Hemant K Chitale

A Materialized View is actually also a physical Table (by the same name) that is created and maintained to store the rows that the MV query is supposed to present.

Since it is also a Table, you can build custom Indexes on it.

Here, my Source Table has an Index on OBJECT_ID :

SQL> create table source_table_1
  2  as select object_id, owner, object_name
  3  from dba_objects
  4  where object_id is not null
  5  /

Table created.

SQL> alter table source_table_1
  2  add constraint source_table_1_pk
  3  primary key (object_id)
  4  /

Table altered.

SQL> create materialized view log on source_table_1;

Materialized view log created.

SQL>

I then build Materialized View with  an additional Index on it :

SQL> create materialized view mv_1
  2  refresh fast on demand
  3  as select object_id as obj_id, owner as obj_owner, object_name as obj_name
  4  from source_table_1
  5  /

Materialized view created.

SQL> create index mv_1_ndx_on_owner
  2  on mv_1 (obj_owner)
  3  /

Index created.

SQL>

Let's see if this Index is usable.

SQL> exec  dbms_stats.gather_table_stats('','MV_1');

PL/SQL procedure successfully completed.

SQL> explain plan for
  2  select obj_owner, count(*)
  3  from mv_1
  4  where obj_owner like 'H%'
  5  group by obj_owner
  6  /

Explained.

SQL> select * from table(dbms_xplan.display);

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 2523122927

------------------------------------------------------------------------------------------
| Id  | Operation            | Name              | Rows  | Bytes | Cost (%CPU)| Time     |
------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT     |                   |     2 |    10 |    15   (0)| 00:00:01 |
|   1 |  SORT GROUP BY NOSORT|                   |     2 |    10 |    15   (0)| 00:00:01 |
|*  2 |   INDEX RANGE SCAN   | MV_1_NDX_ON_OWNER |  5943 | 29715 |    15   (0)| 00:00:01 |
------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------
---------------------------------------------------

   2 - access("OBJ_OWNER" LIKE 'H%')
       filter("OBJ_OWNER" LIKE 'H%')

Note how this Materialized View has a column called "OBJ_OWNER"  (while the Source Table column is called "OWNER") and the Index ("MV_1_NDX_ON_OWNER") on this column is used.


You  would have also noted that you can run DBMS_STATS.GATHER_TABLE_STATS on a Materialized View and it's Indexes.

However, it is NOT a good idea to define your own Unique Indexes (including Primary Key) on a Materialized View.  During the course of a Refresh, the MV may not be consistent and the Unique constraint may be violated.   See Oracle Support Document # 67424.1

In my previous blog post, I've shown a Materialized View that is built as an empty MV and subsequently populated by a Refresh call.

You can also define a Materialized View over an *existing*  (pre-populated) Table.

Let's say you have a Source Table and have built a Replica of it it another Schema or Database.  Building the Replica may have taken an hour or even a few hours.  You now know that the Source Table will have some changes every day and want the Replica to be updated as well.  Instead of executing, say, a TRUNCATE and INSERT, into the Replica every day, you define a Fast Refresh Materialized View over it and let Oracle identify all the changes (which, on a daily basis, could be a small percentage of the total size of the Source/Replica) and update the Replica using a Refresh call.


Here's a quick demo.

SQL> select count(*) from my_large_source;

  COUNT(*)
----------
     72447

SQL> grant select on my_large_source to hr;

Grant succeeded.

SQL> connect hr/HR@orclpdb1
Connected.
SQL> alter session enable parallel dml;

Session altered.

SQL> create table my_large_replica
  2  as select * from hemant.my_large_source
  3  where 1=2;

Table created.

SQL> insert /*+ PARALLEL (8) */
  2  into my_large_replica
  3  select * from hemant.my_large_source;

72447 rows created.

SQL>

So, now, HR has a Replica of the Source Table in the HEMANT schema.  Without any subsequent updates to the Source Table, I create the Materialized View definition, with the "ON PREBUILT TABLE" clause.

SQL> connect hemant/hemant@orclpdb1
Connected.
SQL> create materialized view log on my_large_source;

Materialized view log created.

SQL> grant select, delete on mlog$_my_large_source to hr;

Grant succeeded.

SQL> connect hr/HR@orclpdb1
Connected.
SQL>
SQL> create materialized view my_large_replica
  2  on prebuilt table
  3  refresh fast
  4  as select * from hemant.my_large_source;

Materialized view created.

SQL> select count(*) from hemant.my_large_source;

  COUNT(*)
----------
     72447

SQL> select count(*) from my_large_replica;

  COUNT(*)
----------
     72447

SQL>

I am now ready to add data and Refresh the MV.

SQL> connect hemant/hemant@orclpdb1
Connected.
SQL> desc my_large_source
 Name                                      Null?    Type
 ----------------------------------------- -------- ----------------------------
 ID_COL                                    NOT NULL NUMBER
 PRODUCT_NAME                                       VARCHAR2(128)
 FACTORY                                            VARCHAR2(128)

SQL> insert into my_large_source
  2  values (74000,'Revolutionary Pin','Outer Space');

1 row created.

SQL> commit;

Commit complete.

SQL> select count(*) from mlog$_my_large_source;

  COUNT(*)
----------
         1

SQL>
SQL> connect hr/HR@orclpdb1
Connected.
SQL> select count(*) from hemant.my_large_source;

  COUNT(*)
----------
     72448

SQL> select count(*) from my_large_replica;

  COUNT(*)
----------
     72447

SQL>
SQL> execute dbms_mview.refresh('MY_LARGE_REPLICA','F');

PL/SQL procedure successfully completed.

SQL> select count(*) from my_large_replica;

  COUNT(*)
----------
     72448

SQL>
SQL> select id_col, product_name
  2  from my_large_replica
  3  where factory = 'Outer Space'
  4  /

    ID_COL
----------
PRODUCT_NAME
--------------------------------------------------------------------------------
     74000
Revolutionary Pin


SQL>
SQL> select count(*) from hemant.mlog$_my_large_source;

  COUNT(*)
----------
         0

SQL>

Instead of rebuilding / repopulating the Replica Table with all 72,448 rows, I used the MV definition and the MV Log on the Source Table to copy over that 1 new row.

The above demonstration is against 19c.

Here are two older posts, one in March 2009 and the other in January 2012 on an earlier release of Oracle.

A Materialized View can be created with all the target rows pre-inserted (and subsequently refreshed for changes).  This is the default behaviour.

However, it is possible to define a Materialized View without actually populating it.

You might want to take such a course of action for scenarios like :

1.  Building a number of Materialized Views along with a code migration but not wanting to spend time that would be required to actually populate the MVs  and deferring the population to a subsequent maintenance window after which the code and data will be referenced by the application/users

2.  Building a number of MVs in a Tablespace that is initially small but will be enlarged in the maintenance window to handle the millions of rows that will be inserted

3.  Building an MV definition without actually having all the "clean" Source Table(s) rows currently available, deferring the cleansing of data to a later date and then populating the MV after the cleansing

The BUILD DEFERRED clause comes in handy here.


Let's say that we have a NEW_SOURCE_TABLE (with many rows and/or with rows that are yet to be cleansed) and want to build an "empty" MV on it  (OR that this MV is one of a number of MVs that are being built together simply for migration of dependent code, without the data).

SQL> desc new_source_table
 Name                                      Null?    Type
 ----------------------------------------- -------- ----------------------------
 ID                                        NOT NULL NUMBER
 DATA_ELEMENT_1                                     VARCHAR2(15)
 DATA_ELEMENT_2                                     VARCHAR2(15)
 DATE_COL                                           DATE

SQL>
SQL> create materialized view log on new_source_table;
create materialized view log on new_source_table
*
ERROR at line 1:
ORA-12014: table 'NEW_SOURCE_TABLE' does not contain a primary key constraint


SQL> create materialized view log on new_source_table with rowid;

Materialized view log created.

SQL>
SQL> create materialized view new_mv
  2  build deferred
  3  refresh with rowid
  4  as select id as id_number,
  5  data_element_1 as data_key,
  6  data_element_2 as data_val,
  7  date_col as data_date
  8  from new_source_table
  9  /

Materialized view created.

