Archive for the ‘Oracle 21c’ Category
Oracle Unit Test
A unit test script may contain SQL or PL/SQL statements or it may call another script file that contains SQL or PL/SQL statements. Moreover, a script file is a way to bundle several activities into a single file because most unit test programs typically run two or more instructions as unit tests.
Unconditional Script File
You can write a simple unit test like the example program provided in the Lab 1 Help Section, which includes conditional logic. However, you can write a simpler script that is unconditional and raises exceptions when preconditions do not exist.
The following script file creates a one table and one_s sequence. The DROP TABLE and DROP SEQUENCE statements have the same precondition, which is that the table or sequence must previously exist.
-- Drop table one. DROP TABLE one; -- Crete table one. CREATE TABLE one ( one_id NUMBER , one_text VARCHAR2(10)); -- Drop sequence one_s. DROP SEQUENCE one_s; -- Create sequence one_s. CREATE SEQUENCE one_s; |
After writing the script file, you can save it in the lab2 subdirectory as the unconditional.sql file. After you login to the SQL*Plus environment from the lab2 subdirectory. You call the unconditional.sql script file from inside the SQL*Plus environment with the following syntax:
@unconditional.sql |
It will display the following output, which raises an exception when the one table or one_s sequence does not already exist in the schema or database:
DROP TABLE one * ERROR at line 1: ORA-00942: table or view does not exist Table created. DROP SEQUENCE one_s * ERROR at line 1: ORA-02289: sequence does not exist Sequence created. |
An unconditional script raises exceptions when a precondition of the statement does not exist. The precondition is not limited to objects, like the table or sequence; and the precondition may be specific data in one or several rows of one or several tables. You can avoid raising conditional errors by writing conditional scripts.
Conditional Script File
A conditional script file contains statements that check for a precondition before running a statement, which effectively promotes their embedded statements to a lambda function. The following logic recreates the logic of the unconditional.sql script file as a conditional script file:
-- Conditionally drop a table and sequence. BEGIN FOR i IN (SELECT object_name , object_type FROM user_objects WHERE object_name IN ('ONE','ONE_S') ORDER BY object_type ) LOOP IF i.object_type = 'TABLE' THEN EXECUTE IMMEDIATE 'DROP TABLE '||i.object_name||' CASCADE CONSTRAINTS'; ELSE EXECUTE IMMEDIATE 'DROP SEQUENCE '||i.object_name; END IF; END LOOP; END; / -- Crete table one. CREATE TABLE one ( one_id NUMBER , one_text VARCHAR2(10)); -- Create sequence one_s. CREATE SEQUENCE one_s; |
You can save this script in the lab2 subdirectory as conditional.sql and then unit test it in SQL*Plus. You must manually drop the one table and one_s sequence before running the conditional.sql script to test the preconditions.
You will see that the conditional.sql script does not raise an exception because the one table or one_s sequence is missing. It should generate output to the console, like this:
PL/SQL procedure successfully completed. Table created. Sequence created. |
As a rule, you should always write conditional script files. Unconditional script files throw meaningless errors, which may cause your good code to fail a deployment test that requires error free code.
Selective Aggregation
Selective Aggregation
Learning Outcomes
- Learn how to combine CASE operators and aggregation functions.
- Learn how to selective aggregate values.
- Learn how to use SQL to format report output.
Selective aggregation is the combination of the CASE operator and aggregation functions. Any aggregation function adds, sums, or averages the numbers that it finds; and when you embed the results of a CASE operator inside an aggregation function you get a selective result. The selectivity is determined by the WHEN clause of a CASE operator, which is more or less like an IF statement in an imperative programming language.
The prototype for selective aggregation is illustrated with a SUM function below:
SELECT SUM(CASE WHEN left_operand = right_operand THEN result WHEN left_operand > right_operand THEN result WHEN left_operand IN (SET OF comma-delimited VALUES) THEN result WHEN left_operand IN (query OF results) THEN result ELSE alt_result END) AS selective_aggregate FROM some_table; |
A small example let’s you see how selective aggregation works. You create a PAYMENT table and PAYMENT_S sequence for this example, as follows:
-- Create a PAYMENT table. CREATE TABLE payment ( payment_id NUMBER , payment_date DATE CONSTRAINT nn_payment_1 NOT NULL , payment_amount NUMBER(20,2) CONSTRAINT nn_payment_2 NOT NULL , CONSTRAINT pk_payment PRIMARY KEY (payment_id)); -- Create a PAYMENT_S sequence. CREATE SEQUENCE payment_s; |
After you create the table and sequence, you should insert some data. You can match the values below or choose your own values. You should just insert values for a bunch of rows.
