MacLochlainns Weblog

Michael McLaughlin's Technical Blog

Site Admin

Archive for the ‘Oracle Developer’ tag

Debug PL/SQL Web Pages

without comments

What happens when you can’t get a PL/SQL Web Toolkit to work because it only prints to a web page? That’s more tedious because any dbms_output.put_line command you embed only prints to a SQL*Plus session. The answer is quite simple, you create a test case and test it inside a SQL*Plus environment.

Here’s a sample web page that fails to run successfully …

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
CREATE OR REPLACE
  PROCEDURE html_table_values
  ( name_array   OWA_UTIL.VC_ARR
  , value_array  OWA_UTIL.VC_ARR ) IS
  BEGIN
    /* Print debug to SQL*Plus session. */
    FOR i IN 1..name_array.COUNT LOOP
      DBMS_OUTPUT.put_line('Value ['||name_array(i)||'='||value_array(i)||']');
    END LOOP;
 
    /* Open HTML page with the PL/SQL toolkit. */
    htp.print('<!DOCTYPE html>');
    htp.htmlopen;
    htp.headopen;
    htp.htitle('Test');
    htp.headclose;
    htp.bodyopen;
    htp.line;
    htp.print('Test');
    htp.line;
    htp.bodyclose;
    htp.htmlclose;
END;
/

You can test the program with the following anonymous block as the SYSTEM user, which is equivalent to the following URL:

http://localhost:8080/db/html_table_values?begin=1004&end=1012

The following test program lets you work:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
DECLARE
  x  OWA_UTIL.VC_ARR;
  y  OWA_UTIL.VC_ARR;
BEGIN
  /* Insert first row element. */
  x(1) := 'begin';
  y(1) := '1004';
 
  /* Insert second row element. */
  x(2) := 'end';
  y(2) := '1012';
 
  /* Call the anonymous schema's web page. */
  anonymous.html_table_values(x,y);
END;
/

It should print:

Value [begin=1004]
Value [end=1012]

I hope this helps those looking for a solution.

Written by maclochlainn

May 16th, 2016 at 5:18 pm

SQL Developer & PL/SQL

without comments

While SQL Developer installs with a dbms_output view, some organizations close it before they distribute images or virtual machine (VM) instances. This post shows you how to re-enable the Dbms Output view for SQL Developer.

SQL Developer DBMS_OUTPUT Configuration

SQLDeveloper1

  1. You need to open SQL Developer, which may look like this when the DBMS_OUTPUT view isn’t visible.

SQLDeveloper1

  1. You need to click on the View menu option in SQL Developer and choose the Dbms Output dropdown menu element.

SQLDeveloper1

  1. You should see a grayed-out Dbms Output view.

SQLDeveloper1

  1. You should type a simply “Hello World!” anonymous block program in PL/SQL, like the one shown in the drawing.

SQLDeveloper1

  1. After writing the “Hello World!” anonymous block program in PL/SQL, click the green arrow to start the statement and you will see two things. There is now a Script Output view between your console and Dbms Output views, and it should say “anonymous block completed.” Unfortunately, none of your output is displayed in the Dbms Output view because you need to enable it.

SQLDeveloper1

  1. If you hover over the Dbms Output view’s green arrow, you see the help message that describes the behavior of the green arrow. The Dbms Output green arrow lets you enable the Dbms Output view for output.

SQLDeveloper1

  1. After you click the Dbms Output view’s green arrow, you receive a Select Connection prompt for the view. Make sure you have the right user, and click the OK button to continue.

SQLDeveloper1

  1. After you create the connection for the Dbms Output stream, the view area becomes white rather than gray.

SQLDeveloper1

  1. Click the green arrow to start the statement and you will see the “Hello World!” string in the Dbms Output view.”

As always, I hope this helps those looking for a solution.