SQL>

Notice that my Source Table currently does not have a Primary Key.  The MV Log can be created with the "WITH ROWID" clause in the absence of the Primary Key.
The Materialized View is also built with the ROWID as the Refresh cannot use a Primary Key.
Of course, you may well have a Source Table with a Primary Key.  In that case, you can continue to default using the Primary Key instead of the ROWID

Once the Source Table is properly populated / cleansed and/or the tablespace containing the MV is large enough, the MV is first refreshed with a COMPLETE Refresh and subsequently with FAST Refresh's.

SQL> select count(*) from new_source_table;

  COUNT(*)
----------
       106

SQL> execute dbms_mview.refresh('NEW_MV','C',atomic_refresh=>FALSE);

PL/SQL procedure successfully completed.

SQL>

Subsequently, when one or more rows are inserted/updated in the Source Table, the next Refresh is a Fast Refresh.

SQL> execute dbms_mview.refresh('NEW_MV','F');

PL/SQL procedure successfully completed.

SQL>
SQL> select mview_name, refresh_mode, refresh_method, last_refresh_type
  2  from user_mviews
  3  where mview_name = 'NEW_MV'
  4  /

MVIEW_NAME         REFRESH_M REFRESH_ LAST_REF
------------------ --------- -------- --------
NEW_MV             DEMAND    FORCE    FAST

SQL>

Thus, we started off with an empty MV and later used REFRESHs (COMPLETE and FAST) to populate it.

So far, in previous posts in this series, I have demonstrated Materialized Views that set to REFRESH ON DEMAND.

You can also define a Materialized View that is set to REFRESH ON COMMIT -- i.e. every time DML against the Source Table is committed, the MV is also immediately updated.  Such an MV must be in the same database  (you cannot define an ON COMMIT Refresh across two databases  -- to do so, you have to build your own replication code, possibly using Database Triggers or external methods of 2-phase commit).

Here is a quick demonstration, starting with a Source Table in the HEMANT schema and then building a FAST REFRESH MV in the HR schema.

SQL> show user
USER is "HEMANT"
SQL> create table hemant_source_tbl (id_col number not null primary key, data_col varchar2(30));

Table created.

SQL> grant select on hemant_source_tbl to hr;

Grant succeeded.

SQL> create materialized view log on hemant_source_tbl;

Materialized view log created.

SQL> grant select on mlog$_hemant_source_tbl to hr;

Grant succeeded.

SQL>
SQL> grant create materialized view to hr;

Grant succeeded.

SQL> grant on commit refresh on hemant_source_tbl to hr;

Grant succeeded.

SQL>
SQL> grant on commit refresh on mlog$_hemant_source_tbl to hr;

Grant succeeded.

SQL>

Note : I had to grant the CREATE MATERIALIZED VIEW privilege to HR for this test case. Also, as the MV is to Refresh ON COMMIT, two additional object-level grants on the Source Table and the Materialized View Log are required as the Refresh is across schemas.

SQL> connect hr/HR@orclpdb1
Connected.
SQL> create materialized view hr_mv_on_commit
  2  refresh fast on commit
  3  as select id_col as primary_key_col, data_col as value_column
  4  from hemant.hemant_source_tbl;

Materialized view created.

SQL>

Now that the Materialized View is created successfully, I will test DML against the table and check that an explicit REFRESH call (e.g. DBMS_MVIEW.REFRESH or DBMS_REFRESH.REFRESH) is not required.

SQL> connect hemant/hemant@orclpdb1
Connected.
SQL> insert into hemant_source_tbl values (1,'First');

1 row created.

SQL> commit;

Commit complete.

SQL> select * from hr.hr_mv_on_commit;

PRIMARY_KEY_COL VALUE_COLUMN
--------------- ------------------------------
              1 First

SQL> connect hr/HR@orclpdb1
Connected.
SQL> select * from hr_mv_on_commit;

PRIMARY_KEY_COL VALUE_COLUMN
--------------- ------------------------------
              1 First

SQL>

The Materialized View in the HR schema was refreshed immediately, without an explicit REFRESH call.

Remember : An MV that is to REFRESH ON COMMIT must be in the same database as the Source Table.

So far, all my blog posts in this series cover "single" Materialized Views (even if I have created two MVs, they are independent of each other and can be refreshed at different schedules).

A Refresh Group is what you would define if you want multiple MVs to be refreshed to the same point in time.  This allows for
(a) data from transaction that touch multiple tables
or
(b) views of multiple tables
to be consistent in the target MVs.

For example, if you have SALES_ORDER and LINE_ITEMS tables and the MVs on these are refreshed at different times, you might see the ORDER (Header) without the LINE_ITEMs (or, worse, in the absence of Referential Integrity constraints, LINE_ITEMs without the ORDER (Header) !).

Here's a demo, using the HR  DEPARTMENTS and EMPLOYEES table with corresponding MVs built in the HEMANT schema.

SQL> show user
USER is "HR"
SQL> select count(*) from departments;

  COUNT(*)
----------
        27

SQL> select count(*) from employees;

  COUNT(*)
----------
       107

SQL> 
SQL> grant select on departments to hemant;

Grant succeeded.

SQL> grant select on employees to hemant;

Grant succeeded.

SQL>
SQL> create materialized view log on departments;

Materialized view log created.

SQL> grant select, delete on mlog$_departments to hemant;

Grant succeeded.

SQL>
SQL> create materialized view log on employees;

Materialized view log created.

SQL> grant select, delete on mlog$_employees to hemant;

Grant succeeded.

SQL>
SQL>

Having created the source MV Logs  note that I have to grant privileges to the account (HEMANT) that will be reading and deleting from the MV Logs.

Next, I setup the MVs and the Refresh Group

SQL> show user
USER is "HEMANT"
SQL>
SQL> select count(*) from hr.departments;

  COUNT(*)
----------
        27

SQL> select count(*) from hr.employees;

  COUNT(*)
----------
       107

SQL>
SQL>
SQL> create materialized view mv_dept
  2  refresh fast on demand
  3  as select department_id as dept_id, department_name as dept_name
  4  from hr.departments
  5  /

Materialized view created.

SQL>
SQL> create materialized view mv_emp
  2  refresh fast on demand
  3  as select department_id as dept_id, employee_id as emp_id,
  4  first_name, last_name, hire_date
  5  from hr.employees
  6  /

Materialized view created.

SQL>
SQL> select count(*) from mv_dept;

  COUNT(*)
----------
        27

SQL> select count(*) from mv_emp;

  COUNT(*)
----------
       107

SQL>
SQL> execute dbms_refresh.make(-
> name=>'HR_MVs',-
> list=>'MV_DEPT,MV_EMP',-
> next_date=>sysdate+0.5,-
> interval=>'sysdate+1');

PL/SQL procedure successfully completed.

SQL>
SQL> commit;

Commit complete.

SQL>

Here, I have built two MVs and then a Refresh Group called "HR_MVS".  The first refresh will be 12hours from now and every subsequent refresh will be after 24hours.  (The Refresh Interval must be set to what would be larger than the time taken to execute the actual Refresh).

However, I can manually execute the Refresh after new rows are populated into the source tables. First, I insert new rows

SQL> show user
USER is "HR"
SQL> insert into departments (department_id, department_name)
  2  values
  3  (departments_seq.nextval, 'New Department');

1 row created.

SQL> select department_id
  2  from departments
  3  where department_name = 'New Department';

DEPARTMENT_ID
-------------
          280

SQL> insert into employees(employee_id, first_name, last_name, email, hire_date, job_id, department_id)
  2  values
  3  (employees_seq.nextval, 'Hemant', 'Chitale', 'hkc@myenterprise.com', sysdate, 'AD_VP', 280);

1 row created.

SQL> select employee_id
  2  from employees
  3  where first_name = 'Hemant';

EMPLOYEE_ID
-----------
        208

SQL> commit;

Commit complete.

SQL>

Now that there are new rows, the target MVs must be refreshed together.

SQL> connect hemant/hemant@orclpdb1
Connected.
SQL> execute dbms_refresh.refresh('HR_MVS');

PL/SQL procedure successfully completed.

