MacLochlainns Weblog

Michael McLaughlin's Technical Blog

Site Admin

Archive for the ‘Oracle’ Category

NDS parameters as IN OUT?

without comments

A question posed by a student: “Why are Oracle Native Dynamic SQL (NDS) USING clause parameters IN, IN OUT, or OUT when the RETURNING INTO clause manages output values?” It a great question, isn’t it? The followup question was also great, “How do you implement an example of NDS IN OUT parameters?”

The answer is two fold. First, you should use the USING clause for parameter list input values and the RETURNING INTO clause for return values whenever possible. Second, when it’s not possible you’re generally passing parameters into and out of an NDS PL/SQL anonymous block.

The basic prototype for passing and retrieving values from an NDS statement is:

EXECUTE IMMEDIATE sql_stmt
  USING { IN | IN OUT | OUT } local_variable [, ...]
  RETURNING INTO { IN OUT | OUT } local_variable [, ...];

A quick and hopefully fun example is this parody on Marvel’s The Avengers. The program creates an anonymous block with a super hero of Thor and super villain of Loki, then it uses a USING clause with IN OUT parameters to an anonymous block statement. That’s basically the trick to how you use IN OUT parameters in NDS statements.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
-- Enable SERVEROUTPUT.
SET SERVEROUTPUT ON SIZE UNLIMITED
 
-- Declare an anonymous testing block.
DECLARE
 
  -- Declare two local variables.
  lv_super_hero     VARCHAR2(20) := 'Thor';
  lv_super_villain  VARCHAR2(20) := 'Loki';
 
  -- Declare a null statement variable.
  lv_stmt  VARCHAR2(32767);
 
  -- Declare a local procedure to parse the NDS block.
  PROCEDURE print_code_block (pv_block VARCHAR2) IS
    -- Declare local parsing variables.
    lv_length   INTEGER := 1;
    lv_start    INTEGER := 1;
    lv_end      INTEGER := 1;
  BEGIN
    -- Read line by line on a line return character.
    WHILE NOT (lv_end = 0) LOOP
      -- Check for line returns.
      lv_end := INSTR(lv_stmt,CHR(10),lv_start);
      -- Check whether line return has been read.
      IF NOT lv_end = 0 THEN     
        -- Reset the ending substring value and print substring.
        lv_end := INSTR(lv_stmt,CHR(10),lv_start);
        dbms_output.put_line('| '||SUBSTR(lv_stmt,lv_start,lv_end - lv_start));
      ELSE
        -- Print the last substring with a semicolon and exit the loop.      
        dbms_output.put_line('| '||SUBSTR(lv_stmt,lv_start,LENGTH(lv_stmt) - lv_start)||';');
      END IF;
      -- Reset the beginning of the string.
      lv_start := lv_end + 1;      
    END LOOP;    
  END print_code_block;
 
BEGIN
 
  -- Demonstrate good triumps over evil.
  dbms_output.put_line('The good '||lv_super_hero||' beats up the bad '||lv_super_villain||'!');
 
  -- Assign the anonymous block to the local statement variable.
  lv_stmt := 'DECLARE'||CHR(10)
          || '  lv_super_hero     VARCHAR2(20);'||CHR(10)
          || '  lv_super_villain  VARCHAR2(20);'||CHR(10)
          || 'BEGIN'||CHR(10)
          || '  lv_super_hero '||CHR(58)||'= :pv_super_hero;'||CHR(10)
          || '  lv_super_villain '||CHR(58)||'= :pv_super_villain;'||CHR(10)
          || '  :pv_super_hero '||CHR(58)||'= lv_super_villain;'||CHR(10)
          || '  :pv_super_villain '||CHR(58)||'= lv_super_hero;'||CHR(10)
          || 'END;';
 
  -- Run the NDS program.
  EXECUTE IMMEDIATE lv_stmt USING IN OUT lv_super_hero
                                , IN OUT lv_super_villain;
 
  -- Print the diagnostic code block, that's why it used line returns afterall.  
  dbms_output.put_line('--------------------------------------------------');
  print_code_block(lv_stmt);
  dbms_output.put_line('--------------------------------------------------');
 
  -- Demonstrate the world is upside down without Johnny Depp playing Capt'n Jack.
  dbms_output.put_line('The good '||lv_super_hero||' beats up the bad '||lv_super_villain||'!');
 
END;
/

You’ll get the following printed output:

The good Thor beats up the bad Loki!
--------------------------------------------------
| DECLARE
|   lv_super_hero     VARCHAR2(20);
|   lv_super_villain  VARCHAR2(20);
| BEGIN
|   lv_super_hero := :pv_super_hero;
|   lv_super_villain := :pv_super_villain;
|   :pv_super_hero := lv_super_villain;
|   :pv_super_villain := lv_super_hero;
| END;
--------------------------------------------------
The good Loki beats up the bad Thor!

As always, I hope it helps you understand the concept of the USING clause with IN OUT parameters but I hope there’s always better way.

Written by maclochlainn

June 13th, 2012 at 11:52 pm

Result Cache Functions

without comments

I finally got around to cleaning up old contact me messages. One of the messages raises a question about RESULT_CACHE functions. The writer wanted an example implementing both a standalone schema and package RESULT_CACHE function.

The question references a note from the Oracle Database 11g PL/SQL Programming book (on page 322). More or less, that note points out that at the time of writing a RESULT_CACHE function worked as a standalone function but failed inside a package. When you tried it, you raised the following error message:

PLS-00999: Implementation Restriction (may be temporary)

It’s no longer true in Oracle 11gR2, but it was true in Oracle 11gR1. I actually mentioned in a blog entry 4 years ago.

You can implement a schema RESULT_CACHE function like this:

1
2
3
4
5
6
7
8
CREATE OR REPLACE FUNCTION full_name
( pv_first_name   VARCHAR2
, pv_last_name    VARCHAR2 )
RETURN VARCHAR2 RESULT_CACHE IS
BEGIN  
  RETURN pv_first_name || ' ' || pv_last_name;
END full_name;
/

You would call it like this from a query:

SELECT   full_name(c.first_name, c.last_name)
FROM     contact c;

You can declare a published package RESULT_CACHE function like this:

1
2
3
4
5
6
7
CREATE OR REPLACE PACKAGE cached_function IS
  FUNCTION full_name
  ( pv_first_name   VARCHAR2
  , pv_last_name    VARCHAR2 )
  RETURN VARCHAR2 RESULT_CACHE;
END cached_function;
/

You would implement the function in a package body like this:

1
2
3
4
5
6
7
8
9
10
CREATE OR REPLACE PACKAGE BODY cached_function IS
  FUNCTION full_name
  ( pv_first_name   VARCHAR2
  , pv_last_name    VARCHAR2 )
  RETURN VARCHAR2 RESULT_CACHE IS
  BEGIN  
    RETURN pv_first_name || ' ' || pv_last_name;
  END full_name; 
END cached_function;
/

You would call the package function like this from a query:

SELECT   cached_function.full_name(c.first_name, c.last_name)
FROM     contact c;

I hope this answers the question.

Written by maclochlainn

May 29th, 2012 at 12:31 am

MySQL Striped Views

with 7 comments

A question came up today about how to stripe a MySQL view, and this post shows you how. Along with the question, there was a complaint about why you can’t use session variables in a view definition. It’s important to note two things: there’s a workaround and there’s an outstanding request to add lift the feature limitation in Bug 18433.

A striped view lets authorized users see only part of a table, and is how Oracle Database 11g sets up Virtual Private Databases. Oracle provides both schema (or database) level access and fine-grained control access. Fine grained control involves setting a special session variable during a user’s login. This is typically done by checking the rights in an Access Control List (ACL) and using an Oracle built-in package.

You can do more or less the same thing in MySQL by using stored functions. One function would set the session variable and the other would fetch the value for comparison in a view.