View Anonymous PL/SQL Block →
You can populate data with the anonymous PL/SQL block, which creates 10,000 random rows in the payment table. Please note thatyou will get different payment dates and amounts each time you run the script.
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 | DECLARE -- Create local collection data types. TYPE pmtval IS TABLE OF NUMBER(20,2); TYPE smonth IS TABLE OF VARCHAR2(3); -- Create variable to hold the list of payments. payments PMTVAL := pmtval(); -- Declare month arrays. short_month SMONTH := smonth('JAN','FEB','MAR','APR','MAY','JUN' ,'JUL','AUG','SEP','OCT','NOV','DEC'); -- Declare variable values. month VARCHAR2(3); year NUMBER := '2019'; pmt_date DATE; tpmt_date VARCHAR2(11); -- Declare default number of random payments. payment_number NUMBER := 10000; BEGIN -- Populate payment list. FOR i IN 1..payment_number LOOP payments.EXTEND; SELECT ROUND(dbms_random.value() * 1000,0) || '.' || ROUND(dbms_random.value() * 100,0) INTO payments(payments.COUNT) FROM dual; END LOOP; -- Create and populate payment date and amount. FOR i IN 1..payment_number LOOP -- Assign random month value. month := short_month(dbms_random.value(1,short_month.COUNT)); -- Assign random day of the month value and assemble random date. IF month IN ('JAN','MAR','MAY','JUL','AUG','OCT','DEC') THEN pmt_date := ROUND(dbms_random.value(1,31),0) || '-' || month || '-' || year; ELSIF month IN ('APR','JUN','SEP','NOV') THEN pmt_date := ROUND(dbms_random.value(1,30),0) || '-' || month || '-' || year; ELSE pmt_date := ROUND(dbms_random.value(1,28),0) || '-' || month || '-' || year; END IF; -- Insert values into the PAYMENT table. INSERT INTO payment ( payment_id, payment_date, payment_amount ) VALUES ( payment_s.NEXTVAL, pmt_date, payments(i)); END LOOP; -- Commit the writes. COMMIT; END; / |
After inserting 10,000 rows, you can get an unformatted total with the following query:
-- Query total amount. SELECT SUM(payment_amount) AS payment_total FROM payment; |
It outputs the following:
PAYMENT_TOTAL ------------- 5011091.75 |
You can nest the result inside the TO_CHAR function to format the output, like
-- Query total formatted amount. SELECT TO_CHAR(SUM(payment_amount),'999,999,999.00') AS payment_total FROM payment; |
It outputs the following:
PAYMENT_TOTAL --------------- 5,011,091.75 |
Somebody may suggest that you use a PIVOT function to rotate the data into a summary by month but the PIVOT function has limits. The pivoting key must be numeric and the column values will use only those numeric values.
-- Pivoted summaries by numeric monthly value. SELECT * FROM (SELECT EXTRACT(MONTH FROM payment_date) payment_month , payment_amount FROM payment) PIVOT (SUM(payment_amount) FOR payment_month IN (1,2,3,4,5,6,7,8,9,10,11,12)); |
It outputs the following:
1 2 3 4 5 6 7 8 9 10 11 12 ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- 245896.55 430552.36 443742.63 457860.27 470467.18 466370.71 415158.28 439898.72 458998.09 461378.56 474499.22 246269.18 |
You can use selective aggregation to get the results by a character label, like
SELECT SUM( CASE WHEN EXTRACT(MONTH FROM payment_date) = 1 AND EXTRACT(YEAR FROM payment_date) = 2019 THEN payment_amount END) AS "JAN" , SUM( CASE WHEN EXTRACT(MONTH FROM payment_date) = 2 AND EXTRACT(YEAR FROM payment_date) = 2019 THEN payment_amount END) AS "FEB" , SUM( CASE WHEN EXTRACT(MONTH FROM payment_date) = 3 AND EXTRACT(YEAR FROM payment_date) = 2019 THEN payment_amount END) AS "MAR" , SUM( CASE WHEN EXTRACT(MONTH FROM payment_date) IN (1,2,3) AND EXTRACT(YEAR FROM payment_date) = 2019 THEN payment_amount END) AS "1FQ" , SUM( CASE WHEN EXTRACT(MONTH FROM payment_date) = 4 AND EXTRACT(YEAR FROM payment_date) = 2019 THEN payment_amount END) AS "APR" FROM payment; |
It outputs the following:
JAN FEB MAR 1FQ APR ---------- ---------- ---------- ---------- ---------- 245896.55 430552.36 443742.63 1120191.54 457860.27 |
You can format the output with a combination of the TO_CHAR and LPAD functions. The TO_CHAR allows you to add a formatting mask, complete with commas and two mandatory digits to the right of the decimal point. The reformatted query looks like