Written by maclochlainn

May 13th, 2016 at 10:55 am

Using a Sparse Index

with 2 comments

My vacation from my blog is officially over. The question that I’m answering today is: How can you pass a set of non-sequential ID values to a function and return a result set? You can solve the problem by passing an ADT (Attribute Data Type) or UDT (User Defined Type) variable into a subquery of a cursor. The subquery leverages the TABLE function to translate the ADT or UDT into SQL result set, which is equivalent to a comma-delimited list of values.

You can also solve this problem with Native Dynamic SQL (NDS). However, the person who posed the question didn’t want to use NDS to build out a variable length list of comma-delimited numbers.

You need to create three object types for this example. They are:

  • a list of numbers
  • a record structure, declared as an object type without methods
  • a list of the record structure

These are the SQL commands to create the required data types:

CREATE OR REPLACE
  TYPE list_ids IS TABLE OF NUMBER;
/
CREATE OR REPLACE
  TYPE item_struct IS OBJECT
  ( item_id       NUMBER
  , item_title    VARCHAR2(80)
  , release_date  DATE );
/
CREATE OR REPLACE
  TYPE item_struct_list IS TABLE OF item_struct;
/

Next, you create a nonsynchronous function. It takes a sparsely populated list of values that map to the surrogate key of the column, which is typically the table’s primary key column. It returns a collection of the item_struct object type. This type of function is an object-table function.

The code follows:

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
CREATE OR REPLACE
  FUNCTION nonsynchronous 
  ( pv_list_ids  LIST_IDS ) RETURN item_struct_list IS
    /* Declare a record data structure list. */
    lv_struct_list  ITEM_STRUCT_LIST := item_struct_list();
 
    /* Declare a sparsely indexed list of film items. */
    CURSOR get_items
    ( cv_list_ids  LIST_IDS ) IS
      SELECT   item_id AS item_id
      ,        item_title
      ||       CASE
                 WHEN item_subtitle IS NOT NULL THEN
                   ': '|| item_subtitle
               END AS item_title
      ,        release_date AS release_date
      FROM     item
      WHERE    item_id IN (SELECT *
                           FROM   TABLE(cv_list_ids))
      ORDER BY item_id;
BEGIN
  /* Lood through the sparsely populated list of numbers. */
  FOR i IN get_items(pv_list_ids) LOOP
    lv_struct_list.EXTEND;
    lv_struct_list(lv_struct_list.COUNT) := item_struct( item_id      => i.item_id
                                                       , item_title   => i.item_title
                                                       , release_date => i.release_date );
  END LOOP;
 
  /* Return the record structure list. */
  RETURN lv_struct_list;
END;
/

The foregoing nonsynchronous function uses a nested query that transforms to a result set on lines 18 and 19. In the execution block of the program, it uses a call to the item_struct structure to capture and assign row values to an element of the lv_struct_list variable.

You can now test the nonsynchronous function with the following query:

COL item_id      FORMAT 9999  HEADING "Item|ID #"
COL item_title   FORMAT A40   HEADING "Item Title"
COL release_date FORMAT A11   HEADING "Release|Date"
SELECT   *
FROM     TABLE(nonsynchronous(list_ids(1002, 1013, 1007)));

The query returns the record set as an ordered list in the result set, like:

Item					       Release
 ID # Item Title			       Date
----- ---------------------------------------- -----------
 1002 Star Wars I: Phantom Menace	       04-MAY-99
 1007 RoboCop				       24-JUL-03
 1013 The DaVinci Code			       19-MAY-06

I hope this answers the question about how to get results sets with sparsely populated ID values.