SQL> select count(*) from mv_dept;

  COUNT(*)
----------
        28

SQL> select count(*) from mv_emp;

  COUNT(*)
----------
       108

SQL>
SQL> select * from mv_dept
  2  where dept_id=280;

   DEPT_ID DEPT_NAME
---------- ------------------------------
       280 New Department

SQL> select * from mv_emp
  2  where emp_id=208;

   DEPT_ID     EMP_ID FIRST_NAME           LAST_NAME                 HIRE_DATE
---------- ---------- -------------------- ------------------------- ---------
       280        208 Hemant               Chitale                   12-OCT-19

SQL>

Both MVs have been Refresh'd together as an ATOMIC Transaction.  If either of the two MVs had failed to refresh (e.g. unable to allocate extent to grow the MV), both the INSERTs would be rolled back.  (Note : It is not a necessary requirement that both source tables have new / updated rows, the Refresh Group works even if only one of the two tables has new / updated rows).

Note : I have used DBMS_REFRESH.REFRESH (instead of DBMS_MVIEW.REFRESH) to execute the Refresh.

You can build multiple Refresh Groups, each consisting of *multiple* Source Tables from the same source database.
You would define each Refresh Group to maintain consistency of data across multiple MVs (sourced from different tables).
Besides the Refresh Group on two HR tables, I could have, within the HEMANT schema, more Refresh Groups on FINANCE schema tables as well.

(Can you have a Refresh Group sourcing from tables from different schemas ?  Try that out !)


What's the downside of Refresh Groups ?    
Undo and Redo !  Every Refresh consists of INSERT/UPDATE/DELETE operations on the MVs.  And if any one of the MVs fails to Refresh, the entire set of DMLs (across all the MVs in the Refresh Group) has to *Rollback* !

Till 9i, if you did a COMPLETE Refresh of a Single Materialized View, Oracle would do a TRUNCATE followed by an INSERT.
If you did a COMPLETE Refresh of a *group* of Materialized Views, Oracle would execute DELETE and INSERT so that all the MVs would be consistent to the same point in time.  Thus, if one of the MVs failed to refresh (e.g. the SELECT on the Source Table failed or the INSERT failed, it would be able to do a ROLLBACK of all the MVs to revert them to the status (i.e. all rows that were present) as of the time before the Refresh began.  This would also allow all MVs to be available for queries with the rows as of before the Refresh began, even as the Refresh was running (because the Refresh of the multiple MVs was a single transaction).

In 10g, the behaviour for a *group* of Materialized Views remained the same.  However, for a single MV, the default was now to do a DELETE and INSERT as well.  This would allow the MV to be queryable as well while the Refresh was running.
This change came as a surprise to many customers (including me at a site where I was managing multiple single MVs) !
This change meant that the single MV took longer to run (because DELETEing all the rows takes a long time !) and required much more Undo and Redo space !!

Here's a demonstration in 19c (as in the previous posts in this series) :

First, I start with a new, larger, Source Table  and then build two MVs on it :

SQL> create table source_table_2
  2  as select *
  3  from dba_objects
  4  where object_id is not null
  5  /

Table created.

SQL> alter table source_table_2
  2  add constraint source_table_2_pk
  3  primary key (object_id)
  4  /

Table altered.

SQL> select count(*)
  2  from source_table_2
  3  /

  COUNT(*)
----------
     72366

SQL>
SQL> create materialized view new_mv_2_1
  2  as select object_id, owner, object_name, object_type
  3  from source_table_2
  4  /

Materialized view created.

SQL> create materialized view new_mv_2_2
  2  as select object_id, owner, object_name, object_type
  3  from source_table_2
  4  /

Materialized view created.

SQL>
SQL> select mview_name, refresh_mode, refresh_method, last_refresh_type, fast_refreshable
  2  from user_mviews
  3  where mview_name like 'NEW_MV%'
  4  order by 1
  5  /

MVIEW_NAME       REFRESH_M REFRESH_ LAST_REF FAST_REFRESHABLE
---------------- --------- -------- -------- ------------------
NEW_MV_2_1       DEMAND    FORCE    COMPLETE DIRLOAD_DML
NEW_MV_2_2       DEMAND    FORCE    COMPLETE DIRLOAD_DML

SQL>

Note that it *IS* possible to have two Materialized Views with exactly the same QUERY co-existing.  They may have different REFRESH_METHODs (here both are the same) and/or may have different frequencies of Refresh calls when the REFRESH_MODE is 'DEMAND'

Note also that I did not specify any "refresh on demand" clause so both defaulted to FORCE and DEMAND.

(Question 1 : Why might I have two MVs with the same QUERY and the same REFRESH_METHOD but different frequency or different times when the Refresh is called ?)

(Question 2 : What is DIRLOAD_DML ?)


Now, let me issue two different COMPLETE Refresh calls and trace them.

SQL> execute dbms_mview.refresh('NEW_MV_2_1','C');
SQL> execute dbms_mview.refresh('NEW_MV_2_2','C',atomic_refresh=>FALSE);  -- from a different session

Now, I look at the trace files.

For the NEW_MV_2_1  (where ATOMIC_REFRESH defaulted to TRUE), I see :

/* MV_REFRESH (DEL) */ delete from "HEMANT"."NEW_MV_2_1"

/* MV_REFRESH (INS) */INSERT /*+ BYPASS_RECURSIVE_CHECK */ INTO "HEMANT"."NEW_MV_2_1"("OBJECT_ID","OWNER","OBJECT_NAME","OBJECT_TYPE") SELECT "SOURCE_TABLE_2"."OBJECT_ID","SOURCE_TABLE_2"."OWNER","SOURCE_TABLE_2"."OBJECT_NAME","SOURCE_TABLE_2"."OBJECT_TYPE" FROM "SOURCE_TABLE_2" "SOURCE_TABLE_2"

And for the NEW_MV_2_2 (where ATOMIC_REFRESH was set to FALSE), I see :

LOCK TABLE "HEMANT"."NEW_MV_2_2" IN EXCLUSIVE MODE  NOWAIT

/* MV_REFRESH (DEL) */ truncate table "HEMANT"."NEW_MV_2_2" purge snapshot log

/* MV_REFRESH (INS) */INSERT /*+ BYPASS_RECURSIVE_CHECK APPEND SKIP_UNQ_UNUSABLE_IDX */ INTO "HEMANT"."NEW_MV_2_2"("OBJECT_ID","OWNER","OBJECT_NAME","OBJECT_TYPE") SELECT "SOURCE_TABLE_2"."OBJECT_ID","SOURCE_TABLE_2"."OWNER","SOURCE_TABLE_2"."OBJECT_NAME","SOURCE_TABLE_2"."OBJECT_TYPE" FROM "SOURCE_TABLE_2" "SOURCE_TABLE_2"

So, the default ATOMIC_REFRESH=TRUE caused a DELETE followed by an INSERT while the ATOMIC_REFRESH=FALSE caused a DELETE followed by an INSERT APPEND (a Direct Path Insert).  The second method is much faster.



More information from a tkprof for the NEW_MV_2_1 (ATOMIC_REFRESH=TRUE) is :

INSERT INTO "HEMANT"."NEW_MV_2_1"("OBJECT_ID","OWNER","OBJECT_NAME",
  "OBJECT_TYPE") SELECT "SOURCE_TABLE_2"."OBJECT_ID","SOURCE_TABLE_2"."OWNER",
  "SOURCE_TABLE_2"."OBJECT_NAME","SOURCE_TABLE_2"."OBJECT_TYPE" FROM
  "SOURCE_TABLE_2" "SOURCE_TABLE_2"


call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.01       0.01          0         66          0           0
Execute      0      0.00       0.00          0          0          0           0
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        1      0.01       0.01          0         66          0           0




delete from "HEMANT"."NEW_MV_2_1"


call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.00       0.00          0          0          0           0
Execute      1      1.47       1.77        151        173     224377       72366
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        2      1.47       1.77        151        173     224377       72366

Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 106     (recursive depth: 1)
Number of plan statistics captured: 1

Rows (1st) Rows (avg) Rows (max)  Row Source Operation
---------- ---------- ----------  ---------------------------------------------------
         0          0          0  DELETE  NEW_MV_2_1 (cr=178 pr=151 pw=0 time=1783942 us starts=1)
     72366      72366      72366   INDEX FAST FULL SCAN SYS_C_SNAP$_82SOURCE_TABLE_2_PK (cr=157 pr=150 pw=0 time=54982 us starts=1 cost=42 size=361830 card=72366)(object id 73111)




INSERT /*+ BYPASS_RECURSIVE_CHECK */ INTO "HEMANT"."NEW_MV_2_1"("OBJECT_ID",
  "OWNER","OBJECT_NAME","OBJECT_TYPE") SELECT "SOURCE_TABLE_2"."OBJECT_ID",
  "SOURCE_TABLE_2"."OWNER","SOURCE_TABLE_2"."OBJECT_NAME",
  "SOURCE_TABLE_2"."OBJECT_TYPE" FROM "SOURCE_TABLE_2" "SOURCE_TABLE_2"


call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.00       0.00          0          0          4           0
Execute      1      0.71       0.71          0       2166     152128       72366
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        2      0.71       0.71          0       2166     152132       72366

Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 106     (recursive depth: 1)
Number of plan statistics captured: 1

Rows (1st) Rows (avg) Rows (max)  Row Source Operation
---------- ---------- ----------  ---------------------------------------------------
         0          0          0  LOAD TABLE CONVENTIONAL  NEW_MV_2_1 (cr=2257 pr=0 pw=0 time=723103 us starts=1)
     72366      72366      72366   TABLE ACCESS FULL SOURCE_TABLE_2 (cr=1410 pr=0 pw=0 time=30476 us starts=1 cost=392 size=3980130 card=72366)

Note that the first INSERT was only Parsed but *not* Executed.