Most developers who try this initially meet failure because they try to embed the session variable inside the view, like this trivial example with Hobbits (can’t resist the example with the first installment from Peter Jackson out later this year):

1
2
CREATE VIEW hobbit_v AS
SELECT * FROM hobbit WHERE hobbit_name = @sv_login_name;

The syntax is disallowed, as explained in the MySQL Reference 13.1.20 CREATE VIEW Syntax documentation. The attempt raises the following error message:

ERROR 1351 (HY000): VIEW's SELECT contains a variable or parameter

The fix is quite simple, you write a function that sets the ACL value for the session and another that queries the ACL session value. For the example, I’ve written the SET_LOGIN_NAME and a GET_LOGIN_NAME functions. (If you’re new to stored programs, you can find a 58 page chapter on writing them in my Oracle Database 11g & MySQL 5.6 Developer Handbook or you can use Guy Harrison’s MySQL Stored Procedure Programming.)

You would call the SET_LOGIN_NAME when you connect to the MySQL database as the first thing to implement this type of architecture. You would define the function like the following. (Please note that the example includes all setup statements from the command line and should enable you cutting and pasting it. ;-)):

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
-- Change the delimiter to something other than a semicolon.
DELIMITER $$
 
-- Conditionally drop the function.
DROP FUNCTION IF EXISTS set_login_name$$
 
-- Create the function.
CREATE FUNCTION set_login_name(pv_login_name VARCHAR(20)) RETURNS INT UNSIGNED
BEGIN
 
  /* Declare a local variable to verify completion of the task. */
  DECLARE  lv_success_flag  INT UNSIGNED  DEFAULT FALSE;
 
  /* Check whether the input value is something other than a null value. */
  IF pv_login_name IS NOT NULL THEN
 
    /* Set the session variable and enable the success flag. */
    SET @sv_login_name := pv_login_name;
    SET lv_success_flag := TRUE;
 
  END IF;
 
  /* Return the success flag. */
  RETURN lv_success_flag;
END;
$$
 
-- Change the delimiter back to a semicolon.
DELIMITER ;

You can use a query to set and confirm action like this:

SELECT IF(set_login_name('Frodo')=TRUE,'Login Name Set','Login Name Not Set') AS "Login Name Status";

Or, you can use the actual number 1 in lieu of the TRUE, like this:

SELECT IF(set_login_name('Frodo')=1,'Login Name Set','Login Name Not Set') AS "Login Name Status";

Please check this older post on how MySQL manages logical constants and the realities of TRUE and FALSE constants. A more practical example in an API would be this, which returns zero when unset and one when set:

SELECT set_login_name('Frodo') AS "Login Name Status";

The getter function for this example, simply reads the current value of the MySQL session variable. Like the prior example, it’s ready to run too.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
-- Change the delimiter to something other than a semicolon.
DELIMITER $$
 
-- Conditionally drop the function.
DROP FUNCTION IF EXISTS get_login_name$$
 
-- Create the function.
CREATE FUNCTION get_login_name() RETURNS VARCHAR(20)
BEGIN
  /* Return the success flag. */
  RETURN @sv_login_name;
END;
$$
 
-- Change the delimiter back to a semicolon.
DELIMITER ;

Before you test it, lets create a HOBBIT table, seed it with data, and create a HOBBIT_V view. They’re bundled together in the following microscript:

-- Conditionally drop the table.
DROP TABLE IF EXISTS hobbit;
 
-- Create the table.
CREATE TABLE hobbit
( hobbit_id    INT UNSIGNED
, hobbit_name  VARCHAR(20));
 
-- Seed two rows.
INSERT INTO hobbit VALUES ( 1,'Bilbo'),( 1,'Frodo');
 
-- Conditionally drop the view.
DROP VIEW IF EXISTS hobbit_v;
 
-- Create the function-enabled view.
CREATE VIEW hobbit_v AS
SELECT * FROM hobbit WHERE hobbit_name = get_login_name();

A query to the table after setting the session variable will only return one row, the row with Frodo in the HOBBIT_NAME column. It also guarantees an unfiltered UPDATE statement against the view only updates the single row returned, like this:

UPDATE hobbit_v SET hobbit_id = 2;

In a real solution, there are more steps. For example, you’d want your tables in one database, views in another, and functions and procedures in a library database. However, I hope this helps seed some ideas for those interested in creating fine-grained virtual private databases in MySQL with user-authenticated application controls.

Written by maclochlainn

May 23rd, 2012 at 11:41 pm

Collaborate 2012 – Day 4

with one comment

Last day of Collaborate 2012 and Scott Spendolini, Sumneva, gave a great presentation on APEX. Only caught the beginning Jan Visser’s Perl presentation because of the distance to the Luxor from the Mandalay South Conference Center and anticipated queuing time for checkout.

We can now look forward to Collaborate 2013 in Denver, Colorado.

Back to observing and working with code, here’s a nice article from MacWorld on how you set up a WebDAV on the Mac. While I’m mentioning Mac OS X and development, there’s still no firm upgrade window for the missing text editing tool – TextMate, and WWDC 2012 tickets sold out in two hours.

Written by maclochlainn

April 26th, 2012 at 1:19 pm

Collaborate 2012 – Day 3

without comments

Virtualization is important and Dave Welch from the House of Brick gave a great presentation of experiences with VMWare and Tier 1 databases. It was a comprehensive presentation, but the white paper was easier to follow. The slides were complete but the volume of information was a lot for an hour presentation. Well worth the time though.

Utah Oracle User Group (UTOUG) announced a call for Fall Symposium papers today. The Fall Symposium will be in Salt Lake City on 9/6/2012. If you’re interested in presenting on Oracle or MySQL, the call for presentations will be open until 6/15/2012.

The conference party was tonight, and it provided some nice orderves and pizza. The theme was a return to 1980s music, and some folks really dressed their parts. You can listen to a short snapshot of the band by clicking the image to launch a small video segment.

I’m looking forward to the APEX Behind the Scenes presentation at 8:30 a.m. tomorrow. When the conference is over, I won’t miss the smoke filled air that we walk through from the Luxor to the Mandalay. It’s really amazing that the complex is more than a mile in length. It runs from the Luxor to the Mandalay South Conference Center.

Written by maclochlainn

April 26th, 2012 at 12:31 am

Collaborate 2012 – Day 2

without comments

It seems the Titanic is everywhere, even inside the pyramid of the Luxor hotel. While the Luxor is within the Mandalay Bay complex, it’s about a half mile walk to the conference and a half mile back. We go by the Mandalay Conference Center’s aquarium. We thought it might be interesting but at $18 an admission, we opted to pass on it. It’s amazing to have an aquarium in the desert, but it’s probably not as nice as the Monterey Bay aquarium.

It was interesting to start the day listening to Rich Niemiec on partitioning tables and using Exadata in Oracle. The NoSQL (Not Only SQL) presentations were interesting, as was the upgrading of Oracle 11gR2 in an E-Business Suite environment presentation. Then, I finished the day with what’s new with the Oracle VM Server.

Checking out the exhibit hall I managed to get a signed copy of Rich Niemiec’s Oracle Database 11g Release 2 Performance Tuning Tips & TechniquesOracle Database 11g Release 2 Performance Tuning Tips & Techniques and a copy of MongoDB: The Definitive GuideMongoDB The Definitive Guide.

Written by maclochlainn

April 25th, 2012 at 1:52 am

Collaborate 2012 – Day 1

without comments

Collaborate 2012 started on Sunday but for me I began on Monday. I enjoyed Bob Burgess, SalesForce, presentation on shell scripting for MySQL Administration today. It preceded my presentation in the same room, which I thought was an interesting coincidence since we got our conference credentials together.

I presented on portable SQL between Oracle and MySQL. The presentation went well. Before I took questions, I got to ask them because I had three copies of my new Oracle Press book to give away: Oracle Database 11g and MySQL 5.6 Developer Handbook. Handing out the books served as a nice ice breaker for the audience to ask questions about the presentation.

My favorite question was, “Will Oracle continue to improve MySQL?” My answer to that is always simple because Oracle’s support for MySQL has and continues to be great, “Oracle only spends money on winners and that means MySQL wins.” Oracle product management was in attendance and they re-enforced Oracle’s commitment to MySQL.

At 6 p.m., the Exhibit Hall opened and I checked it out. Cisco hired Kathy Bailey to draw caricatures, and she drew mine as you can see at the left. I’m looking forward to more presentations tomorrow.