Written by maclochlainn

May 11th, 2016 at 1:37 am

REGEXP_LIKE Behavior

with one comment

Often, the biggest problem with regular expressions is that those who use them sometimes don’t use them correctly. A great example occurs in the Oracle Database with the REGEXP_LIKE function. For example, some developer use the following to validate whether a string is a number but it only validates whether the first character is a number.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
DECLARE
  lv_input  VARCHAR2(100);
BEGIN
  /* Assign input value. */
  lv_input := '&input';
 
  /* Check for numeric string. */
  IF REGEXP_LIKE(lv_input,'[[:digit:]]') THEN
    dbms_output.put_line('It''s a number.');
  ELSE
    dbms_output.put_line('It''s a string.');
  END IF;
END;
/

When they test numbers it appears to works, it even appears to work when the test string start with number, but it fails with any string that starts with a character. That’s because the REGEXP_LIKE function on line 8 only checks the first character, but the following checks all the characters in the string.

8
  IF REGEXP_LIKE(lv_inputs(i),'[[:digit:]]{'||LENGTH(lv_inputs(i))||'}') THEN

You can also fix it with the following non-Posix solution:

8
  IF REGEXP_LIKE(lv_input,'[[0-9]]') THEN

You can add a collection to the program and use it to test single-digit, double-digit, and string with a leading integer. Save the program as test.sql and you can test three conditions with one call.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
DECLARE
  /* Declare the local collection type. */
  TYPE inputs IS TABLE OF VARCHAR2(100);
 
  /* Declare a local variable of the collection type. */
  lv_inputs  INPUTS;
BEGIN
  /* Assign the inputs to the collection variable. */
  lv_inputs := inputs('&1','&2','&3');
 
  /* Read through the collection and print whether it's an number or string. */
  FOR i IN 1..lv_inputs.COUNT LOOP
    IF REGEXP_LIKE(lv_inputs(i),'[[:digit:]]{'||LENGTH(lv_inputs(i))||',}') THEN
      dbms_output.put_line('It''s a number.');
    ELSE
      dbms_output.put_line('It''s a string.');
    END IF;
  END LOOP;
END;
/

You can run the test.sql program like this:

SQL> @test.sql 1 12 1a

It prints:

It's a number.
It's a number.
It's a string.

As always, I hope this helps those looking for a solution.

Written by maclochlainn

September 30th, 2015 at 7:23 pm

SQL*Plus Tricks

with 3 comments

Have you ever wondered how to leverage substitution variables in anonymous block programs? There are several tricks that you can use beyond passing numeric and string values to local variable. The generic default appears to take a number unless you cast it as a string but that’s not really the whole story. The first two are standard examples of how to use numeric and string substitution values.

The following accept a numeric substitution value:

1
2
3
4
5
6
7
8
9
10
DECLARE
  lv_input  NUMBER;
BEGIN
  /* Assign substitution value to local variable. */
  lv_input := &input;
 
  /* Print the local variable. */
  dbms_output.put_line('['||lv_input||']');
END;
/

The following accept a string substitution value, casts the input as a string, assigns the string value to a 4,000 character length local variable, checks whether the 4,000 character length is greater than 10, and assigns the first 10 characters to the lv_parse_input variable:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
DECLARE
  lv_unparsed_input  VARCHAR2(4000);
  lv_parsed_input    VARCHAR2(10);
BEGIN
  /* Assign substitution value to local variable. */
  lv_unparsed_input := '&input';
 
  /* Check size of input value. */
  IF LENGTH(lv_unparsed_input) > 10 THEN
    lv_parsed_input := SUBSTR(lv_unparsed_input,1,10);
  END IF;
 
  /* Print the local variable. */
  dbms_output.put_line('Print {lv_parsed_input}: ['||lv_parsed_input||']');
END;
/

Next, let’s examine two tricks. The first passes a case insensitive variable name and the second passes a case sensitive variable name as a parameter to an anonymous block program.