While that for NEW_MV_2_2 (ATOMIC_REFRESH=FALSE) shows :

LOCK TABLE "HEMANT"."NEW_MV_2_2" IN EXCLUSIVE MODE  NOWAIT


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          0          0           0
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        2      0.00       0.00          0          0          0           0




truncate table "HEMANT"."NEW_MV_2_2" purge snapshot log



call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.00       0.00          0          0          1           0
Execute      1      0.06       0.56         13         15        511           0
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        2      0.06       0.57         13         15        512           0



INSERT /*+ BYPASS_RECURSIVE_CHECK APPEND SKIP_UNQ_UNUSABLE_IDX */ INTO
  "HEMANT"."NEW_MV_2_2"("OBJECT_ID","OWNER","OBJECT_NAME","OBJECT_TYPE")
  SELECT "SOURCE_TABLE_2"."OBJECT_ID","SOURCE_TABLE_2"."OWNER",
  "SOURCE_TABLE_2"."OBJECT_NAME","SOURCE_TABLE_2"."OBJECT_TYPE" FROM
  "SOURCE_TABLE_2" "SOURCE_TABLE_2"


call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.01       0.09          0         43          0           0
Execute      1      0.22       0.56          3       1487       1121       72366
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        2      0.24       0.65          3       1530       1121       72366

Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 106     (recursive depth: 1)
Number of plan statistics captured: 1

Rows (1st) Rows (avg) Rows (max)  Row Source Operation
---------- ---------- ----------  ---------------------------------------------------
         0          0          0  LOAD AS SELECT  NEW_MV_2_2 (cr=3688 pr=7 pw=586 time=953367 us starts=1)
     72366      72366      72366   OPTIMIZER STATISTICS GATHERING  (cr=3337 pr=0 pw=0 time=142500 us starts=1 cost=392 size=3980130 card=72366)
     72366      72366      72366    TABLE ACCESS FULL SOURCE_TABLE_2 (cr=1410 pr=0 pw=0 time=40841 us starts=1 cost=392 size=3980130 card=72366)




ALTER INDEX "HEMANT"."SYS_C_SNAP$_83SOURCE_TABLE_2_PK" REBUILD  NOPARALLEL


call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.00       0.08          0          1          1           0
Execute      1      0.11       0.48        586        626        680           0
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        2      0.11       0.56        586        627        681           0

Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 106     (recursive depth: 2)
Number of plan statistics captured: 1

Rows (1st) Rows (avg) Rows (max)  Row Source Operation
---------- ---------- ----------  ---------------------------------------------------
         1          1          1  INDEX BUILD UNIQUE SYS_C_SNAP$_83SOURCE_TABLE_2_PK (cr=632 pr=586 pw=150 time=392351 us starts=1)(object id 0)
     72366      72366      72366   SORT CREATE INDEX (cr=590 pr=586 pw=0 time=148023 us starts=1)
     72366      72366      72366    MAT_VIEW ACCESS FULL NEW_MV_2_2 (cr=590 pr=586 pw=0 time=86149 us starts=1 cost=166 size=361830 card=72366)

So, the ATOMIC_REFRESH=FALSE caused
a. TRUNCATE
b. INSERT APPEND (i.e. Direct Path Insert, minimizing Undo and reducing Redo)
c. REBUILD INDEX

I am not comparing Execution Time for the two Refresh's.  I would rather that you focus on the fact that the DELETE (in ATOMIC_REFRESH=TRUE) can be very expensive (think Undo and Redo) when it has delete, say, millions of rows.  Also, that the INSERT is a regular operation that also causes Undo and Redo to be generated.

ATOMIC_REFRESH=FALSE makes a significant difference to the Undo and Redo generation and will be faster for large Materialized Views.

What is the downside of ATOMIC_REFRESH=FALSE ?  Firstly, the MV will not present any rows to a query that executes against it while the Refresh is in progress.  Secondly, if the Refresh fails, the MV is left in a Truncated state (without rows) until another Refresh is executed.
The ATOMIC_REFRESH=TRUE avoids  these two pitfalls, at the expense of resources (Undo and Redo) and time to refresh.

For more information, see Oracle Support Document #553464.1

In the previous blog post, I had a remark "We'll explore the implications of "REFRESH FAST" and just "REFRESH" alone in a subsequent blog post."

This is in the context of whether it is a FORCE or FAST that shows up as the REFRESH_METHOD.  A FORCE attempts a FAST and, if it can't do so (e.g. the Materialized View Log is not accessible), attempts a COMPLETE Refresh from all the rows of the Source Table.

Other than a MV Log being a requirement, there are constraints on which types of Materialized Views can do a FAST Refresh.

SQL> create materialized view mv_fast_not_possible
  2  refresh fast on demand
  3  as select id, data_element_2, sysdate
  4  from source_table
  5  /
as select id, data_element_2, sysdate
                              *
ERROR at line 3:
ORA-12015: cannot create a fast refresh materialized view from a complex query


SQL> !oerr ora 12015
12015, 00000, "cannot create a fast refresh materialized view from a complex query"
// *Cause:  Neither ROWIDs and nor primary key constraints are supported for
//          complex queries.
// *Action: Reissue the command with the REFRESH FORCE or REFRESH COMPLETE
//          option or create a simple materialized view.

SQL>

Thus, a "complex" query -- here one that adds a SYSDATE column -- cannot use a FAST Refresh.
(For all the restrictions, see Paragraph "5.3.8.4 General Restrictions on Fast Refresh" in the 19c documentation. )

SQL> create materialized view mv_fast_not_possible
  2  refresh force on demand
  3  as select id, data_element_2, sysdate
  4  from source_table
  5  /

Materialized view created.

SQL> select refresh_mode, refresh_method, last_refresh_type
  2  from user_mviews
  3  where mview_name  = 'MV_FAST_NOT_POSSIBLE'
  4  /

REFRESH_M REFRESH_ LAST_REF
--------- -------- --------
DEMAND    FORCE    COMPLETE

SQL>
SQL> insert into source_table
  2  values (2000,'TwoThousand','NewTwoTh',sysdate);

1 row created.

SQL> select * from source_table order by date_col ;

        ID DATA_ELEMENT_1  DATA_ELEMENT_2  DATE_COL
---------- --------------- --------------- ---------
       101 First           One             18-AUG-19
       103 Third           Three           18-AUG-19
       104 Fourth          Updated         09-SEP-19
         5 Fifth           Five            16-SEP-19
         6 Sixth           TwoHundred      19-SEP-19
         7 Seventh         ThreeHundred    19-SEP-19
      2000 TwoThousand     NewTwoTh        25-SEP-19

7 rows selected.

SQL>
SQL> commit;

Commit complete.

SQL> exec dbms_mview.refresh('MV_OF_SOURCE');

PL/SQL procedure successfully completed.

SQL> exec dbms_mview.refresh('MV_2');

PL/SQL procedure successfully completed.

SQL> exec dbms_mview.refresh('MV_FAST_NOT_POSSIBLE');

PL/SQL procedure successfully completed.