Written by maclochlainn

April 24th, 2012 at 1:47 am

Oracle CSV Imports

with one comment

The first step in creating an effective import plan for comma-separated value (CSV) files is recognizing your options in a database. There are several options in an Oracle database. You can read the file with Java, C/C++, C#, PL/SQL (through the UTL_FILE package), PHP, Perl, or any other C-callable programming language; or you can use SQL*Loader as a standalone utility or through externally managed tables (known as external tables). The most convenient and non-programming solution is using external tables.

Adopting external tables as your import solution should drive you to consider how to manage the security surrounding this type of methodology. Host hardening is a critical security step because it shuts down most, hopefully all, unauthorized use of the operating system where the database and external files reside. Next, you need to manage the access to the external tables and ensure that exposure of business sensitive information in CSV files is minimized.

This post explains how to manage access and police (cleanup external files) once they’re read into the database. It assumes you have root-level permissions to the operating system and database. The SYS and SYSTEM accounts have the equivalent of root permissions for database configuration. The rule of thumb with these accounts is simple, manage as much as possible with the SYSTEM account before you use the SYS account.

Setting up the Import File System

While you can do all the setup of virtual directories in Oracle regardless of whether you’ve set them up in the operating system, it’s a good idea to set them up in the OS first. The example is using a Windows 7 OS, so you’ll need to change the directories when working in Linux or Unix. Here are the directories:

C:\Imports\ImportFiles
C:\Imports\ImportLogs

You may take note that there are two directories. That’s because you don’t want to grant write privileges to the Oracle virtual directory where you put the files. You can grant read-only privileges to the virtual directory and read-write privileges to the log directory.

Setting up the Import User/Schema

This step lets you create an IMPORT user/schema in the Oracle database. You need to connect as the SYSTEM user to perform these steps (or another authorized DBA account with adequate privileges):

CREATE USER import IDENTIFIED BY import
DEFAULT TABLESPACE users QUOTA 1000M ON users
TEMPORARY TABLESPACE temp;

After creating the user, grant the privileges like this as the SYSTEM user:

GRANT CREATE CLUSTER, CREATE INDEXTYPE, CREATE OPERATOR
,     CREATE PROCEDURE, CREATE SEQUENCE, CREATE SESSION
,     CREATE SYNONYM, CREATE TABLE, CREATE TRIGGER
,     CREATE TYPE,  CREATE VIEW TO import;

Setting up Virtual Directories

A virtual directory in Oracle acts maps an internal database directory name (known as a virtual directory) to a physical directory of the operating system. You create two virtual directories in this example, one holds read-only permissions to the directory where you’re putting the data file, and the other holds read-write permissions to the directory where you’re writing any log files from the external file process.

Log files are generated from this process when you query the data from the external file. Any error in the files conformity is written to a log file.

CREATE DIRECTORY upload_files AS 'C:\Imports\ImportFiles';
CREATE DIRECTORY upload_logs AS 'C:\Imports\ImportLogs';

After creating the virtual directories in the database, you must grant appropriate access to the user account that will access the data. This grants those permissions to the IMPORT user:

GRANT READ ON DIRECTORY upload_files TO import;
GRANT READ, WRITE ON DIRECTORY upload_logs TO import;

Setting up an External Table

An external table references both the UPLOAD_FILES and UPLOAD_LOGS virtual directories, and the virtual directories must map to physical directories that allow read and write privileges to the Oracle user. Here’s the external table for this example:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
CREATE TABLE item_import_ext_table
( asin_number           VARCHAR2(10)
, item_type             VARCHAR2(15)
, item_title            VARCHAR2(60)
, item_subtitle          VARCHAR2(60)
, item_rating            VARCHAR2(8)
, item_rating_agency     VARCHAR2(4)
, item_release_date      DATE)
  ORGANIZATION EXTERNAL
  ( TYPE oracle_loader
    DEFAULT DIRECTORY upload_files
    ACCESS PARAMETERS
    ( RECORDS DELIMITED BY NEWLINE CHARACTERSET US7ASCII
      BADFILE     'UPLOAD_LOGS':'item_import_ext_table.bad'
      DISCARDFILE 'UPLOAD_LOGS':'item_import_ext_table.dis'
      LOGFILE     'UPLOAD_LOGS':'item_import_ext_table.log'
      FIELDS TERMINATED BY ','
      OPTIONALLY ENCLOSED BY "'"
      MISSING FIELD VALUES ARE NULL )
    LOCATION ('item_import.csv'))
REJECT LIMIT UNLIMITED;

Setting up a Physical File

You should put the following in a item_import.csv physical file (case sensitivity won’t matter on the Windows 7 platform but will matter on the Linux or Unix platforms):

'B000W74EQC','DVD_WIDE_SCREEN','Harry Potter and the Sorcerer''s Stone',,'PG','MPAA','11-DEC-2007'
'B000W746GK','DVD_WIDE_SCREEN','Harry Potter and the Chamber of Secrets',,'PG','MPAA','11-DEC-2007'
'B000W796OM','DVD_WIDE_SCREEN','Harry Potter and the Prisoner of Azkaban',,'PG','MPAA','11-DEC-2007'
'B000E6EK2Y','DVD_WIDE_SCREEN','Harry Potter and the Goblet of Fire',,'PG-13','MPAA','07-MAR-2006'
'B000W7F5SS','DVD_WIDE_SCREEN','Harry Potter and the Order of the Phoenix',,'PG-13','MPAA','11-DEC-2007'
'B002PMV9FG','DVD_WIDE_SCREEN','Harry Potter and the Half-Blood Prince',,'PG','MPAA','08-DEC-2009'
'B001UV4XHY','DVD_WIDE_SCREEN','Harry Potter and the Deathly Hallows, Part 1',,'PG-13','MPAA','15-APR-2011'
'B001UV4XIS','DVD_WIDE_SCREEN','Harry Potter and the Deathly Hallows, Part 2',,'PG-13','MPAA','11-NOV-2011'

Testing the External Table

After putting the item_import.csv file in the C:\Imports\ImportFiles directory, you can test the process at this point by running the following query:

SET PAGESIZE 99
 
COLUMN asin_number        FORMAT A11 HEADING "ASIN #"
COLUMN item_title         FORMAT A46 HEADING "ITEM TITLE"
COLUMN item_rating        FORMAT A6  HEADING "RATING"
COLUMN item_release_date  FORMAT A11 HEADING "RELEASE|DATE"
 
SELECT   asin_number
,        item_title
,        item_rating
,        TO_CHAR(item_release_date,'DD-MON-YYYY') AS item_release_date
FROM     item_import_ext_table;

It should return eight rows.

Extending Access to the Data Dictionary

The physical directory names of virtual directories are hidden from generic users. They’re available in the ALL_DIRECTORIES and DBA_DIRECTORIES administrative view for queries by the SYS, SYSTEM, and any DBA role privileged users.

While a privileged user can query the view, placing the view inside a function or procedure deployed in the privileged user’s schema would raise an ORA-00942 error. That error signals that the table or view does not exist.

This example deploys the view in the SYSTEM schema. That means it requires you make the following grant as the SYS user:

GRANT SELECT ON sys.dba_directories TO system;

After making the grant from the SYS schema to the SYSTEM schema, connect to the SYSTEM schema. Then, create the following GET_DIRECTORY_PATH function:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
CREATE OR REPLACE FUNCTION get_directory_path
( virtual_directory IN VARCHAR2 )
RETURN VARCHAR2 IS
  -- Define RETURN variable.
  directory_path VARCHAR2(256) := '';
  --Define dynamic cursor.
  CURSOR get_directory (virtual_directory VARCHAR2) IS
    SELECT   directory_path
    FROM     sys.dba_directories
    WHERE    directory_name = virtual_directory;
  -- Define a LOCAL exception FOR name violation.
  directory_name EXCEPTION;
  PRAGMA EXCEPTION_INIT(directory_name,-22284);
BEGIN
  OPEN  get_directory (virtual_directory);
  FETCH get_directory
  INTO  directory_path;
  CLOSE get_directory;
  -- RETURN file name.
  RETURN directory_path;
EXCEPTION
  WHEN directory_name THEN
    RETURN NULL;
END get_directory_path;
/

It’s tempting to make the grant on this function to PUBLIC user but that would expose information that any DBA should try and limit. That means you grant EXECUTE privilege only to the IMPORT schema.