COL JAN FORMAT A13 HEADING "Jan" COL FEB FORMAT A13 HEADING "Feb" COL MAR FORMAT A13 HEADING "Mar" COL 1FQ FORMAT A13 HEADING "1FQ" COL APR FORMAT A13 HEADING "Apr" SELECT LPAD(TO_CHAR(SUM( CASE WHEN EXTRACT(MONTH FROM payment_date) = 1 AND EXTRACT(YEAR FROM payment_date) = 2019 THEN payment_amount END),'9,999,999.00'),13,' ') AS "JAN" , LPAD(TO_CHAR(SUM( CASE WHEN EXTRACT(MONTH FROM payment_date) = 2 AND EXTRACT(YEAR FROM payment_date) = 2019 THEN payment_amount END),'9,999,999.00'),13,' ') AS "FEB" , LPAD(TO_CHAR(SUM( CASE WHEN EXTRACT(MONTH FROM payment_date) = 3 AND EXTRACT(YEAR FROM payment_date) = 2019 THEN payment_amount END),'9,999,999.00'),13,' ') AS "MAR" , LPAD(TO_CHAR(SUM( CASE WHEN EXTRACT(MONTH FROM payment_date) IN (1,2,3) AND EXTRACT(YEAR FROM payment_date) = 2019 THEN payment_amount END),'9,999,999.00'),13,' ') AS "1FQ" , LPAD(TO_CHAR(SUM( CASE WHEN EXTRACT(MONTH FROM payment_date) = 4 AND EXTRACT(YEAR FROM payment_date) = 2019 THEN payment_amount END),'9,999,999.00'),13,' ') AS "APR" FROM payment; |
It displays the formatted output:
Jan Feb Mar 1FQ Apr ------------- ------------- ------------- ------------- ------------- 245,896.55 430,552.36 443,742.63 1,120,191.54 457,860.27 |
INSERT Statement
INSERT Statement
Learning Outcomes
- Learn how to use positional- and named-notation in INSERT statements.
- Learn how to use the VALUES clause in INSERT statements.
- Learn how to use subqueries in INSERT statements.
The INSERT statement lets you enter data into tables and views in two ways: via an INSERT statement with a VALUES clause and via an INSERT statement with a query. The VALUES clause takes a list of literal values (strings, numbers, and dates represented as strings), expression values (return values from functions), or variable values.
Query values are results from SELECT statements that are subqueries (covered earlier in this appendix). INSERT statements work with scalar, single-row, and multiple-row subqueries. The list of columns in the VALUES clause or SELECT clause of a query (a SELECT list) must map to the positional list of columns that defines the table. That list is found in the data dictionary or catalog. Alternatively to the list of columns from the data catalog, you can provide a named list of those columns. The named list overrides the positional (or default) order from the data catalog and must provide at least all mandatory columns in the table definition. Mandatory columns are those that are not null constrained.
Oracle databases differ from other databases in how they implement the INSERT statement. Oracle doesn’t support multiple-row inserts with a VALUES clause. Oracle does support default and override signatures as qualified in the ANSI SQL standards. Oracle also provides a multiple- table INSERT statement. This section covers how you enter data with an INSERT statement that is based on a VALUES clause or a subquery result statement. It also covers multiple-table INSERT statements.
The INSERT statement has one significant limitation: its default signature. The default signature is the list of columns that defines the table in the data catalog. The list is defined by the position and data type of columns. The CREATE statement defines the initial default signature, and the ALTER statement can change the number, data types, or ordering of columns in the default signature.
The default prototype for an INSERT statement allows for an optional column list that overrides the default list of columns. When you provide the column list you choose to implement named-notation, which is the right way to do it. Relying on the insertion order of the columns is a bad idea. An INSERT statement without a list of column names is a position-notation statement. Position-notation is bad because somebody can alter that order and previously written INSERT statements will break or put data in the wrong columns.
Like methods in OOPLs, an INSERT statement without the optional column list constructs an instance (or row) of the table using the default constructor. The override constructor for a row is defined by any INSERT statement when you provide an optional column list. That’s because it overrides the default constructor.
The generic prototype for an INSERT statement is confusing when it tries to capture both the VALUES clause and the result set from a query. Therefore, I’ve opted to provide two generic prototypes.
Insert by value
The first uses the VALUES clause:
INSERT INTO table_name [( column1, column2, column3, ...)] VALUES ( value1, value2, value3, ...); |
Notice that the prototype for an INSERT statement with the result set from a query doesn’t use the VALUES clause at all. A parsing error occurs when the VALUES clause and query both occur in an INSERT statement.