SQL>
SQL> select mview_name, refresh_mode,refresh_method,last_refresh_type, last_refresh_date
  2  from user_mviews
  3  order by last_refresh_date
  4  /

MVIEW_NAME            REFRESH_M REFRESH_ LAST_REF LAST_REFR
--------------------- --------- -------- -------- ---------
MV_OF_SOURCE          DEMAND    FAST     FAST     25-SEP-19
MV_2                  DEMAND    FORCE    FAST     25-SEP-19
MV_FAST_NOT_POSSIBLE  DEMAND    FORCE    COMPLETE 25-SEP-19

SQL>

MV_FAST_NOT_POSSIBLE will always undergo a COMPLETE Refresh using REFRESH_METHOD='FORCE'.

MV_2 has REFRESH_METHOD='FORCE' because it was created with "refresh on demand" with the "fast" keyword missing.  Nevertheless, it is a "simple" Materialized View so does a FAST Refresh.

MV_OF_SOURCE was created with "refresh fast on demand", so it is already configured as REFRESH_METHOD='FAST'

Now that we have two Materialized Views against a Source table, how can we identify the relationship via the data dictionary ?

This is the query to the data dictionary in the database where the Source Table exists :

SQL> l
  1  select v.owner MV_Owner, v.name MV_Name, v.snapshot_site, v.refresh_method,
  2  l.log_table MV_Log_Name, l.master MV_Source,
  3  to_char(l.current_snapshots,'DD-MON-RR HH24:MI:SS') Last_Refresh_Date
  4  from dba_registered_snapshots v, dba_snapshot_logs l
  5  where v.snapshot_id = l.snapshot_id
  6* and l.log_owner = 'HEMANT'
SQL> /

MV_OWNER MV_NAME          SNAPSHOT_SITE      REFRESH_MET MV_LOG_NAME        MV_SOURCE             LAST_REFRESH_DATE
-------- ---------------- ------------------ ----------- ------------------ --------------------- ------------------
HEMANT   MV_OF_SOURCE     ORCLPDB1           PRIMARY KEY MLOG$_SOURCE_TABLE SOURCE_TABLE          16-SEP-19 22:41:04
HEMANT   MV_2             ORCLPDB1           PRIMARY KEY MLOG$_SOURCE_TABLE SOURCE_TABLE          16-SEP-19 22:44:37

SQL>

I have run the query on the DBA_REGISTERED_SNAPSHOTS and DBA_SNAPSHOT_LOGS because the join on SNAPSHOT_ID is not available between DBA_REGISTERED_MVIEWS and DBA_MVIEW_LOGS.  Similarly, the CURRENT_SNAPSHOTS column is also not available in DBA_MVIEW_LOGS.  These two columns are important when you have *multiple* MViews against the same Source Table.

Note the "Snapshot_Site" is required because the Materialized View can be in a different database.  In this example, the MViews are in the same database as the Source Table. 

The target database containing the MViews will not have the Source Table "registered" into a data dictionary view.  The Source Table will be apparently from the QUERY column of DBA_MVIEWS (also, if the Source Table is in a different database, look at the MASTER_LINK column to identify the Database Link that connects to the source database).


UPDATE :  In case you are wondering what query you'd write against the database containing the Materialized View(s), you can simply query DBA_MVIEWS.

SQL> l
  1  select mview_name, query, master_link, refresh_mode, refresh_method,
  2  last_refresh_type, to_char(last_refresh_date,'DD-MON-RR HH24:MI:SS') Last_Refresh_Date
  3  from dba_mviews
  4  where owner = 'HEMANT'
  5* order by 1 desc
SQL> /

MVIEW_NAME
------------
QUERY
--------------------------------------------------------------------------------
MASTER_LINK  REFRESH_M REFRESH_ LAST_REF LAST_REFRESH_DATE
------------ --------- -------- -------- ---------------------------
MV_OF_SOURCE
SELECT "SOURCE_TABLE"."ID" "ID","SOURCE_TABLE"."DATA_ELEMENT_1" "DATA_ELEMENT_1"
,"SOURCE_TABLE"."DATA_ELEMENT_2" "DATA_ELEMENT_2","SOURCE_TABLE"."DATE_COL" "DAT
E_COL" FROM "SOURCE_TABLE" "SOURCE_TABLE"
             DEMAND    FAST     FAST     16-SEP-19 22:41:04

MV_2
select id, data_element_2
from source_table
             DEMAND    FORCE    FAST     16-SEP-19 22:44:37


SQL>

Here, the MASTER_LINK would specify the name of the Database Link used to connect to the Master (i.e. Source) table, if it was a different database.

REFRESH_MODE is ON DEMAND so that the MVs can be refreshed by either scheduled jobs or manually initiated calls -- as I've done in previous blog posts.  (The alternative can be ON COMMIT, if the Source Table and MV are in the same database).

LAST_REFRESH_TYPE is FAST, meaning that the refresh was able to use the MV Log on the Source Table to identify changes and merge them into the MV.  See the entries from the trace file that I've shown in the previous blog post.

Note the difference in the two REFRESH_METHOD values for the two MVs.
MV_OF_SOURCE was created as "refresh fast on demand" while "MV_2" was created as "refresh on demand".

We'll explore the implications of "REFRESH FAST" and just "REFRESH" alone in a subsequent blog post.

Question : Why does the QUERY look so different between MV_OF_SOURCE and MV_2 ?

Continuing the previous post, what happens when there is an UPDATE to the source table ?

SQL> select * from source_table;

        ID DATA_ELEMENT_1  DATA_ELEMENT_2  DATE_COL
---------- --------------- --------------- ---------
         1 First           One             18-AUG-19
         3 Third           Three           18-AUG-19
         4 Fourth          Four            18-AUG-19

SQL> select * from mlog$_source_table;

no rows selected

SQL> select * from rupd$_source_table;

no rows selected

SQL>
SQL> update source_table
  2  set data_element_2 = 'Updated', date_col=sysdate
  3  where id=4;

1 row updated.

SQL> select * from rupd$_source_table;

no rows selected

SQL> commit;

Commit complete.

SQL> select * from rupd$_source_table;

no rows selected

SQL> select * from mlog$_source_table;

        ID SNAPTIME$ D O
---------- --------- - -
CHANGE_VECTOR$$
--------------------------------------------------------------------------------
     XID$$
----------
         4 01-JAN-00 U U
18
8.4443E+14


SQL>

So, it is clear that UPDATES, too, go to the MLOG$ table.

What about multi-row operations ?

SQL> update source_table set id=id+100;

3 rows updated.

SQL> select * from rupd$_source_table;

no rows selected

SQL> select * from mlog$_source_table;

        ID SNAPTIME$ D O
---------- --------- - -
CHANGE_VECTOR$$
--------------------------------------------------------------------------------
     XID$$
----------
         4 01-JAN-00 U U
18
8.4443E+14

         1 01-JAN-00 D O
00
1.4075E+15

       101 01-JAN-00 I N
FF
1.4075E+15

         3 01-JAN-00 D O
00
1.4075E+15

       103 01-JAN-00 I N
FF
1.4075E+15

         4 01-JAN-00 D O
00
1.4075E+15

       104 01-JAN-00 I N
FF
1.4075E+15


7 rows selected.

SQL>

Wow ! Three rows updated in the Source Table translated to 6 rows in the MLOG$ table ! Each update row was represented by an DMLTYPE$$='D' and OLD_NEW$$='O'  followed by a DMLTYPE$$='I' and OLD_NEW$$='N'.   So that should mean "delete the old row from the materialized view and insert the new row into the materialized view" ??

(For the time being, we'll ignore SNAPTIME$$ being '01-JAN-00').

So an UPDATE to the Source Table of a Materialized View can be expensive during the UPDATE (as it creates two entries in the MLOG$ table) and for subsequent refresh's as well !

What happens when I refresh the Materialized View ?

SQL> execute dbms_session.session_trace_enable;

PL/SQL procedure successfully completed.

SQL> execute dbms_mview.refresh('MV_OF_SOURCE');

PL/SQL procedure successfully completed.

SQL> execute dbms_session.session_trace_disable;

PL/SQL procedure successfully completed.