This grant should be made as the SYSTEM user:

GRANT EXECUTE ON get_directory_path TO import;

After granting the EXECUTE privilege to the IMPORT user, connect to the IMPORT schema and create a synonym to the GET_DIRECTORY_PATH function. The syntax for that command is:

CREATE SYNONYM get_directory_path FOR system.get_directory_path;

You can now test your access to the function with the following query from the IMPORT schema:

SELECT get_directory_path('UPLOAD_FILES') FROM dual;

You should return the following if you’ve got everything working at this point:

GET_DIRECTORY_PATH('UPLOAD_FILES')
------------------------------------
C:\Imports\ImportFiles

At this point, you’ve completed the second major configuration component. You now need the ability to read files outside the database, which can be done with Java in Oracle 10g or Oracle 11g (that’s not possible in Oracle 10g XE or Oracle 11g XE because they don’t support an internal JVM). The

Reading Virtual Directory Files

The GET_DIRECTORY_PATH function provides you with the ability to read the Oracle data catalog and find the absolute directory path of a virtual directory. In this framework, you need this value to find whether the item_import.csv physical file is present in the file system before you read the file.

There doesn’t appear to be a neat little function to read an external directory. At least, there’s not one in the UTL_FILE or DBMS_LOB packages where you’d think it should be found. Unfortunately, that leaves us with two alternatives. One is to write an external library in C, C++, or C#. Another is to write an internal Java library that reads the file system. You accomplish this by granting permissions to a target directory or directories.

The first step is to create a scalar array of VARCHAR2 variables, like

CREATE OR REPLACE TYPE file_list AS TABLE OF VARCHAR2(255);
/

The second step is to write the Java library file. You can write it three ways. One accepts default error handling and the others override the default exception handling. If you’re new to Java, you should take the basic library with default handling. If you’ve more experience, you may want to override the helpful message with something that causes the developer to check with the DBA or simply suppress the message to enhance security.

You should note that the database connection is an Oracle Database 11g internal database connection. The connection only does one thing. It allows you to map the ArrayDescriptor to a schema-level SQL collection type. The element types of these collections should be scalar variables, like DATE, NUMBER, or VARCHAR2 data types.

The more advanced method overrides exception handling by suppressing information about the java.properties settings. You can do it by catching the natively thrown exception and re-throw it or ignore it. The example ignores it because handling it in Java reports an unhandled exception at the PL/SQL or SQL layer, which leads end users to think you have a major design problem.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
CREATE OR REPLACE AND COMPILE JAVA SOURCE NAMED "ListVirtualDirectory" AS
 
  // Import required classes.
  import java.io.*;
  import java.security.AccessControlException;
  import java.sql.*;
  import java.util.Arrays;
  import oracle.sql.driver.*;
  import oracle.sql.ArrayDescriptor;
  import oracle.sql.ARRAY;
 
  // Define the class.
  public class ListVirtualDirectory {
 
    // Define the method.
    public static ARRAY getList(String path) throws SQLException {
 
    // DECLARE variable AS a NULL, required because OF try-catch block.
    ARRAY listed = NULL;
 
    // Define a connection (this IS FOR Oracle 11g).
    Connection conn = DriverManager.getConnection("jdbc:default:connection:");
 
    // USE a try-catch block TO trap a Java permission error ON the directory.
    try {
      // DECLARE a class WITH the file list.
      File directory = NEW File(path);
 
      // DECLARE a mapping TO the schema-level SQL collection TYPE.
      ArrayDescriptor arrayDescriptor = NEW ArrayDescriptor("FILE_LIST",conn);
 
      // Translate the Java String[] TO the Oracle SQL collection TYPE.
      listed = NEW ARRAY(arrayDescriptor,conn,((Object[]) directory.list())); }
    catch (AccessControlException e) {}
  RETURN listed; }}
/

You can’t call an internal Java library without a PL/SQL wrapper function. Here’s the wrapper function for this Java library:

CREATE OR REPLACE FUNCTION list_files(path VARCHAR2) RETURN FILE_LIST IS
LANGUAGE JAVA
NAME 'ListVirtualDirectory.getList(java.lang.String) return oracle.sql.ARRAY';
/

You MUST grant the Oracle Database’s internal JVM authority to read the external directory before you can return the directory contents. Any attempt to read a directory without the proper permissions raises an ORA-29532 exception.

The following is an anonymous block to grant permissions to a directory. You must grant a minimum of read permissions but since you’ll also delete this file later in the post you should grant read, write, and delete. You must run it from the SYSDBA role as the SYS user.

1
2
3
4
5
6
7
BEGIN
  DBMS_JAVA.GRANT_PERMISSION('IMPORT'
                             ,'SYS:java.io.FilePermission'
                             ,'C:\Imports\ImportFiles'
                             ,'read,write,delete');
END;
/

While you’re connected, it’s a good idea to grant the same privileges to your log directory:

1
2
3
4
5
6
7
BEGIN
  DBMS_JAVA.GRANT_PERMISSION('IMPORT'
                            ,'SYS:java.io.FilePermission'
                            ,'C:\Imports\ImportLogs'
                            ,'read,write,delete');
END;
/

You should now be able to read the contents of an external file from another PL/SQL block or from a SQL statement. Here’s an example of the SQL statement call that uses everything developed to this point:

SELECT   column_value AS "File Names"
FROM     TABLE(list_files(get_directory_path('UPLOAD_FILES')));

It should return the item_import.csv physical file as the only file in the physical directory, like:

File Names
-----------------
item_import.csv

Mapping an External Table to a source File

The next step leverages the user segment of the Oracle Database’s data catalog and all the components developed above to find and display the external table and external file. This query returns the results:

COLUMN TABLE_NAME FORMAT A30
COLUMN file_name  FORMAT A30
 
SELECT   xt.table_name
,        xt.file_name
FROM    (SELECT   uxt.TABLE_NAME
         ,        ixt.column_value AS file_name
         FROM     user_external_tables uxt CROSS JOIN
         TABLE(list_files(get_directory_path(uxt.default_directory_name))) ixt) xt
JOIN     user_external_locations xl
ON       xt.table_name = xl.table_name AND xt.file_name = xl.location;

It should return the following:

TABLE_NAME                     FILE_NAME
------------------------------ ------------------------------
ITEM_IMPORT_EXT_TABLE          item_import.csv

You can migrate the query into the following function. It returns a zero when the file isn’t found and a one when it is found.

CREATE OR REPLACE FUNCTION external_file_found
( table_in VARCHAR2 ) RETURN NUMBER IS
  -- Define a default return value.
  retval NUMBER := 0;
 
  -- Decalre a cursor to find external tables.
  CURSOR c (cv_table VARCHAR2) IS
    SELECT   xt.table_name
    ,        xt.file_name
    FROM    (SELECT   uxt.TABLE_NAME
             ,        ixt.column_value AS file_name
             FROM     user_external_tables uxt CROSS JOIN
             TABLE(list_files(get_directory_path(uxt.default_directory_name))) ixt) xt
    JOIN     user_external_locations xl ON xt.table_name = xl.table_name
    AND      xt.file_name = xl.location AND xt.table_name = UPPER(cv_table);
BEGIN
  FOR i IN c(table_in) LOOP
    retval := 1;
  END LOOP;
  RETURN retval;
END;
/

With the EXTERNAL_FILE_FOUND function, you can create a function that returns rows when the external file is found and no rows when the external file isn’t found. The following view hides the logic required to make that work:

CREATE OR REPLACE VIEW item_import AS
SELECT   *
FROM     item_import_ext_table
WHERE    external_file_found('ITEM_IMPORT_EXT_TABLE') = 1;

Conveniently, you can now query the ITEM_IMPORT view without the risk of raising the following error when the file is missing:

SELECT * FROM item_import_ext_table
*
ERROR at line 1:
ORA-29913: error IN executing ODCIEXTTABLEOPEN callout
ORA-29400: DATA cartridge error
KUP-04040: file item_import.csv IN UPLOAD_FILES NOT found

You can now grant the SELECT privilege on the ITEM_IMPORT view to your application schema, like:

GRANT SELECT ON item_import TO application;

After granting the SELECT privilege on the ITEM_IMPORT view to the APPLICATION schema, you can create a synonym to hide the IMPORT schema.