The second prototype uses a query and excludes the VALUES clause. The subquery may return one to many rows of data. The operative rule is that all columns in the query return the same number of rows of data, because query results should be rectangles—rectangles made up of one to many rows of columns.
Insert by subquery
Here’s the prototype for an INSERT statement that uses a subquery:
INSERT INTO table_name [( column1, column2, column3, ...)] ( SELECT value1, value2, value3, ... FROM table_name WHERE ...); |
A query, or SELECT statement, returns a SELECT list. The SELECT list is the list of columns, and it’s evaluated by position and data type. The SELECT list must match the definition of the table or the override signature provided.
Default signatures present a risk of data corruption through insertion anomalies, which occur when you enter bad data in tables. Mistakes transposing or misplacing values can occur more frequently with a default signature, because the underlying table structure can change. As a best practice, always use named notation by providing the optional list of values; this should help you avoid putting the right data in the wrong place.
The following subsections provide examples that use the default and override syntax for INSERT statements in Oracle databases. The subsections also cover multiple-table INSERT statements and a RETURNING INTO clause, which is an extension of the ANSI SQL standard. Oracle uses the RETURNING INTO clause to manage large objects, to return autogenerated identity column values, and to support some of the features of Oracle’s dynamic SQL. Note that Oracle also supports a bulk INSERT statement, which requires knowledge of PL/SQL.
Insert by Values →
An INSERT statement with a VALUES clause can only insert one row at a time in and Oracle database. Other databases, like Microsoft SQL Server and MySQL allow you to insert a comma delimited set of values inside the VALUES clause. Oracle adheres to the ANSI standard that support single row inserts with a VALUES clause and multiple row inserts with a subquery.
Inserting by the VALUES clause is the most common type of INSERT statement. It’s most useful when interacting with single-row inserts.
You typically use this type of INSERT statement when working with data entered through end-user web forms. In some cases, users can enter more than one row of data using a form, which occurs, for example, when a user places a meal order in a restaurant and the meal and drink are treated as order items. The restaurant order entry system would enter a single-row in the order table and two rows in the order_item table (one for the meal and the other for the drink). PL/SQL programmers usually handle the insertion of related rows typically inside a loop structure when they use dynamic INSERT statements. Dynamic inserts are typically performed using NDS (Native Dynamic SQL) statements.
Oracle supports only a single-row insert through the VALUES clause. Multiple-row inserts require an INSERT statement from a query.
The VALUES clause of an INSERT statement accepts scalar values, such as strings, numbers, and dates. It also accepts calls to arrays, lists, or user-defined object types, which are called flattened objects. Oracle supports VARRAY as arrays and nested tables as lists. They can both contain elements of a scalar data type or user-defined object type.
The following sections discuss how you use the VALUES clause with scalar data types, how you convert various data types, and how you use the VALUES clause with nested tables and user-defined object data types.
Inserting Scalar Data Types
Instruction Details →
This section shows you how to INSERT scalar values into tables.
The basic syntax for an INSERT statement with a VALUES clause can include an optional override signature between the table name and VALUES keyword. With an override signature, you designate the column names and the order of entry for the VALUES clause elements. Without an override signature, the INSERT signature checks the definition of the table in the database catalog. The positional order of the column in the data catalog defines the positional, or default, signature for the INSERT statement. As shown previously, you can discover the structure of a table in Oracle with the DESCRIBE command issued at the SQL*Plus command line:
DESCRIBE table_name |
You’ll see the following after describing the rental table in SQL*Plus:
Name Null? Type ------------------------------------ -------- -------- RENTAL_ID NOT NULL NUMBER CUSTOMER_ID NOT NULL NUMBER CHECK_OUT_DATE NOT NULL DATE RETURN_DATE DATE CREATED_BY NOT NULL NUMBER CREATION_DATE NOT NULL DATE LAST_UPDATED_BY NOT NULL NUMBER LAST_UPDATE_DATE NOT NULL DATE |
The rental_id column is a surrogate key, or an artificial numbering sequence. The combination of the customer_id and check_out_date columns serves as a natural key because a DATE data type is a date-time value. If it were only a date, the customer would be limited to a single entry for each day, and limiting customer rentals to one per day isn’t a good business model.
The basic INSERT statement would require that you look up the next sequence value before using it. You should also look up the surrogate key column value that maps to the row where your unique customer is stored in the contact table. For this example, assume the following facts:
- Next sequence value is 1086
- Customer’s surrogate key value is 1009
- Current date-time is represented by the value from the SYSDATE function
- Return date is the fifth date from today
- User adding and updating the row has a primary (surrogate) key value of 1
- Creation and last update date are the value returned from the SYSDATE function.
