0 exflg: 0
Itl Xid Uba Flag Lck Scn/Fsc
0x01 0x000a.020.000010ad 0x00000000.0000.00 ---- 0 fsc 0x0000.00000000
0x02 0x0000.000.00000000 0x00000000.0000.00 ---- 0 fsc 0x0000.00000000
0x03 0x0000.000.00000000 0x00000000.0000.00 ---- 0 fsc 0x0000.00000000
bdba: 0x0100008b
data_block_dump,data header at 0x7f96a71b327c
===============
tsiz: 0x1f80
hsiz: 0x18
pbl: 0x7f96a71b327c
76543210
flag=--------
ntab=1
nrow=3
frre=-1
fsbo=0x18
fseo=0x1f63
avsp=0x1f4b
tosp=0x1f4b
0xe:pti[0] nrow=3 offs=0
0x12:pri[0] offs=0x1f77
0x14:pri[1] offs=0x1f6e
0x16:pri[2] offs=0x1f63
block_row_dump:
tab 0, row 0, @0x1f77
tl: 9 fb: --H-FL-- lb: 0x0 cc: 3
col 0: [ 1] 31
col 1: [ 1] 61
col 2: [ 1] 63
tab 0, row 1, @0x1f6e
tl: 9 fb: --H-FL-- lb: 0x0 cc: 3
col 0: [ 1] 32
col 1: [ 1] 62
col 2: [ 1] 63
tab 0, row 2, @0x1f63
tl: 11 fb: --H-FL-- lb: 0x0 cc: 4
col 0: [ 1] 33
col 1: [ 1] 62
col 2: [ 1] 63
col 3: [ 1] 33
end_of_block_dump
End dump data blocks tsn: 4 file#: 4 minblk 139 maxblk 139
结论
- _ADD_COL_OPTIM_ENABLED=true生效后，可以实现大表的快速新增带缺省值的非空列，对于生产环境较有意义。
- 使用该特性增加列，需注意后期一些操作的影响，例如：not null改为null、alter table move等。
- 该参数优化仅对新增带缺省值的非空列有效，对带缺省值的空列无效。
Oracle 12c又引入了_add_nullable_column_with_default_optim参数，可以对新增带缺省值的空列也有效。