CREATE SYNONYM item_import FOR item_import;

At this point, many developers feel they’re done. Enclosing the results in a schema-level function provides more utility than a view. The next section shows you how to replace the view with a schema-level function.

Replacing the View with an Object Table Function

Inside a schema-level function, you can assign the results from the query to a SQL collection of an object type. The object type should mirror the structure of the table, like the following:

1
2
3
4
5
6
7
8
9
CREATE OR REPLACE TYPE item_import_object IS OBJECT
( asin_number         VARCHAR2(10)
, item_type           VARCHAR2(15)
, item_title          VARCHAR2(60)
, item_subtitle       VARCHAR2(60)
, item_rating         VARCHAR2(8)
, item_rating_agency  VARCHAR2(4)
, item_release_date   DATE);
/

After creating the object type that mirrors the structure of the ITEM_IMPORT_EXT_TABLE table, you need to create a list like collection of the object type. The nested table collection type acts like a list in Oracle:

1
2
3
CREATE OR REPLACE TYPE item_import_object_table IS
  TABLE OF item_import_object;
/

After defining the object type and collection, you can access them in the following type of function:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
CREATE OR REPLACE FUNCTION external_file_contents
( table_in VARCHAR2 ) RETURN item_import_object_table IS
 
  -- Define a local counter.
  lv_counter NUMBER := 1;
 
  -- Construct an empty collection of ITEM_IMPORT_OBJECT data types.
  lv_item_import_table ITEM_IMPORT_OBJECT_TABLE := item_import_object_table();
 
  -- Decalre a cursor to find external tables.
  CURSOR c (cv_table VARCHAR2) IS
    SELECT   *
    FROM     item_import_ext_table
    WHERE    external_file_found(cv_table) = 1;
 
BEGIN
  FOR i IN c(table_in) LOOP
    lv_item_import_table.EXTEND;
    lv_item_import_table(lv_counter) := item_import_object(i.asin_number
                                                          ,i.item_type
                                                          ,i.item_title
                                                          ,i.item_subtitle
                                                          ,i.item_rating
                                                          ,i.item_rating_agency
                                                          ,i.item_release_date);
    lv_counter := lv_counter + 1;
  END LOOP;
 
  /*
   *  This is where you can place autonomous function calls:
   *  ======================================================
   *   - These can read source and log files, and write them
   *     to CLOB attributes for later inspection or review.
   *   - These can call Java libraries to delete files, but
   *     you should note that Java deletes any file rather
   *     than moving it to the trash bin (where you might
   *     recover it.
   */
 
  RETURN lv_item_import_table;
END;
/

Between the assignment to the collection and the return statement of the function, you have the ability of calling any number of autonomous functions. Any schema-level function can call autonomous functions that read and write tables with DML statements, like the INSERT, UPDATE, and DELETE statements. You can also call schema-functions that wrap Java libraries that delete external files.

You can confirm that the steps work by running the following query with or without the SQL*Plus formatting:

/*
 *  SQL*Plus formatting.
 */
SET PAGESIZE 99
 
COLUMN asin_number        FORMAT A11 HEADING "ASIN #"
COLUMN item_title         FORMAT A46 HEADING "ITEM TITLE"
COLUMN item_rating        FORMAT A6  HEADING "RATING"
COLUMN item_release_date  FORMAT A11 HEADING "RELEASE|DATE"
 
/*
 *  Query works only when item_import.csv file is present.
 */
SELECT   asin_number
,        item_title
,        item_rating
,        TO_CHAR(item_release_date,'DD-MON-YYYY') AS item_release_date
FROM     TABLE(external_file_contents('ITEM_IMPORT_EXT_TABLE'));

It should return the following from SQL*Plus:

                                                                  RELEASE
ASIN #      ITEM TITLE                                     RATING DATE
----------- ---------------------------------------------- ------ -----------
B000W74EQC  Harry Potter and the Sorcerer's Stone          PG     11-DEC-2007
B000W746GK  Harry Potter and the Chamber of Secrets        PG     11-DEC-2007
B000W796OM  Harry Potter and the Prisoner of Azkaban       PG     11-DEC-2007
B000E6EK2Y  Harry Potter and the Goblet of Fire            PG-13  07-MAR-2006
B000W7F5SS  Harry Potter and the Order of the Phoenix      PG-13  11-DEC-2007
B002PMV9FG  Harry Potter and the Half-Blood Prince         PG     08-DEC-2009
B001UV4XHY  Harry Potter and the Deathly Hallows, Part 1   PG-13  15-APR-2011
B001UV4XIS  Harry Potter and the Deathly Hallows, Part 2   PG-13  11-NOV-2011

The creation of the schema-level function lets you recreate the ITEM_IMPORT view. The following view would encapsulate (or hide) the presence of the function, which hides all the infrastructure components developed before this section (see line 14 in the function):

1
2
3
CREATE OR REPLACE VIEW item_import AS
SELECT   *
FROM     TABLE(external_file_contents('ITEM_IMPORT_EXT_TABLE'));

Implementing a Managed Import Process

During any import the information from the import process is exposed and one or more items may fail during the import process. That means the source file and loading log files must be preserved immediately after reading the data successfully. This is done by loading the data source file and log, discard, and bad import files into database tables. Only the source and log files exist when all rows are well formed, but the log files are reused for any subsequent load and require human inspection to isolate a specific upload.

The best way to implement this requires creating individual tables to hold each of the four potential large objects. The ITEM_MASTER table holds a transactional primary key and a table name for the import table. The primary key of the ITEM_MASTER table is the base key for imports and the ITEM_DATA, ITEM_LOG, ITEM_DISCARD, and ITEM_BAD tables hold foreign keys that point back to the ITEM_MASTER table’s primary key. These tables also hold a character large object column (CLOB), which will hold the respective source data file or log, discard, or bad files.

The following create the tables for the logging framework:

CREATE TABLE import_master
( import_master_id  NUMBER CONSTRAINT pk_import_master PRIMARY KEY
, import_table      VARCHAR2(30));
 
-- Create sequence for import master.
CREATE SEQUENCE import_master_s;
 
-- Create import table.
CREATE TABLE import_data
( import_data_id    NUMBER CONSTRAINT pk_import_data PRIMARY KEY
, import_master_id  NUMBER
, import_data       CLOB
, CONSTRAINT fk_import_data FOREIGN KEY (import_data_id)
  REFERENCES import_master (import_master_id))
LOB (import_data) STORE AS BASICFILE item_import_clob
(TABLESPACE users ENABLE STORAGE IN ROW CHUNK 32768
 PCTVERSION 10 NOCACHE LOGGING
 STORAGE (INITIAL 1048576
          NEXT    1048576
          MINEXTENTS 1
          MAXEXTENTS 2147483645));
 
-- Create sequence for import master.
CREATE SEQUENCE import_data_s;
 
-- Create import table.
CREATE TABLE import_log
( import_log_id     NUMBER CONSTRAINT pk_import_log PRIMARY KEY
, import_master_id  NUMBER
, import_log        CLOB
, CONSTRAINT fk_import_log FOREIGN KEY (import_log_id)
  REFERENCES import_master (import_master_id))
LOB (import_log) STORE AS BASICFILE item_import_log_clob
(TABLESPACE users ENABLE STORAGE IN ROW CHUNK 32768
 PCTVERSION 10 NOCACHE LOGGING
 STORAGE (INITIAL 1048576
          NEXT    1048576
          MINEXTENTS 1
          MAXEXTENTS 2147483645));
 
-- Create sequence for import master.
CREATE SEQUENCE import_log_s;
 