An INSERT statement must include a list of values that match the positional data types of the database catalog, or it must use an override signature for all mandatory columns.
You can now write the following INSERT statement, which relies on the default signature:
Name Null? Type ------------------------------------ -------- -------- RENTAL_ID NOT NULL NUMBER CUSTOMER_ID NOT NULL NUMBER CHECK_OUT_DATE NOT NULL DATE RETURN_DATE DATE CREATED_BY NOT NULL NUMBER CREATION_DATE NOT NULL DATE LAST_UPDATED_BY NOT NULL NUMBER LAST_UPDATE_DATE NOT NULL DATE |
The rental_id column is a surrogate key, or an artificial numbering sequence. The combination of the customer_id and check_out_date columns serves as a natural key because a DATE data type is a date-time value. If it were only a date, the customer would be limited to a single entry for each day, and limiting customer rentals to one per day isn’t a good business model.
The basic INSERT statement would require that you look up the next sequence value before using it. You should also look up the surrogate key column value that maps to the row where your unique customer is stored in the contact table. For this example, assume the following facts:
- Next sequence value is 1086
- Customer’s surrogate key value is 1009
- Current date-time is represented by the value from the SYSDATE function
- Return date is the fifth date from today
- User adding and updating the row has a primary (surrogate) key value of 1
- Creation and last update date are the value returned from the SYSDATE function.
An INSERT statement must include a list of values that match the positional data types of the database catalog, or it must use an override signature for all mandatory columns.
You can now write the following INSERT statement, which relies on the default signature:
SQL> INSERT INTO rental 2 VALUES 3 ( 1086 4 , 1009 5 , SYSDATE 6 , TRUNC(SYSDATE + 5) 7 ,1 8 , SYSDATE 9 , 1 10 , SYSDATE); |
If you weren’t using SYSDATE for the date-time value on line 5, you could manually enter a date-time with the following Oracle proprietary syntax:
5 , TO_DATE('15-APR-2011 12:53:01','DD-MON-YYYY HH24:MI:SS') |
The TO_DATE function is an Oracle-specific function. The generic conversion function would be the CAST function. The problem with a CAST function by itself is that it can’t handle a format mask other than the database defaults (‘DD-MON-RR‘ or ‘DD-MON-YYYY‘). For example, consider this syntax:
5 , CAST('15-APR-2011 12:53:02' AS DATE) |
It raises the following error:
5 , CAST('15-APR-2011 12:53:02' AS DATE) FROM dual * ERROR AT line 1: ORA-01830: DATE format picture ends before converting entire input string |
You actually need to double cast this type of format mask when you want to store it as a DATE data type. The working syntax casts the date-time string as a TIMESTAMP data type before recasting the TIMESTAMP to a DATE, like
5 , CAST(CAST('15-APR-2011 12:53:02' AS TIMESTAMP) AS DATE) |
Before you could write the preceding INSERT statement, you would need to run some queries to find the values. You would secure the next value from a rental_s1 sequence in an Oracle database with the following command:
SQL> SELECT rental_s1.NEXTVAL FROM dual; |
This assumes two things, because sequences are separate objects from tables. First, code from which the values in a table’s surrogate key column come must appear in the correct sequence. Second, a sequence value is inserted only once into a table as a primary key value.
In place of a query that finds the next sequence value, you would simply use a call against the .nextval pseudocolumn in the VALUES clause. You would replace line 3 with this:
3 ( rental_s1.NEXTVAL |
The .nextval is a pseudocolumn, and it instantiates an instance of a sequence in the current session. After a call to a sequence with the .nextval pseudocolumn, you can also call back the prior sequence value with the .currval pseudocolumn.
Assuming the following query would return a single-row, you can use the contact_id value as the customer_id value in the rental table:
SQL> SELECT contact_id 2 FROM contact 3 WHERE last_name = 'Potter' 4 AND first_name = 'Harry'; |
Taking three steps like this is unnecessary, however, because you can call the next sequence value and find the valid customer_id value inside the VALUES clause of the INSERT statement. The following INSERT statement uses an override signature and calls for the next sequence value on line 11. It also uses a scalar subquery to look up the correct customer_id value with a scalar subquery on lines 12 through 15.
