Oracle Database 12c前幾天正式釋出了，如果學習一個新版本的資料庫？我通常是從New Features Guide文件看起，先通覽文件的目錄，遇到感興趣的新功能點，就開始做實驗來驗證這個新功能。當然，這之前需要先把新版本的資料庫安裝好，先把新版本的全部文件下載到本地，這樣即使你坐在飛機上也可以有文件可查。

這次我的計劃是實驗一下Identity型別的欄位，這個欄位可以用來作主鍵，會自動遞增，這種型別的欄位在SQL Server中早就存在，但是Oracle直到12c才推出這個功能。

通常我不會用sys使用者進行任何實驗（除非是驗證sysdba的新功能），因此總是會先建立一個我自己的dba使用者。
在12c中建立這個使用者首先就遇到了錯誤。

SQL> CREATE USER kamus IDENTIFIED BY oracle DEFAULT tablespace users;
CREATE USER kamus IDENTIFIED BY oracle DEFAULT tablespace users
            *
ERROR at line 1:
ORA-65096: invalid common USER OR ROLE name

對於一個不熟悉的錯誤，第一件事情不是去Google（如果你說Baidu那就更幼稚了），而是用oerr實用程式來看看Oracle自己對這個錯誤是怎麼解釋的。為什麼我喜歡非Windows環境中的Oracle？oerr的存在也是很大一個原因。

[[email protected] ~]$ oerr ora 65096
65096, 00000, "invalid common user or role name"
// *Cause:  An attempt was made to create a common user or role with a name
//          that wass not valid for common users or roles.  In addition to
//          the usual rules for user and role names, common user and role
//          names must start with C## or c## and consist only of ASCII
//          characters.
// *Action: Specify a valid common user or role name.
//

錯誤資訊的解析非常明確地告知“試圖建立一個通用使用者，必需要用C##或者c##開頭”，這時候心裡會有疑問，什麼是common user？但是我通常不會先急著去翻文件，而是先把手頭的事情做完，也就是先把使用者建立上。

SQL> CREATE USER c##kamus IDENTIFIED BY oracle DEFAULT tablespace users;   USER created.   SQL> GRANT dba TO c##kamus;   GRANT succeeded.   SQL> SELECT USERNAME,COMMON,ORACLE_MAINTAINED FROM dba_users;   USERNAME                       COM O
------------------------------ --- -
AUDSYS                         YES Y
GSMUSER                        YES Y
SPATIAL_WFS_ADMIN_USR          YES Y
SPATIAL_CSW_ADMIN_USR          YES Y
APEX_PUBLIC_USER               YES Y
SYSDG                          YES Y
DIP                            YES Y
SYSBACKUP                      YES Y
MDDATA                         YES Y
GSMCATUSER                     YES Y
C##KAMUS                       YES N
SYSKM                          YES Y
XS$NULL                        YES Y
OJVMSYS                        YES Y
ORACLE_OCM                     YES Y
OLAPSYS                        YES Y
SI_INFORMTN_SCHEMA             YES Y
DVSYS                          YES Y
ORDPLUGINS                     YES Y
XDB                            YES Y
ANONYMOUS                      YES Y
CTXSYS                         YES Y
ORDDATA                        YES Y
GSMADMIN_INTERNAL              YES Y
APPQOSSYS                      YES Y
APEX_040200                    YES Y
WMSYS                          YES Y
DBSNMP                         YES Y
ORDSYS                         YES Y
MDSYS                          YES Y
DVF                            YES Y
FLOWS_FILES                    YES Y
SYS                            YES Y
SYSTEM                         YES Y
OUTLN                          YES Y
LBACSYS                        YES Y   36 ROWS selected.

建立C##KAMUS使用者成功之後，再返回去解決心中的疑問，什麼是common user？在聯機文件的左上角搜尋關鍵字common user，會得到如下的結果。

通常我會先瀏覽Concept，如果看完覺得心中疑問已經解決，就會返回繼續作之前的實驗，不會再瀏覽其他的連結；如果想要查詢怎麼做，比如說如何建立Common User，才會繼續去看Task部分。這樣的好處是可以保持專注而不至於被過多文件分心。

但是由於Common User這個概念幾乎是嶄新的，所以我很有興趣繼續探索一下，跟Common User相對的Local User該如何建立。繼續去看Task當然是個方法，但是這裡我選擇的是直接去看SQL Language Reference，因為我們知道一定是在Create User語法裡面會有不同的定義，進入Create User語法頁面，直接搜尋common user，就可以看到如下這段話。

CONTAINER Clause

To create a local user in a pluggable database (PDB), ensure that the current container is that PDB and specify CONTAINER = CURRENT. To create a common user, ensure that the current container is the root and specify CONTAINER = ALL. The name of the common user must begin with C## or c##. If you omit this clause and the current container is a PDB, then CONTAINER = CURRENT is the default. If you omit this clause and the current container is the root, then CONTAINER = ALL is the default.