SQL>

The session trace file shows these operations (I've excluded a large number of recursive SQLs and SQLs that were sampling the data for optimisation of execution plans):

update "HEMANT"."MLOG$_SOURCE_TABLE" 
set snaptime$$ = :1  
where snaptime$$ > to_date('2100-01-01:00:00:00','YYYY-MM-DD:HH24:MI:SS')

/* QSMQ VALIDATION */ ALTER SUMMARY "HEMANT"."MV_OF_SOURCE" COMPILE

select 1 from "HEMANT"."MLOG$_SOURCE_TABLE" 
where snaptime$$ > :1 
and ((dmltype$$ IN ('I', 'D')) or  (dmltype$$ = 'U' and old_new$$ in ('U', 'O') 
and   sys.dbms_snapshot_utl.vector_compare(:2, change_vector$$) = 1)) 
and rownum = 1

SELECT /*+ NO_MERGE(DL$) ROWID(MAS$) ORDERED USE_NL(MAS$) NO_INDEX(MAS$) PQ_DISTRIBUTE(MAS$,RANDOM,NONE) */ 
COUNT(*) cnt  
FROM ALL_SUMDELTA DL$, "HEMANT"."SOURCE_TABLE"  MAS$  
WHERE DL$.TABLEOBJ# = :1 AND DL$.TIMESTAMP > :2 AND DL$.TIMESTAMP <= :3 
AND  MAS$.ROWID BETWEEN DL$.LOWROWID AND DL$.HIGHROWID

select dmltype$$, count(*) cnt from "HEMANT"."MLOG$_SOURCE_TABLE"  
where snaptime$$ > :1 and snaptime$$ <= :2  
group by dmltype$$  order by dmltype$$

delete from "HEMANT"."MLOG$_SOURCE_TABLE" 
where snaptime$$ <= :1

and this being the refresh (merge update) of the target MV

DELETE FROM "HEMANT"."MV_OF_SOURCE" SNAP$ 
WHERE "ID" IN 
(SELECT * FROM (SELECT MLOG$."ID" 
  FROM "HEMANT"."MLOG$_SOURCE_TABLE" MLOG$ 
  WHERE "SNAPTIME$$" > :1 AND ("DMLTYPE$$" != 'I')) 
  AS OF SNAPSHOT(:B_SCN) )

/* MV_REFRESH (MRG) */ MERGE INTO "HEMANT"."MV_OF_SOURCE" "SNA$" USING 
(SELECT * FROM (SELECT CURRENT$."ID",CURRENT$."DATA_ELEMENT_1",CURRENT$."DATA_ELEMENT_2",CURRENT$."DATE_COL" 
FROM (SELECT "SOURCE_TABLE"."ID" "ID","SOURCE_TABLE"."DATA_ELEMENT_1" "DATA_ELEMENT_1","SOURCE_TABLE"."DATA_ELEMENT_2" "DATA_ELEMENT_2","SOURCE_TABLE"."DATE_COL" "DATE_COL" 
FROM "SOURCE_TABLE" "SOURCE_TABLE") CURRENT$, 
(SELECT DISTINCT MLOG$."ID" FROM "HEMANT"."MLOG$_SOURCE_TABLE" MLOG$ WHERE "SNAPTIME$$" > :1 
AND ("DMLTYPE$$" != 'D')) LOG$ WHERE CURRENT$."ID" = LOG$."ID") AS OF SNAPSHOT(:B_SCN) )"AV$" ON ("SNA$"."ID" = "AV$"."ID") 
WHEN MATCHED THEN UPDATE  SET "SNA$"."DATA_ELEMENT_1" = "AV$"."DATA_ELEMENT_1","SNA$"."DATA_ELEMENT_2" = "AV$"."DATA_ELEMENT_2","SNA$"."DATE_COL" = "AV$"."DATE_COL" 
WHEN NOT MATCHED THEN INSERT  (SNA$."ID",SNA$."DATA_ELEMENT_1",SNA$."DATA_ELEMENT_2",SNA$."DATE_COL") 
VALUES (AV$."ID",AV$."DATA_ELEMENT_1",AV$."DATA_ELEMENT_2",AV$."DATE_COL")

So, we see a large number of intensive operations against the MLOG$ Materialized View Log object.

And on the MV, there is a DELETE followed by a MERGE (UPDATE/IINSERT)


Two takeaways :
1.  Updating the Source Table of a Materialized View can have noticeable overheads
2.  Refreshing a Materialized View takes some effort on the part of the database

(Did you notice the strange year 2100 date in the update of the MLOG$ table?
.
.
.
.
.
.

A series of posts in 2017 on 12cR1 RAC :

1.  1. Grid  Infrastructure Install  (Dec-16)

2.  2. Convert AdminManaged Database to PolicyManaged   (Dec-16)

3.  3. Convert PolicyManaged Database back to AdminManaged   (Jan-17)

4.  4. Adding a Disk of a Different Size  (Jan-17)

5.  5. Relocating OCR and VoteDisk  (Jan-17)

6.  6. Running the Cluster Verification Utility  (Feb-17)

7.  7. OCR Commands  (Mar-17)

8.  8a. Setting Up Single Instance DG Standby for RAC  (Mar-17)

9.  8b. DUPLICATE DATABASE FOR STANDBY  (Mar-17)

10.  8c. Ignorable "Errors" During the DUPLICATE  (Mar-17)

11.  8d. Registering the two databases in DataGuard Broker  (Mar-17)

12.  8e. Redo Shipping and Apply (RAC to nonRAC)  (Apr-17)

13.  8f. Accessing Data in the PDB in the Standby  (Apr-17)

14.  8g. Switchover from RAC Primary to SingleInstance Standby  (Apr-17)

15.  8h. DataGuard Switchover (RAC to nonRAC) messages  (Apr-17)

16.  8i. Switchback from SingleInstance to RAC  (Apr-17)

17.  9. Adding a Service to a PDB in RAC  (Apr-17)

18.  10. Video on Database Startup  (May-17)

So, the 2 single-row INSERTs did create two entries in the MV Log and the REFRESH of the Materialized View did a cleanup of those two entries.

I haven't yet explored :
a.  UPDATEs
b. Multi-Row Operations

Basic Replication, starting with Read Only Snapshots has been available in Oracle since  V7.   This was doable with the "CREATE SNAPSHOT" command.

In 8i, the term was changed from "Snapshot" to "Materialized View"  and the "CREATE MATERIALIZED VIEW" command was introduced, while "CREATE SNAPSHOT" was still supported.

Just as CREATE SNAPSHOT is still available in 19c,  DBMS_SNAPSHOT.REFRESH is also available.


























Not that I recommend that you use CREATE SNAPSHOT and DBMS_SNAPSHOT anymore.  DBAs and Developers should have been using CREATE MATERIALIZED VIEW and DBMS_REFRESH since 8i.

In the next few blog posts (this will be a very short series) I will explore Basic Replication.  Let me know if you want to see it in 11.2 and 12c as well.

This blog has now achieved 2million PageViews :




(The "drop" at the end is the count for only the first week of August 2019).

Although this blog began in December 2006, the PageViews counts start with 8,176 in July 2010.  So, effectively, this blog has had 2million PageViews in 9years.

The first 1million PageViews were achieved in March 2015.

Unfortunately, the rate at which I have been publishing has declined since 2017 (36 posts in 2017, 30 in 2018 and only 8 so far this year).  I apologise for this.  Hopefully, I should be able to add more posts in the coming months.

A recent series of posts on Partitioning

1.  1.  Introduction   (Aug-18)

2.  2.  Simple Range Partitioning  (Aug-18)

3.  3a.  Indexes on a Partitioned Table  (Aug-18)

4.  3b.  More Indexes on a Partitioned Table  (Aug-18)

5.  3c.  Unique Index[es] on a Partitioned Table  (Aug-18)

6.  3d.  Partial Indexing (in 11g)  (Sep-18)

7.  4.  Row Movement  (Sep-18)

8.  5.  List Partitioning  (Sep-18)

9.  6.  Hash Partitioning  (Sep-18)

10.  7.  Interval Partitioning  (Oct-18)

11.  8.  Reference Partitioning  (Nov-18)

12.  9.  System Partitioning  (Nov-18)

13.  10.  Virtual Column Based Partitioning  (Nov-18)

14.  11.  Composite Partitioning  (Dec-18)

15.  12.  Data Dictionary Queries  (Dec-18)

16.  13a.  Relocating a Partition  (Dec-18)

17. 13b.  Splitting a Partition  (Jan-19)

18.  13c.  Merging Partitions  (Mar-19)

19.  13d.  TRUNCATE and DROP Partitions and Global Indexes  (Mar-19)

20.  14.  Converting a non-Partitioned Table to a Partitioned Table  (Mar-19)

21.  15.  Online Modification of Partitioning Type (Strategy)  (Apr-19)

22.  16.  Hybrid Partitioning  (May-19)

Oracle 19c introduces Hybrid Partitioning whereby you can have external and internal Partitions co-existing.  External Partitions are on storage (filesystem) outside the database.