SQL> INSERT INTO rental 2 ( rental_id 3 , customer_id 4 , check_out_date 5 , return_date 6 , created_by 7 , creation_date 8 , last_updated_by 9 , last_update_date ) 10 VALUES 11 ( rental_s1.NEXTVAL 12 ,(SELECT contact_id 13 FROM contact 14 WHERE last_name = 'Potter' 15 AND first_name = 'Harry') 16 , SYSDATE 17 , TRUNC(SYSDATE + 5) 18 , 1 19 , SYSDATE 20 , 3 21 , SYSDATE); |
When a subquery returns two or more rows because the conditions in the WHERE clause failed to find and return a unique row, the INSERT statement would fail with the following message:
,(SELECT contact_id * ERROR AT line 3: ORA-01427: single-ROW subquery returns more than one ROW |
In fact, the statement could fail when there are two or more “Harry Potter” names in the data set because three columns make up the natural key of the contact table. The third column is the member_id, and all three should be qualified inside a scalar subquery to guarantee that it returns only one row of data.
Handling Oracle’s Large Objects
Instruction Details →
This section shows you how to INSERT large object values into tables.
Oracle’s large objects present a small problem when they’re not null constrained in the table definition. You must insert empty object containers or references when you perform an INSERT statement.
Assume, for example, that you have the following three large object columns in a table:
Name Null? Type ------------------------------- -------- ----------------------- ITEM_DESC NOT NULL CLOB ITEM_ICON NOT NULL BLOB ITEM_PHOTO BINARY FILE LOB |
The item_desc column uses a CLOB (Character Large Object) data type, and it is a required column; it could hold a lengthy description of a movie, for example. The item_icon column uses a BLOB (Binary Large Object) data type, and it is also a required column. It could hold a graphic image. The item_photo column uses a binary file (an externally managed file) but is fortunately null allowed or an optional column in any INSERT statement. It can hold a null or a reference to an external graphic image.
Oracle provides two functions that let you enter an empty large object, and they are:
EMPTY_BLOB() EMPTY_CLOB() |
Although you could insert a null value in the item_photo column, you can also enter a reference to an Oracle database virtual directory file. Here’s the syntax to enter a valid BFILE name with the BFILENAME function call:
10 , BFILENAME('VIRTUAL_DIRECTORY_NAME', 'file_name.png') |
You can insert a large character or binary stream into BLOB and CLOB data types by using the stored procedures and functions available in the dbms_lob package. Chapter 13 covers the dbms_lob package.
You can use an empty_clob function or a string literal up to 32,767 bytes long in a VALUES clause. You must use the dbms_lob package when you insert a string that is longer than 32,767 bytes. That also changes the nature of the INSERT statement and requires that you append the RETURNING INTO clause. Here’s the prototype for this Oracle proprietary syntax:
INSERT INTO some_table [( column1, column2, column3, ...)] VALUES ( value1, value2, value3, ...) RETURNING column1 INTO local_variable; |
The local_variable is a reference to a procedural programming language. It lets you insert a character stream into a target CLOB column or insert a binary stream into a BLOB column.
Capturing the Last Sequence Value
Instruction Details →
This section shows you how to INSERT a new sequence in a parent table and a copy of that new sequence as a foreign key value in a child table.
Sometimes you insert into a series of tables in the scope of a transaction. In this scenario, one table gets the new sequence value (with a call to sequence_name.nextval) and enters it as the surrogate primary key, and another table needs a copy of that primary key to enter into a foreign key column. While scalar subqueries can solve this problem, Oracle provides the .currval pseudocolumn for this purpose.
The steps to demonstrate this behavior require a parent table and a child table. The parent table is defined as follows:
Name Null? Type ------------------------------------ -------- -------------- PARENT_ID NOT NULL NUMBER PARENT_NAME VARCHAR2(10) |
The parent_id column is the primary key for the parent table. You include the parent_id column in the child table. In the child table, the parent_id column holds a copy of a valid primary key column value as a foreign key to the parent table.
Name Null? Type ------------------------------------ -------- -------------- CHILD_ID NOT NULL NUMBER PARENT_ID NUMBER PARENT_NAME VARCHAR2(10) |
After creating the two tables, you can manage inserts into them with the .nextval and .currval pseudocolumns. The sequence calls with the .nextval pseudocolumn insert primary key values, and the sequence calls with the .currval pseudocolumn insert foreign key values.
You would perform these two INSERT statements as a group:
SQL> INSERT INTO parent 2 VALUES 3 ( parent_s1.NEXTVAL 4 ,'One Parent'); SQL> INSERT INTO child 2 VALUES 3 ( child_s1.NEXTVAL 4 , parent_s1.CURRVAL 5 ,'One Child'); |
The .currval pseudocolumn for any sequence fetches the value placed in memory by call to the .nextval pseudocolumn. Any attempt to call the .currval pseudocolumn before the .nextval pseudocolumn raises an ORA-02289 exception. The text message for that error says the sequence doesn’t exist, which actually means that it doesn’t exist in the scope of the current session. Line 4 in the insert into the child table depends on line 3 in the insert into the parent table.