也就是說我們一定要先登入進一個PDB，才可以建立本地使用者，那麼如何知道現在的sqlplus是登入進了哪個DB呢？這個疑問其實是一個很簡單的聯想，既然需要去一個地方，那麼一定有方法知道我現在在什麼地方，通過簡單地查詢文件，可以得知以下的方法。現在確實在CDB中。

SQL> SHOW con_name   CON_NAME
------------------------------
CDB$ROOT   SQL> SELECT SYS_CONTEXT ('USERENV', 'CON_NAME') FROM DUAL;   SYS_CONTEXT('USERENV','CON_NAME')
-----------------------------------------------
CDB$ROOT

dbca建庫的時候，有一個新選項是同時建立PDB，我勾選過（對於dbca中出現的新選項，如果不是條件不允許，我都會選中進行測試），建立了名字為pdbtest的PDB，那麼現在我想嘗試登入這個PDB，去建立一個Local User。如何登入PDB？Administrator’s Guide中有專門新的一個章節“Part VI Managing a Multitenant Environment”來描述如何管理多租戶環境，瀏覽目錄就可以直接找到“Connecting to a PDB with SQL*Plus”這部分，如下所示。

You can use the following techniques to connect to a PDB with the SQL*Plus CONNECT command:

Database connection using easy connect
Database connection using a net service name

那嘗試直接使用easy connect來登入PDB。

$ sqlplus sys/[email protected]:15210/pdbtest AS sysdba   SQL*Plus: Release 10.2.0.4.0 - Production ON Sat Jul 6 21:44:42 2013   Copyright (c) 1982, 2007, Oracle.  ALL Rights Reserved.     Connected TO:
Oracle DATABASE 12c Enterprise Edition Release 12.1.0.1.0 - 64bit Production
WITH the Partitioning, OLAP, Advanced Analytics AND REAL Application Testing options   SQL>  SELECT SYS_CONTEXT ('USERENV', 'CON_NAME') FROM DUAL;   SYS_CONTEXT('USERENV','CON_NAME')
------------------------------------------
PDBTEST   Elapsed: 00:00:00.01   SQL>  SELECT NAME,PDB FROM dba_services;
 SELECT NAME,PDB FROM dba_services
                      *
ERROR at line 1:
ORA-01219: DATABASE OR pluggable DATABASE NOT OPEN: queries allowed ON fixed TABLES OR views ONLY

PDB沒有Open？嘗試開啟。無法使用startup命令。這是顯而易見的。 原因是我使用了舊版本的SQL*Plus(如上所示是10.2.0.4.0)連線到12c資料庫的PDB中，某些新特性不被支援。

SQL> startup
ORA-24543: instance startup OR shutdown NOT allowed IN pluggable DATABASE

使用12c自帶的SQL*Plus登入，就可以使用startup命令將PDB開啟，使用SQL*Plus管理PDB的詳細命令可以參看文件描述。

SQL> SHOW USER
USER IS "SYS"
SQL> startup
Pluggable DATABASE opened.
SQL>
SQL> SHOW con_name   CON_NAME
------------------------------
PDBTEST

或者可以使用如下語句開啟PDB。

SQL> ALTER PLUGGABLE DATABASE OPEN;   Operation 227 succeeded.   Elapsed: 00:00:02.44   SQL> SELECT NAME, OPEN_MODE, RESTRICTED, OPEN_TIME FROM V$PDBS;   NAME			       OPEN_MODE  RES OPEN_TIME
------------------------------ ---------- --- ------------------------------------------------
PDBTEST 		       READ WRITE NO  06-JUL-13 09.48.57.260 PM

到此，可以建立Local User了。

SQL> CREATE USER kamus IDENTIFIED BY oracle;   USER created.   Elapsed: 00:00:00.10
SQL> GRANT dba TO kamus;   GRANT succeeded.   Elapsed: 00:00:00.03