This declares an anonymous block program that uses a substitution value as a variable name:

1
2
3
4
5
6
7
DECLARE
  mine  VARCHAR2(10) := 'Default';
BEGIN
  /* Print the local variable's value. */
  dbms_output.put_line('Print {mine} variable value: ['||&input||']');
END;
/

When you run the anonymous block, you’re prompted for an input variable. You provide a case insensitive variable name as the input value:

Enter value for input: MINE
old   5:   dbms_output.put_line('['||&input||']');
new   5:   dbms_output.put_line('['||MINE||']');
Print {mine} variable value: [Default]

The downside of this approach, yields an ORA-06550 and PLS-00201 exception. Neither of these can be caught because Oracle raises the errors during parsing when the variable name isn’t a 100% case insensitive match. The same type of problem occurs in the next example when the input variable isn’t a 100% case sensitive match.

You can rewrite the program to handle case insensitive variables like this:

1
2
3
4
5
6
7
DECLARE
  "Mine"  VARCHAR2(10) := 'Default';
BEGIN
  /* Print the local variable's value. */
  dbms_output.put_line('Print {mine} variable value: ['||"&input"||']');
END;
/

When you run the anonymous block, you’re prompted for an input variable. You provide a case sensitive variable name as the input value:

Enter value for input: Mine
old   5:   dbms_output.put_line('['||&input||']');
new   5:   dbms_output.put_line('['||"Mine"||']');
Print {Mine} variable value: [Default]

Hope this helps those looking for a solution.

Written by maclochlainn

September 24th, 2015 at 1:19 am

Using CALIBRATE_IO

without comments

Using Oracle’s Resource Manager requires you to understand the IO dynamics. The first step requires you to run the CALIBRATE_IO procedure from the DBMS_RESOURCE_MANAGER package.

Oracle provides some great examples about how to use the CALIBRATE_IO procedure of the DBMS_RESOURCE_MANAGER package in the Oracle Database Database PL/SQL Packages and Types Reference. The CALIBRATE_IO procedure returns the best answer when you provide a valid number of files, which you can capture by querying the V$ASM_DISK view.

The following code queries the view and assigns the value to a session level variable:

CLEAR BREAKS
CLEAR COLUMNS
CLEAR COMPUTES
 
VARIABLE files NUMBER
 
BEGIN
  SELECT COUNT(DISTINCT name) disks
  INTO :files
  FROM v$asm_disk;
END;
/

When you have the number of files, you can calibrate the IO with the following anonymous block. The query should always work but just in case the NVL function on line 9 assigns the default number of files.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
DECLARE
  lv_num_physical_disks BINARY_INTEGER; — v$asm_disk
  lv_max_latency BINARY_INTEGER := 10;
  lv_max_iops BINARY_INTEGER;
  lv_max_mbps BINARY_INTEGER;
  lv_actual_latency BINARY_INTEGER;
BEGIN
  /* Assign actual files to anonymous block variable. */
  lv_num_physical_disks := NVL(:files,2);
 
  /* Run the calibrate_io procedure. */
  DBMS_RESOURCE_MANAGER.CALIBRATE_IO(
      num_physical_disks => lv_num_physical_disks
    , max_latency => lv_max_latency
    , max_iops => lv_max_iops
    , max_mbps => lv_max_mbps
    , actual_latency => lv_actual_latency);
END;
/

You can query the results like this:

SELECT max_iops
,      max_mbps
,      max_pmbps
,      latency
,      num_physical_disks
FROM   dba_rsrc_io_calibrate;

It should show results like these:

MAX_IOPS MAX_MBPS MAX_PMBPS LATENCY NUM_PHYSICAL_DISKS
-------- -------- --------- ------- ------------------
    8894      443       294       9                 18

Hope this helps those using the CALIBRATE_IO procedure of the DBMS_RESOURCE_MANAGER package.

Written by maclochlainn

August 31st, 2015 at 8:59 pm

Free Oracle Tuning Book

with 2 comments

Quick Start Guide to Oracle Query TuningWho can resist a free Rich Nimeiec book on SQL Tuning? O.K., those who know everything can resist. If you’re like me, this is an opportunity to learn from Rich. Click on the book image or this link to get a free copy, or if you want to pay $10 for a copy click here to buy Quick Start Guide to Oracle Query Tuning: Tips for DBAs and Developers from Amazon.com.