-- Create import table.
CREATE TABLE import_discard
( import_discard_id  NUMBER CONSTRAINT pk_import_discard PRIMARY KEY
, import_master_id   NUMBER
, import_discard     CLOB
, CONSTRAINT fk_import_discard FOREIGN KEY (import_discard_id)
  REFERENCES import_master (import_master_id))
LOB (import_discard) STORE AS BASICFILE item_import_discard_clob
(TABLESPACE users ENABLE STORAGE IN ROW CHUNK 32768
 PCTVERSION 10 NOCACHE LOGGING
 STORAGE (INITIAL 1048576
          NEXT    1048576
          MINEXTENTS 1
          MAXEXTENTS 2147483645));
 
-- Create sequence for import master.
CREATE SEQUENCE import_discard_s;
 
-- Create import table.
CREATE TABLE import_bad
( import_bad_id     NUMBER CONSTRAINT pk_import_bad PRIMARY KEY
, import_master_id  NUMBER
, import_bad        CLOB
, CONSTRAINT fk_import_bad FOREIGN KEY (import_bad_id)
  REFERENCES import_master (import_master_id))
LOB (import_bad) STORE AS BASICFILE item_import_bad_clob
(TABLESPACE users ENABLE STORAGE IN ROW CHUNK 32768
 PCTVERSION 10 NOCACHE LOGGING
 STORAGE (INITIAL 1048576
          NEXT    1048576
          MINEXTENTS 1
          MAXEXTENTS 2147483645));
 
-- Create sequence for import master.
CREATE SEQUENCE import_bad_s;

The tables set the targets for uploading the source and log files. You should note that the table name is also the column name for the CLOB column, this becomes convenient when supporting a Native Dynamic SQL (NDS) statement in a single autonomous function. The LOAD_CLOB_FROM_FILE function supports reading the external source and log files and writing them their respective tables.

There is a DEADLOCK possibility with this type of architecture. It requires that the base row in the IMPORT_MASTER table is committed before attempting inserts into one of the dependent tables. A call to the function raises an error when the primary key column hasn’t been committed before hand.

You already set the access privileges for the DBMS_LOB package when you granted them to the UPLOAD_FILES and UPLOAD_LOGS virtual directories. This function only requires read permissions, which were granted to both virtual directories.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
CREATE OR REPLACE FUNCTION load_clob_from_file
( pv_src_file_name  IN VARCHAR2
, pv_virtual_dir    IN VARCHAR2
, pv_table_name     IN VARCHAR2
, pv_column_name    IN VARCHAR2
, pv_foreign_key    IN NUMBER ) RETURN NUMBER IS
 
  -- Declare placeholder for sequence generated primary key.
  lv_primary_key  NUMBER;
 
  -- Declare default return value.
  lv_retval  NUMBER := 0;
 
  -- Declare local variables for DBMS_LOB.LOADCLOBFROMFILE procedure.
  des_clob    CLOB;
  src_clob    BFILE := BFILENAME(pv_virtual_dir,pv_src_file_name);
  des_offset  NUMBER := 1;
  src_offset  NUMBER := 1;
  ctx_lang    NUMBER := dbms_lob.default_lang_ctx;
  warning     NUMBER;
 
  -- Declare pre-reading size.
  src_clob_size  NUMBER;
 
  -- Declare variables for handling NDS sequence value.
  lv_sequence          VARCHAR2(30);
  lv_sequence_output   NUMBER;
  lv_sequence_tagline  VARCHAR2(10) := '_s.nextval';
 
  -- Define local variable for Native Dynamic SQL (NDS) Statement.
  stmt  VARCHAR2(2000);
 
  -- Declare the function as an autonomous transaction.
  PRAGMA AUTONOMOUS_TRANSACTION;
 
BEGIN
 
  -- Open file only when found.
  IF dbms_lob.fileexists(src_clob) = 1  AND NOT dbms_lob.isopen(src_clob) = 1 THEN
    src_clob_size := dbms_lob.getlength(src_clob);
    dbms_lob.open(src_clob,dbms_lob.lob_readonly);
  END IF;
 
  -- Concatenate the sequence name with the tagline.
  lv_sequence := pv_table_name || lv_sequence_tagline;
 
  -- Assign the sequence through an anonymous block.
  stmt := 'BEGIN '
       || '  :output := '||lv_sequence||';'
       || 'END;';
 
  -- Run the statement to extract a sequence value through NDS.
  EXECUTE IMMEDIATE stmt USING IN OUT lv_sequence_output;
 
  --  Create a dynamic statement that works for all source and log files.
  -- ----------------------------------------------------------------------
  --  NOTE: This statement requires that the row holding the primary key
  --        has been committed because otherwise it raises the following
  --        error because it can't verify the integrity of the foreign
  --        key constraint.
  -- ----------------------------------------------------------------------
  --        DECLARE
  --        *
  --        ERROR at line 1:
  --        ORA-00060: deadlock detected while waiting for resource
  --        ORA-06512: at "IMPORT.LOAD_CLOB_FROM_FILE", line 50
  --        ORA-06512: at line 20
  -- ----------------------------------------------------------------------  
  stmt := 'INSERT INTO '||pv_table_name||' '||CHR(10)||
          'VALUES '||CHR(10)||
          '('||lv_sequence_output||CHR(10)||
          ','||pv_foreign_key||CHR(10)||
          ', empty_clob())'||CHR(10)||
          'RETURNING '||pv_column_name||' INTO :locator';
 
  -- Run dynamic statement.
  EXECUTE IMMEDIATE stmt USING OUT des_clob;
 
  -- Read and write file to CLOB, close source file and commit.
  dbms_lob.loadclobfromfile( dest_lob     => des_clob
                           , src_bfile    => src_clob
                           , amount       => dbms_lob.getlength(src_clob)
                           , dest_offset  => des_offset
                           , src_offset   => src_offset
                           , bfile_csid   => dbms_lob.default_csid
                           , lang_context => ctx_lang
                           , warning      => warning );
 
  -- Close open source file.
  dbms_lob.close(src_clob);
 
  -- Commit write and conditionally acknowledge it.
  IF src_clob_size = dbms_lob.getlength(des_clob) THEN
    COMMIT;
    lv_retval := 1;
  ELSE
    RAISE dbms_lob.operation_failed;
  END IF;
 
  RETURN lv_retval;  
END load_clob_from_file;
/

You can test this procedure against the data source file with the following script file:

-- Insert a sample row in the master table.
INSERT INTO import_master
VALUES (import_master_s.nextval,'ITEM_IMPORT_EXT_TABLE');
 
-- Record the row value to avoid deadlock on uncommitted master record.
COMMIT;
 
-- Test program for loading CLOB files.
DECLARE
 
  -- Declare testing variables.
  lv_file_name     VARCHAR2(255) := 'item_import.csv';
  lv_virtual_dir   VARCHAR2(255) := 'UPLOAD_FILES';
  lv_table_name    VARCHAR2(30)  := 'IMPORT_DATA';
  lv_column_name   VARCHAR2(30)  := 'IMPORT_DATA';
  lv_foreign_key   NUMBER;
 
BEGIN
 
  -- Assign the current value of the sequence to a local variable.
  lv_foreign_key := import_master_s.currval;
 
  -- Check if you can read and insert a CLOB column.
  IF load_clob_from_file(lv_file_name
                        ,lv_virtual_dir
                        ,lv_table_name
                        ,lv_table_name
                        ,lv_foreign_key) = 1 THEN
 
    -- Display a successful subordinate routine.
    dbms_output.put_line('Subordinate routine succeeds.');
  ELSE
    -- Display a failed subordinate routine.
    dbms_output.put_line('Subordinate routine fails.');
  END IF;
 
END load_clob_from_file;
/

You can test this procedure against the log file with the following script file:

DECLARE
 
  -- Declare testing variables.
  lv_file_name     VARCHAR2(255) := 'item_import_ext_table.log';
  lv_virtual_dir   VARCHAR2(255) := 'UPLOAD_LOGS';
  lv_table_name    VARCHAR2(30)  := 'IMPORT_LOG';
  lv_column_name   VARCHAR2(30)  := 'IMPORT_LOG';
  lv_foreign_key   NUMBER;
 
BEGIN
 
  -- Assign the current value of the sequence to a local variable.
  lv_foreign_key := import_master_s.currval;
 
  dbms_output.put_line('Foreign key ['||lv_foreign_key||']');
 
  -- Check if you can read and insert a CLOB column.
  IF load_clob_from_file(lv_file_name
                        ,lv_virtual_dir
                        ,lv_table_name
                        ,lv_table_name
                        ,lv_foreign_key) = 1 THEN
 
    -- Display a successful subordinate routine.
    dbms_output.put_line('Subordinate routine succeeds.');
  ELSE
    -- Display a failed subordinate routine.
    dbms_output.put_line('Subordinate routine fails.');
  END IF;
 
END;
/

You now have the ability to read and store the source and log files in CLOB columns. The next step is to write a master function that writes the master row and calls the LOAD_CLOB_FROM_FILE function for the source file and each of the log files. That’s what the CLEANUP_EXTERNAL_FILES function provides.