那麼在一個PDB中可以看到多少使用者呢？可以看到CDB中的使用者嗎？這又是一個簡單的聯想，學習的過程其實是一個發散再收縮的迴圈。看來不可以，只能看到自己的使用者，當然這裡有很多Common User。可以看到即使是在PDB中，cdb_檢視也是可以使用的。

SQL> SELECT CON_ID,COUNT(*) FROM cdb_users GROUP BY con_id;   CON_ID   COUNT(*)
---------- ----------
	 3	   38   Elapsed: 00:00:00.03

那麼再回到CDB中看一下，會是什麼情況？可以看到所有Container中的使用者都可以查詢到。

SQL> SELECT CON_ID,COUNT(*) FROM cdb_users GROUP BY con_id;   CON_ID   COUNT(*)
---------- ----------
         1         36
         2         35
         3         38   Elapsed: 00:00:00.08

終於，我可以回到最開始的實驗目標上去了，在PDB中建立了T1表，id列為Identity型別。

SQL> CREATE TABLE t1 (id NUMBER GENERATED AS IDENTITY);   TABLE created.   Elapsed: 00:00:00.22

根據文件描述，Identity型別仍然是通過Sequence來實現的，那麼應該是自動建立了一個Sequence，果然如此。在你學習的過程中會多此一步來查詢一下Sequence檢視嗎？

SQL> SELECT SEQUENCE_NAME FROM user_sequences;   SEQUENCE_NAME
----------------------------------------------------------------------------------------------
ISEQ$$_91620   Elapsed: 00:00:00.00

預設建立的Sequence，CACHE_SIZE是20，開始值是1，這都跟單獨建立的Sequence預設值一樣。

SQL> SELECT * FROM user_sequences;   SEQUENCE_NAME															  MIN_VALUE  MAX_VALUE
-------------------------------------------------------------------------------------------------------------------------------- ---------- ----------
INCREMENT_BY C O CACHE_SIZE LAST_NUMBER PARTITION_COUNT S K
------------ - - ---------- ----------- --------------- - -
ISEQ$$_91620																  1 1.0000E+28
	   1 N N	 20	      1 		N N     Elapsed: 00:00:00.00

插入一條資料試一下，報錯報錯還是報錯。所以如果是generated always的identity列，如果只有這一列，沒法插入資料。

SQL> INSERT INTO t1 VALUES('');
INSERT INTO t1 VALUES('')
*
ERROR at line 1:
ORA-32795: cannot INSERT INTO a generated always IDENTITY COLUMN     Elapsed: 00:00:00.01   SQL> INSERT INTO t1 VALUES(ISEQ$$_91620.NEXTVAL);
INSERT INTO t1 VALUES(ISEQ$$_91620.NEXTVAL)
*
ERROR at line 1:
ORA-32795: cannot INSERT INTO a generated always IDENTITY COLUMN     Elapsed: 00:00:00.00   SQL> INSERT INTO t1 VALUES(NULL);
INSERT INTO t1 VALUES(NULL)
*
ERROR at line 1:
ORA-32795: cannot INSERT INTO a generated always IDENTITY COLUMN     Elapsed: 00:00:00.00

換GENERATED BY DEFAULT ON NULL 型別試一下，Wait，如果刪除了表，對應的序列會自動刪除嗎？理論上應該會，當然還是要測試一下。

SQL> DROP TABLE t1;   TABLE dropped.   Elapsed: 00:00:00.19

序列還在？

SQL> SELECT * FROM user_sequences;   SEQUENCE_NAME															  MIN_VALUE  MAX_VALUE
-------------------------------------------------------------------------------------------------------------------------------- ---------- ----------
INCREMENT_BY C O CACHE_SIZE LAST_NUMBER PARTITION_COUNT S K
------------ - - ---------- ----------- --------------- - -
ISEQ$$_91620																  1 1.0000E+28
	   1 N N	 20	      1 		N N     Elapsed: 00:00:00.00

再建一張表。

SQL> CREATE TABLE t2 (id NUMBER GENERATED BY DEFAULT AS IDENTITY);   TABLE created.   Elapsed: 00:00:00.01

現在是2個序列了。

SQL> SELECT * FROM user_sequences;   SEQUENCE_NAME															  MIN_VALUE  MAX_VALUE
-------------------------------------------------------------------------------------------------------------------------------- ---------- ----------
INCREMENT_BY C O CACHE_SIZE LAST_NUMBER PARTITION_COUNT S K
------------ - - ---------- ----------- --------------- - -
ISEQ$$_91620																  1 1.0000E+28
	   1 N N	 20	      1 		N N   ISEQ$$_91622																  1 1.0000E+28
	   1 N N	 20	      1 		N N     Elapsed: 00:00:00.00