The book is four chapters long, is a 129 pages in length, and is in a PDF format. The outline is:

  1. Query Tuning: Developer and Beginning DBA
  2. Query Tuning: Basics for DBAs and Developers
  3. Advanced Performance Tuning
  4. Tips for Tuning When You Have Everything Tuned

Enjoy reading it. His more comprehensive book is Oracle Database 11g Release 2 Performance Tuning Tips & Techniques (Oracle Press) and it’s $30, but it’s written for an advanced audience (more or less OCA or higher).

Written by maclochlainn

August 31st, 2015 at 11:24 am

Use an object in a query?

with 2 comments

Using an Oracle object type’s instance in a query is a powerful capability. Unfortunately, Oracle’s SQL syntax doesn’t make it immediately obvious how to do it. Most get far enough to put it in a runtime view (a subquery in the FROM clause), but then they get errors like this:

SELECT	 instance.get_type()
         *
ERROR AT line 4:
ORA-00904: "INSTANCE"."GET_TYPE": invalid identifier

The problem is how Oracle treats runtime views, which appears to me as a casting error. Somewhat like the ORDER BY clause irregularity that I noted in July, the trick is complete versus incomplete syntax. The following query fails and generates the foregoing error:

1
2
3
4
SELECT instance.get_type() AS object_type
,      instance.to_string() AS object_content
FROM  (SELECT dependent()AS instance
       FROM   dual);

If you add a table alias, or name, to the runtime view on line 4, it works fine:

1
2
3
4
SELECT cte.instance.get_type() AS object_type
,      cte.instance.to_string() AS object_content
FROM  (SELECT dependent() AS instance
       FROM   dual) cte;

That is the trick. You use an alias for the query, which assigns the alias like a table reference. The reference lets you access instance methods in the scope of a query. Different columns in the query’s SELECT-list may return different results from different methods from the same instance of the object type.

You can also raise an exception if you forget the open and close parentheses for a method call to a UDT, which differs from how Oracle treats no argument functions and procedures. That type of error would look like this:

SELECT cte.instance.get_type AS object_type
       *
ERROR AT line 1:
ORA-00904: : invalid identifier

It is an invalid identifier because there’s no public variable get_type, and a method is only found by using the parenthesis and a list of parameters where they’re required.

The object source code is visible by clicking on the expandable label below.

As always, I hope this helps those solving problems.

Written by maclochlainn

August 22nd, 2015 at 5:23 pm

ORDER BY CASE

with 8 comments

Sometimes I give parts of a solution to increase the play time to solve a problem. I didn’t anticipate a problem when showing how to perform a sort operation with a CASE statement. It’s a sweet solution when you need to sort something differently than a traditional ascending or descending sort.

I gave my students this ORDER BY clause as an example:

  CASE
    WHEN filter = 'Debit' THEN 1
    WHEN filter = 'Credit' THEN 2
    WHEN filter = 'Total' THEN 3
  END;

It raises the following error in MySQL for students:

ERROR 1064 (42000): You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near 'ORDER BY
  CASE
    WHEN filter = 'Debit' THEN 1
    WHEN filter = 'Credit' THEN' at line 6

It raises the following error in Oracle for some students:

  CASE
  *
ERROR AT line 7:
ORA-01785: ORDER BY item must be the NUMBER OF a SELECT-list expression

So, I built a little test case to replicate the problem and error message they encountered:

SQL> SELECT 'Debit' AS filter FROM dual
  2  UNION ALL
  3  SELECT 'Credit' AS filter FROM dual
  4  UNION ALL
  5  SELECT 'Total' AS filter FROM dual
  6  ORDER BY
  7    CASE
  8      WHEN filter = 'Debit' THEN 1
  9      WHEN filter = 'Credit' THEN 2
 10      WHEN filter = 'Total' THEN 3
 11    END;