Let's say we have a List Partitioned table for the widgets that we manufacture. The table is Partitioned by WIDGET_CLASS_ID, based on an ISO standard.  So all companies that manufacture widgets adopt the same WIDGET_CLASS_ID:

SQL> desc widgets_list
 Name                                      Null?    Type
 ----------------------------------------- -------- ----------------------------
 WIDGET_CLASS_ID                                    VARCHAR2(5)
 WIDGET_ID                                          VARCHAR2(32)
 WIDGET_NAME                                        VARCHAR2(32)
 WIDGET_DESCRIPTION                                 VARCHAR2(128)

SQL>
SQL> l
  1  select table_name, partitioning_type, partition_count
  2  from user_part_tables
  3* where table_name = 'WIDGETS_LIST'
SQL> /

TABLE_NAME                       PARTITION PARTITION_COUNT
-------------------------------- --------- ---------------
WIDGETS_LIST                     LIST                    3

SQL>
SQL> l
  1  select partition_name,high_value, num_rows
  2  from user_tab_partitions
  3* where table_name = 'WIDGETS_LIST'
SQL> /

PARTITION_NAME   HIGH_VALUE         NUM_ROWS
---------------- ---------------- ----------
P_A              'A'                    1520
P_B              'B'                     520
P_C              'C'                     119

SQL>

Later, another widget manufacturer that manufactures widgets of CLASS_ID 'X' is acquired.  The WIDGETS_LIST table is in a non-Oracle database and is received as a CSV file.  We accept the CSV file onto a filesystem location :

sh-4.2$ pwd
/home/oracle/ACQUIRED_COMPANY
sh-4.2$ cat AC_Widgets_List.CSV
'X','ABCXX2','The1','cddfdaxx'
'X','XXD2','The2','dda3'
'X','XRC34','The3','ff33355312'
sh-4.2$

So, we have a CSV file "AC_Widgets_List.CSV" listing the widgets manufactured by this company. We want to add it to our WIDGETS_LIST table.

Enter user-name: / as sysdba

Connected to:
Oracle Database 19c Enterprise Edition Release 19.0.0.0.0 - Production
Version 19.3.0.0.0

SQL> alter session set container=ORCLPDB1;

Session altered.

SQL> create directory acquired_company as '/home/oracle/ACQUIRED_COMPANY';

Directory created.

SQL> grant read, write on directory acquired_company to hemant;

Grant succeeded.

SQL>
SQL> connect hemant/hemant@ORCLPDB1
Connected.
SQL>
SQL> l
  1  alter table widgets_list
  2  add external partition attributes (
  3  type oracle_loader
  4  default directory acquired_company
  5  access parameters (
  6  fields terminated by ','
  7  (widget_class_id, widget_id, widget_name, widget_description)
  8  )
  9* )
SQL> /

Table altered.

SQL>
SQL> l
  1  alter table widgets_list
  2  add partition P_ACQ_CO values ('X')
  3* external location ('AC_Widgets_List.CSV')
SQL> /

Table altered.

SQL>
SQL> exec dbms_stats.gather_table_stats('','WIDGETS_LIST');

PL/SQL procedure successfully completed.

SQL>
SQL> l
  1  select partition_name, high_value, num_rows
  2  from user_tab_partitions
  3  where table_name = 'WIDGETS_LIST'
  4* order by partition_position
SQL> /

PARTITION_NAME                   HIGH_VALUE         NUM_ROWS
-------------------------------- ---------------- ----------
P_A                              'A'                    1520
P_B                              'B'                     520
P_C                              'C'                     119
P_ACQ_CO                         'X'                       3

SQL>
SQL> l
  1* select * from widgets_list partition (P_ACQ_CO)
SQL> /

WIDGET_CLASS WIDGET_ID                        WIDGET_NAME
------------ -------------------------------- --------------------------------
WIDGET_DESCRIPTION
--------------------------------------------------------------------------------
'X'          'ABCXX2'                         'The1'
'cddfdaxx'

'X'          'XXD2'                           'The2'
'dda3'

'X'          'XRC34'                          'The3'
'ff33355312'


SQL>

The rows in the "AC_Widgets_List.CSV" file are now visible as rows in a *Partition* in our Oracle Table WIDGETS_LIST.
Of course, these being external, cannot be modified by INSERT/UPDATE/DELETE DML.

The External Attribute Type that I used is ORACLE_LOADER to use the SQL Loader libraries on a filesystem file.  Oracle 19c also supports ORACLE_DATAPUMP, ORACLE_HDFS and ORACLE_HIVE to reference files stored in other types of storage.

Hybrid Partitions are supported with single-level Range and List partitioning methods.  ALTER TABLE to ADD, DROP and RENAME Partitions is supported.

An External Partition can be Exchanged with an External Non-Partitioned Table only.
.
.
UPDATE :  Later, if I update the CSV file (using an external editor) to remove the quotation mark :

sh-4.2$ cat AC_Widgets_List.CSV
X,ABCXX2,The1,cddfdaxx
X,XXD2,The2,dda3
X,XRC34,The3,ff33355312
sh-4.2$

SQL> l
  1* select * from widgets_list partition (P_ACQ_CO)
SQL> /

WIDGET_CLASS_ID  WIDGET_ID                        WIDGET_NAME
---------------- -------------------------------- --------------------------------
WIDGET_DESCRIPTION
------------------------------------------------------------------------------------
X                ABCXX2                           The1
cddfdaxx

X                XXD2                             The2
dda3

X                XRC34                            The3
ff33355312


SQL>

So, it is possible to edit the External Partition using other methods (here I used "vi" on Linux)
.
.
.

Oracle 18c introduces the ability to convert a Partitioned Table from one Type to another -- e.g. from Hash Partitioning to Range Partitioning.  This is effectively a change of the Partitioning strategy for a table without actually having to manually rebuild the table.

I start with a Hash Partitioned Table.

SQL> create table customers(customer_id number, customer_name varchar2(200), customer_city_code number)
  2  partition by hash (customer_id) partitions 4;

Table created.

SQL> select partitioning_type from user_part_tables
  2  where table_name = 'CUSTOMERS'
  3  /

PARTITION
---------
HASH

SQL> select partition_name from user_tab_partitions
  2  where table_name = 'CUSTOMERS'
  3  /

PARTITION_NAME
--------------------------------------------------------------------------------
SYS_P221
SYS_P222
SYS_P223
SYS_P224

SQL>
SQL> insert into  customers
  2  select dbms_random.value(1,1000001), dbms_random.string('X',25), mod(rownum,5)
  3  from dual
  4  connect by level < 1000001
  5  /

1000000 rows created.

SQL> commit;

Commit complete.

SQL> exec dbms_stats.gather_table_stats('','CUSTOMERS');

PL/SQL procedure successfully completed.

SQL> select partition_name, num_rows
  2  from user_tab_partitions
  3  where table_name = 'CUSTOMERS'
  4  /

PARTITION_NAME     NUM_ROWS
---------------- ----------
SYS_P221             250090
SYS_P222             249563
SYS_P223             250018
SYS_P224             250329

SQL>

I now want to convert this Hash Partitioned Table to a Range Partitioned Table online.

SQL> alter table customers
  2  modify
  3  partition by range (customer_id)
  4  (partition P_100K values less than (100001),
  5   partition P_200K values less than (200001),
  6   partition P_300K values less than (300001),
  7   partition P_400K values less than (400001),
  8   partition P_500K values less than (500001),
  9   partition P_600K values less than (600001),
 10   partition P_700K values less than (700001),
 11   partition P_800K values less than (800001),
 12   partition P_900K values less than (900001),
 13   partition P_1MIL values less than (1000001),
 14   partition P_2MIL values less than (2000001),
 15   partition P_MAXVALUE values less than (MAXVALUE))
 16  online;

Table altered.

SQL>
SQL> select partitioning_type
  2  from user_part_tables
  3  where table_name = 'CUSTOMERS'
  4  /

PARTITION
---------
RANGE

SQL> exec dbms_stats.gather_table_stats('','CUSTOMERS');

PL/SQL procedure successfully completed.