寫完整的Drop語句試一下。

SQL> DROP TABLE t2 cascade CONSTRAINT purge;   TABLE dropped.   Elapsed: 00:00:00.12

後建立的序列已經被自動刪除了，之前建立的還在。

SQL> SELECT * FROM user_sequences;   SEQUENCE_NAME															  MIN_VALUE  MAX_VALUE
-------------------------------------------------------------------------------------------------------------------------------- ---------- ----------
INCREMENT_BY C O CACHE_SIZE LAST_NUMBER PARTITION_COUNT S K
------------ - - ---------- ----------- --------------- - -
ISEQ$$_91620																  1 1.0000E+28
	   1 N N	 20	      1 		N N     Elapsed: 00:00:00.00

兩者的不同應該是purge，如果被刪除的表還在回收站中，序列是會保留的，Make sense，因為表還可能從回收站裡面再restore回來，需要保證序列仍然有效。那麼清空回收站實驗一下。

SQL> purge recyclebin;   Recyclebin purged.   Elapsed: 00:00:00.04

果然，序列也相應地被刪除了。

SQL> SELECT * FROM user_sequences;   no ROWS selected   Elapsed: 00:00:00.00

再回到正題，建立T3表，插入一條資料。

SQL> CREATE TABLE t3 (id NUMBER GENERATED BY DEFAULT ON NULL AS IDENTITY);   TABLE created.   Elapsed: 00:00:00.01
SQL> INSERT INTO t3 VALUES(NULL);   1 ROW created.   Elapsed: 00:00:00.01

序列的LAST_NUMBER已經增加為21。

SQL> SELECT * FROM user_sequences;   SEQUENCE_NAME															  MIN_VALUE  MAX_VALUE
-------------------------------------------------------------------------------------------------------------------------------- ---------- ----------
INCREMENT_BY C O CACHE_SIZE LAST_NUMBER PARTITION_COUNT S K
------------ - - ---------- ----------- --------------- - -
ISEQ$$_91624																  1 1.0000E+28
	   1 N N	 20	     21 		N N

後臺如何操作的？使用10046 trace，再插入幾條資料。

SQL> INSERT INTO t3 VALUES(NULL);   1 ROW created.   SQL> INSERT INTO t3 VALUES(NULL);   1 ROW created.   SQL> SELECT * FROM t3;   ID
----------
	 1
	 2
	 3

檢視10046 trace的結果。可以看到執行計劃中直接呼叫了SEQUENCE，就跟之前插入記錄的時候明確指定SEQ.NEXTVAL一樣。其實Oracle的實現方法非常簡單，這一列其實就是Number型別，然後將這一列的Default值設定為”KAMUS”.”ISEQ$$_91624″.nextval，僅此而已。

--tkprof 10046 trace
insert into t3
values
(null)     call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.00       0.00          0          0          0           0
Execute      1      0.00       0.00          0          1          3           1
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        2      0.00       0.00          0          1          3           1   Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 104
Number of plan statistics captured: 1   Rows (1st) Rows (avg) Rows (max)  Row Source Operation
---------- ---------- ----------  ---------------------------------------------------
         0          0          0  LOAD TABLE CONVENTIONAL  (cr=1 pr=0 pw=0 time=90 us)
         1          1          1   SEQUENCE  ISEQ$$_91624 (cr=0 pr=0 pw=0 time=14 us)     Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  SQL*Net message to client                       1        0.00          0.00
  SQL*Net message from client                     1        5.28          5.28
********************************************************************************

到此為止，可以休息一下了，從ORA-65096開始大概花費了1個多小時的時間，我學習到了：
1. 什麼是Common User，什麼是Local User？
2. 如何查詢現在的環境是CDB還是某個PDB？
3. 如何登入PDB？
4. 如何啟動PDB？
5. PDB和CDB中檢視看到的內容有怎樣的不同？
6. 如何建立Identity型別的列？
7. 刪除表以後，對應的Sequence如何處理？
8. Oracle後臺對於Identity列是如何處理的？

你是不是也是這樣學習的呢？

[email protected]
謝謝留言中Asher的建議，增加如下測試。
使用DBMS_METADATA.GET_DDL獲取到的DDL資訊，已經是符合12c語法的樣式了，顯示出了Sequence的具體資訊。