They said, great but how can you fix it? That’s simple, with a Common Table Expression (CTE) in Oracle or with an inline view in MySQL. The Oracle CTE solution is:

  1  WITH results AS
  2  (SELECT 'Debit' AS filter FROM dual
  3   UNION ALL
  4   SELECT 'Credit' AS filter FROM dual
  5   UNION ALL
  6   SELECT 'Total' AS filter FROM dual)
  7  SELECT filter
  8  FROM   results
  9  ORDER BY
 10    CASE
 11	 WHEN filter = 'Debit'  THEN 1
 12	 WHEN filter = 'Credit' THEN 2
 13	 WHEN filter = 'Total'  THEN 3
 14    END;

There are two MySQL solutions. One simply removes the FROM dual clauses from the query components and the other uses an inline view in the FROM clause. This is the inline view:

SELECT filter
FROM  (SELECT 'Debit' AS filter FROM dual
       UNION ALL
       SELECT 'Credit' AS filter FROM dual
       UNION ALL
       SELECT 'Total' AS filter FROM dual) resultset
ORDER BY
  CASE
    WHEN filter = 'Debit' THEN 1
    WHEN filter = 'Credit' THEN 2
    WHEN filter = 'Total' THEN 3
  END;

This is the solution without the FROM dual clauses:

SELECT 'Debit' AS filter
UNION ALL
SELECT 'Credit' AS filter
UNION ALL
SELECT 'Total' AS filter
ORDER BY
  CASE
    WHEN filter = 'Debit' THEN 1
    WHEN filter = 'Credit' THEN 2
    WHEN filter = 'Total' THEN 3
  END;

Both MySQL solutions yield the following:

+--------+
| filter |
+--------+
| Debit  |
| Credit |
| Total  |
+--------+
3 rows in set (0.00 sec)

It puts the fabricating query inside a result set, and then lets you use the column alias to filter the set. If you have a better approach, please share it.

Written by maclochlainn

July 8th, 2015 at 10:06 pm

Mac SQL Developer Install

without comments

This how you install SQL Developer on Mac OS Yosemite. The first thing you need to do is download and install Java 8, not Java 7 on your Mac OS Yosemite as suggested on some web sites. You can determine whether or not Java is installed by running the following command:

Mac-Pro-3:~ username$ java -version
No Java runtime present, requesting install.

You must accept the Java license to install Java 8 on the Mac OS X operating system:

YosemiteInstallJava_01

You have the option of installing the Java SDK or JDK. I’ve opted to install Netbeans 8 with JDK 8u45, as you can tell from the screen capture after you launched the file:

YosemiteInstallJava_02

It is a standard Mac OS installation, which is why I didn’t bother showing any dialog messages. After installing the Java JDK or SDK, you should download SQL Developer 4.1 from Oracle’s web site. Below is a screen shot of the Oracle download web page where I’ve accepted the license agreement:

SQLDeveloperDownload

If you attempt to launch the installation and you’ve set your Mac Security to the “Mac App Store and identified developers” setting, you should raise the following exception:

SQLDeveloperInstall_01

If you reset the Mac Security to an “Anywhere” setting, you can install Oracle SQL Developer on Yosemite. Just make sure you reset it to the “Mac App Store and identified developers” setting after you install SQL Developer.

If you launch SQL Developer with the Security “Anywhere” setting, it displays the following dialog:

SQLDeveloperInstall_02

After you launch the program, you will see the following progress dialog:

SQLDeveloperInstall_03

The last step of the installation launches SQL Developer, as shown below:

SQLDeveloperInstall_04

Click the Connections icon to create an initial connection, like the following:

SQLDeveloperInstall_05

After connecting to the database, you can write and execute a query as shown in the next screen capture:

SQLDeveloperInstall_06

As always, I hope that this helps those who require an example to install SQL Server on a Mac OS.

Written by maclochlainn

June 12th, 2015 at 3:08 am