You can use comments in INSERT statements to map to columns in the table. For example, the following shows the technique for the child table from the preceding example:
SQL> INSERT INTO child 2 VALUES 3 ( child_s1.NEXTVAL -- CHILD_ID 4 , parent_s1.CURRVAL -- PARENT_ID 5 ,'One Child') -- CHILD_NAME 6 / |
Comments on the lines of the VALUES clause identify the columns where the values are inserted. A semicolon doesn’t execute this statement, because a trailing comment would trigger a runtime exception. You must use the semicolon or forward slash on the line below the last VALUES element to include the last comment.
Insert by Subquery Results →
An INSERT statement with a subquery can insert one to many rows of data into any table provided the SELECT-list of the subquery matches the data dictionary definition of the table or the named-notation list provided by the INSERT statement. An INSERT statement with a subquery cannot have a VALUES keyword in it, or it raises an error.
The generic prototype for an INSERT statement follows the pattern of an INSERT statement by value prototype with one exception, it excludes the VALUES keyword and replaces the common delimited list of values with a SELECT-list from a subquery. If you want to rely on the positional definition of the table, exclude the list of comma delimited column values. The optional comma-delimited list of column values is necessary when you want to insert columns in a different order or exclude optional columns.
The generic prototype is:
INSERT INTO table_name [( column1, column2, column3, ...)] ( SELECT value1, value2, value3, ... FROM table_name WHERE ...); |
The subquery, or SELECT statement, must return a SELECT-list that maps to the column definition in the data dictionary or the optional comma-delimited column list.
PostgreSQL Arrays
If you’re wondering about this post, it shows the basic array of a set of integers and strings before showing you how to create nested tables of data in PostgreSQL. By the way, they’re not called nested tables in PostgreSQL, like they are in Oracle but perform like their Oracle cousins.
Let’s create a table with an auto-incrementing column and two arrays, one array of integers and another of strings:
-- Conditionally drop the demo table. DROP TABLE IF EXISTS demo; -- Create the test table. CREATE TABLE demo ( demo_id serial , demo_number integer[5] , demo_string varchar(5)[7]); |
You can insert test values like this:
INSERT INTO demo (demo_number, demo_string) VALUES ( array[1,2,3,4,5] , array['One','Two','Three','Four','Five','Six','Seven']); |
Then, you can query them with this unnest function, like:
SELECT unnest(demo_number) AS numbers , unnest(demo_string) AS strings FROM demo; |
It returns:
numbers | strings ---------+--------- 1 | One 2 | Two 3 | Three 4 | Four 5 | Five | Six | Seven (7 rows) |
You may note that the two arrays are asymmetrical. It only becomes an issue when you navigate the result in a PL/pgSQL cursor or imperative programming language, like Python.
Now, let’s do something more interesting like work with a composite user-defined type, like the player structure. You would create the composite user-defined type with this syntax:
-- Conditionally drop the player type. DROP TYPE IF EXISTS player; -- Create the player type. CREATE TYPE player AS ( player_no integer , player_name varchar(24) , player_position varchar(14) , ab integer , r integer , h integer , bb integer , rbi integer ); |
You can create a world_series table that include a players column that uses an array of player type, like
-- Conditionally drop the world_series table. DROP TABLE IF EXISTS world_series; -- Create the player type. CREATE TABLE world_series ( world_series_id serial , team varchar(24) , players player[30] , game_no integer , year integer ); |
If you’re familiar with the Oracle Database, you’d have to specify a tested table in the syntax. Fortunately, PostgreSQL doesn’t require that.
Insert two rows with the following statement:
INSERT INTO world_series ( team , players , game_no , year ) VALUES ('San Francisco Giants' , array[(24,'Willie Mayes','Center Fielder',5,0,1,0,0)::player ,(5,'Tom Haller','Catcher',4,1,2,0,2)::player] , 4 , 1962 ); |
You can append to the array with the following syntax. A former student and I have a disagreement on whether this is shown in the PostgreSQL 8.15.4 Modifying Array documentation. I believe it can be inferred from the document and he doesn’t believe so. Anyway, here’s how you add an element to an existing array in a table with the UPDATE statement:
UPDATE world_series SET players = (SELECT array_append(players,(7,'Henry Kuenn','Right Fielder',3,0,0,1,0)::player) FROM world_series) WHERE team = 'San Francisco Giants' AND year = 1962 AND game_no = 4; |
Like Oracle’s nested tables, PostgreSQL’s arrays of composite user-defined types requires writing a PL/pgSQL function. I’ll try to add one of those shortly in another blog entry to show you how to edit and replace entries in stored arrays of composite user-defined types.