SQL> SELECT dbms_metadata.GET_DDL('TABLE','T3') FROM dual;   DBMS_METADATA.GET_DDL('TABLE','T3')
--------------------------------------------------------------------------------   CREATE TABLE "KAMUS"."T3"
   (	"ID" NUMBER GENERATED BY DEFAULT ON NULL AS IDENTITY MINVALUE 1 MAXVALUE 99
99999999999999999999999999 INCREMENT BY 1 START WITH 1 CACHE 20 NOORDER  NOCYCLE
  NOT NULL ENABLE,
	"COMMENTS" VARCHAR2(100)
   ) SEGMENT CREATION IMMEDIATE
  PCTFREE 10 PCTUSED 40 INITRANS 1 MAXTRANS 255
 NOCOMPRESS LOGGING
  STORAGE(INITIAL 65536 NEXT 1048576 MINEXTENTS 1 MAXEXTENTS 2147483645
  PCTINCREASE 0 FREELISTS 1 FREELIST GROUPS 1
  BUFFER_POOL DEFAULT FLASH_CACHE DEFAULT CELL_FLASH_CACHE DEFAULT)
  TABLESPACE "USERS"

系統自動產生的序列無法手工修改屬性。

SQL> ALTER SEQUENCE "ISEQ$$_91624" INCREMENT BY 10;
ALTER SEQUENCE "ISEQ$$_91624" INCREMENT BY 10
*
ERROR at line 1:
ORA-32793: cannot ALTER a system-generated SEQUENCE   SQL> host oerr ora 32793
32793,0000, "cannot alter a system-generated sequence"
// *Cause:  An attempt was made TO ALTER a system-generated SEQUENCE.
// *Action: A system-generated SEQUENCE, such AS one created FOR an
//          IDENTITY COLUMN, cannot be altered.

系統自動產生的序列也不允許刪除。

SQL> DROP SEQUENCE "ISEQ$$_91624";
DROP SEQUENCE "ISEQ$$_91624"
              *
ERROR at line 1:
ORA-32794: cannot DROP a system-generated SEQUENCE   SQL> host oerr ora 32794
32794,0000, "cannot drop a system-generated sequence"
// *Cause:  An attempt was made TO DROP a system-generated SEQUENCE.
// *Action: A system-generated SEQUENCE, such AS one created FOR an
//          IDENTITY COLUMN, cannot be dropped.

在11gR2中錯誤資訊編號在ORA-32790和ORA-32800之間是空白，而12c使用了這其間的8個錯誤號作為新特性的報錯。

ORA-32791: prebuilt table managed column cannot have a default on null
Cause: An attempt was made to create a materialized view on a prebuilt table that has a managed column with a default on null expression.
Action: Either remove the default on null property, or do not include the column in the materialized view definition.

ORA-32792: prebuilt table managed column cannot be an identity column
Cause: An attempt was made to create a materialized view on a prebuilt table that has a managed column that is an identity column.
Action: Either remove the identity property, or do not include the column in the materialized view definition.

ORA-32793: cannot alter a system-generated sequence
Cause: An attempt was made to alter a system-generated sequence.
Action: A system-generated sequence, such as one created for an identity column, cannot be altered.

ORA-32794: cannot drop a system-generated sequence
Cause: An attempt was made to drop a system-generated sequence.
Action: A system-generated sequence, such as one created for an identity column, cannot be dropped.

ORA-32795: cannot insert into a generated always identity column
Cause: An attempt was made to insert a value into an identity column created with GENERATED ALWAYS keywords.
Action: A generated always identity column cannot be directly inserted. Instead, the associated sequence generator must provide the value.

ORA-32796: cannot update a generated always identity column
Cause: An attempt was made to update an identity column created with GENERATED ALWAYS keywords.
Action: A generated always identity column cannot be directly updated.

ORA-32797: identity column sequence mismatch in ALTER TABLE EXCHANGE PARTITION
Cause: The two tables specified in the EXCHANGE have identity columns with sequences that are neither both increasing nor decreasing.
Action: Ensure that the identity columns have sequences with INCREMENT BY having the same sign.

ORA-32798: cannot use ANSI RIGHT or FULL outer join with a left correlation
Cause: An attempt was made to use a lateral view with a left correlation to the first operand of an ANSI RIGHT or FULL outer join.
Action: Rewrite the query without the left correlation.