Unfortunately, the Java logic requires using the logical and operation, which is two ampersands (&&). This requires that you turn off substitution variables in SQL*Plus. You do that by disabling DEFINE, like this:

SET DEFINE OFF

You can compile this Java library file after you’ve disabled LOAD_CLOB_FROM_FILE:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
CREATE OR REPLACE FUNCTION cleanup_external_files
( table_in           VARCHAR2
, data_directory_in  VARCHAR2
, log_directory_in   VARCHAR2 ) RETURN NUMBER IS
 
  -- Declare a local Attribute Data Type (ADT).
  TYPE list IS TABLE OF VARCHAR2(3);
 
  -- Declare a collection.
  lv_extension LIST := list('csv','log','bad','dis');
 
  -- Define a default return value.
  retval NUMBER := 0;
 
  -- Declare base target table name.
  lv_target_table  VARCHAR2(30) := 'IMPORT';
  lv_foreign_key   NUMBER;
 
  -- Decalre a cursor to find external tables.
  CURSOR check_source (cv_table_name VARCHAR2) IS
    SELECT   xt.file_name
    FROM    (SELECT   uxt.TABLE_NAME
             ,        ixt.column_value AS file_name
             FROM     user_external_tables uxt CROSS JOIN
             TABLE(list_files(get_directory_path(uxt.default_directory_name))) ixt) xt
    JOIN     user_external_locations xl ON xt.TABLE_NAME = xl.TABLE_NAME
    AND      xt.file_name = xl.location AND xt.TABLE_NAME = UPPER(cv_table_name);
 
  -- Declare a cursor to find files and compare for one input file name.
  CURSOR check_logs (cv_file_name VARCHAR2) IS
    SELECT   list.column_value
    FROM     TABLE(list_files(get_directory_path('UPLOAD_LOGS'))) list
    JOIN    (SELECT cv_file_name AS file_name FROM dual) FILTER
    ON       list.column_value = FILTER.file_name;
 
  -- Declare the function as autonomous.
  PRAGMA AUTONOMOUS_TRANSACTION;
 
BEGIN
 
  -- Master loop to check for source and log files.  
  FOR i IN check_source (table_in) LOOP
 
    -- Assign next sequence value to local variable.
    lv_foreign_key := import_master_s.nextval;
 
    -- Write the master record and commit it for the autonomous threads.
    INSERT INTO import_master
    VALUES (lv_foreign_key,'ITEM_IMPORT_EXT_TABLE');
    COMMIT;
 
    -- Process all file extensions.    
    FOR j IN 1..lv_extension.COUNT LOOP
 
      -- The source data file is confirmed by the CHECK_SOURCE cursor.
      IF lv_extension(j) = 'csv' THEN
 
        --  Load the source data file.
        -- ----------------------------------------------------------
        --  The RETVAL holds success or failure, this approach 
        --  suppresses an error when the file can't be loaded.
        --  It should only occur when there's no space available 
        --  in the target table.
        retval := load_clob_from_file(i.file_name
                                     ,data_directory_in
                                     ,lv_target_table||'_DATA'
                                     ,lv_target_table||'_DATA'
                                     ,lv_foreign_key);
                                     lv_foreign_key := lv_foreign_key + 1;
      ELSE
 
        -- Verify that log file exists before attempting to load it.
        FOR k IN check_logs (LOWER(table_in)||'.'||lv_extension(j)) LOOP
 
          --  Load the log, bad, or dis(card) file.
          -- ----------------------------------------------------------
          --  The RETVAL holds success or failure, as mentioned above.
          retval := load_clob_from_file(LOWER(table_in)||'.'||lv_extension(j)
                                       ,log_directory_in
                                       ,lv_target_table||'_'||lv_extension(j)
                                       ,lv_target_table||'_'||lv_extension(j)
                                       ,lv_foreign_key);
        END LOOP;
      END IF;
    END LOOP;
    retval := 1;
  END LOOP;
  RETURN retval;
END;
/

Deleting Files from Virtual Directories

After you’ve read the files through a query and uploaded the source and log files to the database, you need to cleanup the files. This can be done by using another Java library function, provided you granted read, write, and delete privileges to the internal Java permissions file.

The DeleteFile Java library deletes files from the file system. It doesn’t put them in the trash can for final delete, it removes them completely.

Now you can build the Java library that lets you delete a file. A quick caveat, this code includes an AND logical operator that is two ampersands (&&). SQL uses an ampersand (&) for substitution variables. You’ll need to suppress that behavior when you run this code.

You do that by issuing the following command to disable substitution variables in SQL*Plus:

1
SET DEFINE OFF

You create the DeleteFile library like this:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
CREATE OR REPLACE AND COMPILE JAVA SOURCE NAMED "DeleteFile" AS
  // Java import statements
  import java.io.File;
  import java.security.AccessControlException;
 
  // Class definition.
  public class DeleteFile
  {
    // Define variable(s).
    private static File file;
 
    // Define copyTextFile() method.
    public static void deleteFile(String fileName) throws AccessControlException {
 
      // CREATE files FROM canonical file names.
      file = NEW File(fileName);
 
      // DELETE file(s).
      IF (file.isFile() && file.delete()) {}}}
/

You need a PL/SQL Wrapper to call the library, and here it is:

1
2
3
4
CREATE OR REPLACE PROCEDURE delete_file (dfile VARCHAR2) IS
LANGUAGE JAVA
NAME 'DeleteFile.deleteFile(java.lang.String)';
/

You can call this separately or embed it inside the UPLOAD_LOGS function, which saves re-writing the logic to find any source or log files.

This has provided you with an external table import framework. You can extend the framework by wrapping the query in an object table function. Such a function would afford you the opportunity to cleanup the source and log files after the query operation.

Written by maclochlainn

March 5th, 2012 at 12:19 am

How to use object types?

with 5 comments

A tale of Oracle SQL object types, their constructors, and how you use them. This demonstrates what you can and can’t do and gives brief explanations about why.

The following creates a base SAMPLE_OBJECT data type and a sample_table
collection of the base SAMPLE_OBJECT data type.

CREATE OR REPLACE TYPE sample_object IS OBJECT
(id       NUMBER
,name     VARCHAR2(30));
/
 
CREATE OR REPLACE TYPE sample_table IS TABLE OF sample_object;
/

If the base SAMPLE_OBJECT data type were a Java object, the default constructor of an empty call parameter list would allow you to construct an instance variable. This doesn’t work for an Oracle object type because the default constructor is a formal parameter list of the object attributes in the positional order of their appearance in the declaration statement.

The test case on this concept is:

1
2
3
4
5
6
DECLARE
  lv_object_struct SAMPLE_OBJECT := sample_object();
BEGIN
  NULL;
END;
/

Running the program raises the following exception, which points to the object instance constructor from line 2 above:

  lv_object_struct SAMPLE_OBJECT := sample_object();
                                    *
ERROR at line 2:
ORA-06550: line 2, column 37:
PLS-00306: wrong number or types of arguments in call to 'SAMPLE_OBJECT'
ORA-06550: line 2, column 20:
PL/SQL: Item ignored

Changing the instantiation call to the Oracle design default, two null values let you create
an instance of the SAMPLE_OBJECT type. The following shows that concept, which works when the base object type allows null values.