SQL> col high_value format a12
SQL> select partition_name, high_value, num_rows
  2  from user_tab_partitions
  3  where table_name = 'CUSTOMERS'
  4  order by partition_position
  5  /

PARTITION_NAME   HIGH_VALUE     NUM_ROWS
---------------- ------------ ----------
P_100K           100001           100116
P_200K           200001            99604
P_300K           300001            99941
P_400K           400001           100048
P_500K           500001            99841
P_600K           600001            99920
P_700K           700001           100081
P_800K           800001           100024
P_900K           900001           100123
P_1MIL           1000001          100302
P_2MIL           2000001               0
P_MAXVALUE       MAXVALUE              0

12 rows selected.

SQL>

The Hash Partitioned Table is now converted to a Range Partitioned Table.  The number of Partitions has been changed.  And the operation was performed online with the ONLINE keyword added to the ALTER TABLE ... statement.  The UPDATE INDEXES clauses can also be used to update existing Indexes on the Table.

An excellent article that anyone promising, developing, maintaining or using any system that is non-trivial should read :

https://embeddedartistry.com/blog/2019/4/1/what-can-software-organizations-learn-from-the-boeing-737-max-saga

Pre-12cRelease2, there were only three methods to convert a non-Partitioned Table to a Partitioned Table

(a) Create a new, empty, Partitioned Table and copy (using INSERT .... AS SELECT ... ) all the data from the non-Partitioned Table to the new, Partitioned Table (and subsequently rename the new Partitioned Table after renaming or dropping the old non-Partitioned Table)

(b) Create a new, empty, Partitioned Table and use EXCHANGE PARTITION to switch the non-Partitioned Table into the Partitioned Table (and then run subsequent SPLIT PARTITION or ADD PARTITION commands as needed to create the additional Partitions)

(c) Create an interim Partitioned Table and use DBMS_REDEFINITION to do an online copy of the data to the interim Partitioned Table and automatically switch the name at the end


12.2 introduced the ability to use ALTER TABLE  ... MODIFY PARTITION ... to convert a non-Partitioned Table to a Partitioned Table

I start with a non-Partitioned Table :

SQL> select table_name, partitioned
  2  from user_tables
  3  where table_name = 'SALES_DATA_NONPARTITIONED'
  4  /

TABLE_NAME                     PAR
------------------------------ ---
SALES_DATA_NONPARTITIONED      NO

SQL> select index_name, uniqueness, partitioned
  2  from user_indexes
  3  where table_name = 'SALES_DATA_NONPARTITIONED'
  4  /

INDEX_NAME                     UNIQUENES PAR
------------------------------ --------- ---
SALES_DATA_UK                  UNIQUE  NO

SQL> 

I then convert it to a Range-Partitioned Table.

SQL> alter table sales_data_nonpartitioned
  2  modify 
  3  partition by range (sale_date)
  4  (
  5  partition P_2015 values less than (to_date('01-JAN-2016','DD-MON-YYYY')),
  6  partition P_2016 values less than (to_date('01-JAN-2017','DD-MON-YYYY')),
  7  partition P_2017 values less than (to_date('01-JAN-2018','DD-MON-YYYY')),
  8  partition P_2018 values less than (to_date('01-JAN-2019','DD-MON-YYYY')),
  9  partition P_2019 values less than (to_date('01-JAN-2020','DD-MON-YYYY')),
 10  partition p_MAXVALUE values less than (MAXVALUE) 
 11  )
 12  online 
 13  update indexes
 14  /

Table altered.

SQL> 
SQL> alter table sales_data_nonpartitioned rename to sales_data;

Table altered.

SQL>
SQL> select partition_name, high_value
  2  from user_tab_partitions
  3  where table_name = 'SALES_DATA'
  4  order by partition_position
  5  /

PARTITION_NAME   HIGH_VALUE
---------------- --------------------------
P_2015           TO_DATE(' 2016-01-01 00:00
P_2016           TO_DATE(' 2017-01-01 00:00
P_2017           TO_DATE(' 2018-01-01 00:00
P_2018           TO_DATE(' 2019-01-01 00:00
P_2019           TO_DATE(' 2020-01-01 00:00
P_MAXVALUE       MAXVALUE

6 rows selected.

SQL> 
SQL> select index_name, partitioned, uniqueness, status
  2  from user_indexes
  3  where table_name = 'SALES_DATA'
  4  /

INDEX_NAME                     PAR UNIQUENES STATUS
------------------------------ --- --------- --------
SALES_DATA_UK                  NO  UNIQUE    VALID

SQL> 

The SALES_DATA_NONPARTITIONED was converted to a Range Partitioned Table.  If I didn't have to rename the table (e.g. if the table name was actually, properly SALES_DATA only), then there would be no need to lock the table as the RENAME command does.

A TRUNCATE or DROP Partition makes Global Indexes on a Partitioned Table UNUSABLE.

You may be lucky if the target partition was empty, resulting in Oracle maintaining Global Indexes as valid.  However, the accepted rule is that you either (a) use the UPDATE INDEXES clause [resulting in the TRUNCATE or DROP taking longer to run, effectively locking the table partitions] OR  (b) do a REBUILD of the Indexes that become UNUSABLE after the TRUNCATE or DROP.

12c has introduced what it calls Asynchronous Global Index Maintenance.  With this feature present, the TRUNCATE or DROP runs much faster as a DDL without actually removing the target rows from the Global Indexes [but still requires the UPDATE INDEXES clause to be specified]

So, now in my 12.2 database I have these two Indexes on SALES_DATA :

SQL> select index_name, partitioned, status
  2  from user_indexes
  3  where table_name = 'SALES_DATA'
  4  order by 2,1
  5  /

INDEX_NAME                     PAR STATUS
------------------------------ --- --------
SALES_DATA_PK                  NO  VALID
SALES_DATA_LCL_NDX_1           YES N/A

SQL> 

I then TRUNCATE a non-empty Partition and check the Indexes

SQL> alter table sales_data truncate partition P_2015 update indexes;

Table truncated.

SQL>
SQL> select index_name, partitioned, status, orphaned_entries
  2  from user_indexes
  3  where table_name = 'SALES_DATA'
  4  order by 2,1
  5  /

INDEX_NAME                     PAR STATUS   ORP
------------------------------ --- -------- ---
SALES_DATA_PK                  NO  VALID    YES
SALES_DATA_LCL_NDX_1           YES N/A      NO

SQL> 

The ORPHANED_ENTRIES column indicates that SALES_DATA_PK is subject to Asynchronous Index Maintenance.

This is the job that will do the Index Maintenance at 2am  :

SQL> l
  1  select owner, job_name, last_start_date, next_run_Date
  2  from dba_scheduler_jobs
  3* where job_name = 'PMO_DEFERRED_GIDX_MAINT_JOB'
SQL> /

OWNER
---------------------------------------------------------------------------
JOB_NAME
---------------------------------------------------------------------------
LAST_START_DATE
---------------------------------------------------------------------------
NEXT_RUN_DATE
---------------------------------------------------------------------------
SYS
PMO_DEFERRED_GIDX_MAINT_JOB
20-MAR-19 10.18.51.215433 AM UTC
21-MAR-19 02.00.00.223589 AM UTC


SQL> !date
Wed Mar 20 20:05:24 SGT 2019

SQL> 

So, I could
(1) wait for the next run of the job OR
(2) manually trigger the job (which will scan the entire database for all indexes that require such maintenance) OR
(3) Execute  DBMS_PART.CLEANUP_GIDX  to initiate the maintenance for the specific index OR
(4) Execute an ALTER INDEX REBUILD to make the Index USABLE again.

SQL> execute dbms_part.cleanup_gidx('HEMANT','SALES_DATA');

PL/SQL procedure successfully completed.

SQL> select index_name, partitioned, status, orphaned_entries
  2  from user_indexes
  3  where table_name = 'SALES_DATA'
  4  order by 2,1
  5  /

INDEX_NAME                     PAR STATUS   ORP
------------------------------ --- -------- ---
SALES_DATA_PK                  NO  VALID    NO
SALES_DATA_LCL_NDX_1           YES N/A      NO

SQL> 

Note that the argument to CLEANUP_GIDX is the *Table Name*, not an Index Name.


Here I have demonstrated a TRUNCATE Partition, but the same method would be usable for a DROP Partition.