1
2
3
4
5
6
DECLARE
  lv_object_struct SAMPLE_OBJECT := sample_object(NULL, NULL);
BEGIN
  NULL;
END;
/

If you want to have a null parameter constructor for an object type, you must implement a type and type body with an overloaded no argument constructor, like this:

1
2
3
4
5
CREATE OR REPLACE TYPE sample_object IS OBJECT
( id       NUMBER
, name     VARCHAR2(30)
, CONSTRUCTOR FUNCTION sample_object RETURN SELF AS RESULT);
/
1
2
3
4
5
6
7
8
9
CREATE OR REPLACE TYPE BODY sample_object IS
  CONSTRUCTOR FUNCTION sample_object RETURN SELF AS RESULT IS
    sample_obj SAMPLE_OBJECT := sample_object(NULL,NULL);
  BEGIN
    SELF := sample_obj;
    RETURN;
  END sample_object;
END;
/

Unlike Java, the addition of an overloaded constructor doesn’t drop the default constructor. You can also create a single parameter constructor that leverages the sequence like this:

1
2
3
4
5
6
CREATE OR REPLACE TYPE sample_object IS OBJECT
( id       NUMBER
, name     VARCHAR2(30)
, CONSTRUCTOR FUNCTION sample_object RETURN SELF AS RESULT
, CONSTRUCTOR FUNCTION sample_object (pv_name VARCHAR2) RETURN SELF AS RESULT);
/
1
2
3
4
5
6
7
8
9
10
11
12
13
14
CREATE OR REPLACE TYPE BODY sample_object IS
  CONSTRUCTOR FUNCTION sample_object RETURN SELF AS RESULT IS
    sample_obj SAMPLE_OBJECT := sample_object(sample_object_id.NEXTVAL,NULL);
  BEGIN
    SELF := sample_obj;
  END sample_object;
  CONSTRUCTOR FUNCTION sample_object (pv_name VARCHAR2) RETURN SELF AS RESULT IS
    sample_obj SAMPLE_OBJECT := sample_object(sample_object_id.NEXTVAL,pv_name);
  BEGIN
    SELF := sample_obj;
    RETURN;
  END sample_object;
END;
/

You can test the final object type and body with this anonymous block of code:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
SET SERVEROUTPUT ON SIZE UNLIMITED
 
DECLARE
  lv_object_struct1 SAMPLE_OBJECT := sample_object();
  lv_object_struct2 SAMPLE_OBJECT := sample_object('User Name');
  lv_object_struct3 SAMPLE_OBJECT := sample_object(1001,'User Name');
BEGIN
  dbms_output.put_line('lv_object_struct1.id   ['||lv_object_struct1.id||']');
  dbms_output.put_line('lv_object_struct1.name ['||lv_object_struct1.name||']');
  dbms_output.put_line('lv_object_struct2.id   ['||lv_object_struct2.id||']');
  dbms_output.put_line('lv_object_struct2.name ['||lv_object_struct2.name||']');
  lv_object_struct2.name := 'Changed Name';
  dbms_output.put_line('lv_object_struct2.id   ['||lv_object_struct2.id||']');
  dbms_output.put_line('lv_object_struct2.name ['||lv_object_struct2.name||']');
  dbms_output.put_line('lv_object_struct3.id   ['||lv_object_struct3.id||']');
  dbms_output.put_line('lv_object_struct3.name ['||lv_object_struct3.name||']');
END;
/

It prints to console:

lv_object_struct1.id   [1]
lv_object_struct1.name []
lv_object_struct2.id   [2]
lv_object_struct2.name [User Name]
lv_object_struct2.id   [2]
lv_object_struct2.name [Changed Name]
lv_object_struct3.id   [1001]
lv_object_struct3.name [User Name]

Hope this helps those looking for a quick syntax example and explanation.

Written by maclochlainn

February 14th, 2012 at 8:14 pm

Function or Procedure?

with 7 comments

Somebody asked for a simple comparison between a PL/SQL pass-by-value function and pass-by-reference procedure, where the procedure uses only an OUT mode parameter to return the result. This provides examples of both, but please note that a pass-by-value function can be used in SQL or PL/SQL context while a pass-by-reference procedure can only be used in another anonymous of named block PL/SQL program.

The function and procedure let you calculate the value of a number raised to a power of an exponent. The third parameter lets you convert the exponent value to an inverse value, like 2 to 1/2.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
CREATE OR REPLACE FUNCTION find_root_function
( pv_number   BINARY_DOUBLE
, pv_power    BINARY_DOUBLE
, pv_inverse  BINARY_INTEGER DEFAULT 0 ) RETURN BINARY_DOUBLE IS
 
  -- Declare local variable for return value.
  lv_result   BINARY_DOUBLE;
 
BEGIN
 
  -- If the inverse value is anything but zero calculate the inverse of the power.
  IF pv_inverse = 0 THEN
    lv_result := POWER(pv_number,pv_power);
  ELSE
    lv_result := POWER(pv_number,(1 / pv_power));
  END IF;
 
  RETURN lv_result;
END find_root_function;
/

You can test it with these to queries against the dual table:

SELECT TO_CHAR(find_root_function(4,3),'99,999.90') FROM dual;
SELECT TO_CHAR(find_root_function(125,3,1),'99,999.90') FROM dual;

The procedure does the same thing as the function. The difference is that the fourth parameter to the procedure returns the value rather than a formal return type like a function.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
CREATE OR REPLACE PROCEDURE find_root_procedure
( pv_number   IN     BINARY_DOUBLE
, pv_power    IN     BINARY_DOUBLE
, pv_inverse  IN     BINARY_INTEGER DEFAULT 0
, pv_return      OUT BINARY_DOUBLE ) IS
 
BEGIN
 
  -- If the inverse value is anything but zero calculate the inverse of the power.
  IF pv_inverse = 0 THEN
    pv_return := POWER(pv_number,pv_power);
  ELSE
    dbms_output.put_line('here');
    pv_return := POWER(pv_number,(1 / pv_power));
  END IF;
 
END find_root_procedure;
/

You can test it inside an anonymous block PL/SQL program, like this:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
DECLARE
 
  -- Declare input variables.
  lv_input   BINARY_DOUBLE;
  lv_power   BINARY_DOUBLE;
  lv_inverse BINARY_INTEGER;
  lv_output  BINARY_DOUBLE;
 
BEGIN
 
  -- Assign input values to variables.
  lv_input := '&1';
  lv_power := '&2';
  lv_inverse := '&3';
 
  -- Test raising to a power.
  find_root_procedure(lv_input, lv_power, lv_inverse, lv_output);
  dbms_output.put_line(TO_CHAR(lv_output,'99,999.90'));
 
  -- Test raising to an inverse power.
  find_root_procedure(lv_input, lv_power, lv_inverse, lv_output);
  dbms_output.put_line(TO_CHAR(lv_output,'99,999.90'));
 
END;
/

You can test it inside an anonymous block PL/SQL program, like the following example. For reference, the difference between PL/SQL and the SQL*Plus environment is large. The EXECUTE call is correct in SQL*Plus but would be incorrect inside a PL/SQL block for a Native Dynamic SQL (NDS) call. Inside a PL/SQL block you would use EXECUTE IMMEDIATE because it dispatches a call from the current running scope to a nested scope operation (see comment below).

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
-- SQL*Plus Test.
VARIABLE sv_input BINARY_DOUBLE
VARIABLE sv_power BINARY_DOUBLE
VARIABLE sv_inverse BINARY_DOUBLE
VARIABLE sv_output  BINARY_DOUBLE
 
-- Verify the null value of the session variable.
SELECT :sv_output AS ":sv_output" FROM dual;
 
BEGIN
 
  -- Prompt for local assignments and initialize output variable.
  :sv_input   := '&1';
  :sv_power   := '&2';
  :sv_inverse := '&3';
  :sv_output  := 0;
 
END;
/
 
-- Run the procedure in the SQL*Plus scope.
EXECUTE find_root_procedure(:sv_input, :sv_power, :sv_inverse, :sv_output);
 
-- Query the new value of the session variable.
SELECT TO_CHAR(:sv_output,'99,999.90') AS ":output" FROM dual;

As usual, I hope this helps folks beyond the one who asked. Comments are always welcome.

Written by maclochlainn

January 31st, 2012 at 5:00 pm