You can query the unnested rows and get a return set like a Python tuple with the following query:
SELECT unnest(players) AS player_list FROM world_series WHERE team = 'San Francisco Giants' AND year = 1962 AND game_no = 4; |
It returns the three rows from the players array:
player_list ---------------------------------------------- (24,"Willie Mayes","Center Field",5,0,1,0,0) (5,"Tom Haller",Catcher,4,1,2,0,2) (7,"Henry Kuenn","Right Fielde",3,0,0,1,0) (3 rows) |
It returns the data set in entry-order. If we step outside of the standard 8.15 Arrays PostgreSQL Documentation, you can do much more with arrays (or nested tables). The balance of this example demonstrates some new syntax that helps you achieve constructive outcomes in PostgreSQL.
You can use a Common Table Expression (CTE) to get the columnar display of the player composite user-defined type. This type of solution is beyond the standard , like:
WITH list AS (SELECT unnest(players) AS row_result FROM world_series WHERE team = 'San Francisco Giants' AND year = 1962 AND game_no = 4) SELECT (row_result).player_name , (row_result).player_no , (row_result).player_position FROM list; |
If you’re unfamiliar with accessing composite user-defined types, I wrote a post on that 7 years ago. You can find the older blog entry PostgreSQL Composites on my blog.
It returns only the three requested columns of the player composite user-defined type:
player_name | player_no | player_position --------------+-----------+----------------- Willie Mayes | 24 | Center Fielder Tom Haller | 5 | Catcher Henry Kuenn | 7 | Right Fielder (3 rows) |
You should note that the data is presented in an entry-ordered manner when unnested alone in the SELECT-list. That behavior changes when the SELECT-list includes non-array data.
The easiest way to display data from the non-array and array columns is to list them inside the SELECT-list of the CTE, like:
WITH list AS (SELECT game_no AS game , year , unnest(players) AS row_result FROM world_series WHERE team = 'San Francisco Giants' AND year = 1962 AND game_no = 4) SELECT game , year , (row_result).player_name , (row_result).player_no , (row_result).player_position FROM list; |
It returns an ordered set of unnested rows when you include non-array columns, like:
game | year | player_name | player_no | player_position ------+------+--------------+-----------+----------------- 4 | 1962 | Henry Kuenn | 7 | Right Fielder 4 | 1962 | Tom Haller | 5 | Catcher 4 | 1962 | Willie Mayes | 24 | Center Fielder (3 rows) |
While you can join the world_series table to the unnested array rows (returned as a derived table, its a bad idea. The mechanics to do it require you to return the primary key column in the same SELECT-list of the CTE. Then, you join the CTE list to the world_series table by using the world_series_id primary key.
However, there is no advantage to an inner join approach and it imposes unnecessary processing on the database server. The odd rationale that I heard when I noticed somebody was using a CTE to base-table join was: “That’s necessary so they could use column aliases for the non-array columns.” That’s not true because you can use the aliases inside the CTE, as shown above when game is an alias to the game_no column.
As always, I hope this helps those looking to solve a problem in PostgreSQL.
Oracle Container User
After you create and provision the Oracle Database 21c Express Edition (XE), you can create a c##student container user with the following two step process.
- Create a c##student Oracle user account with the following command:
CREATE USER c##student IDENTIFIED BY student DEFAULT TABLESPACE users QUOTA 200M ON users TEMPORARY TABLESPACE temp;
- Grant necessary privileges to the newly created c##student user:
GRANT CREATE CLUSTER, CREATE INDEXTYPE, CREATE OPERATOR , CREATE PROCEDURE, CREATE SEQUENCE, CREATE SESSION , CREATE TABLE, CREATE TRIGGER, CREATE TYPE , CREATE VIEW TO c##student;
As always, I hope this helps those looking for how to do something that’s less than clear because everybody uses tools.
Tiny SQL Developer
The first time you launch SQL Developer, you may see a very small or tiny display on the screen. With some high resolution screens the text is unreadable. Unless you manually configure the sqldeveloper shortcut, you generally can’t use it.
On my virtualization on a 27″ screen it looks like:
As an Administrator user, you right click the SQLDeveloper icon and click the Compatibility tab, which should look like the following dialog. You need to check the Compatibility Mode, which by default is unchecked with Windows 8 displayed in the select list.
Check the Compatibility Mode box and the select list will no longer be gray scaled. Click on the select list box and choose Windows 7. After the change you should see the following:
After that change, you need to click on the Change high DPI settings gray scaled button, which will display the following dialog box.
Click the Override high DPI scaling behavior check box. It will change the gray highlighted Scaling Performed by select box to white. Then, you click the Scaling Performed by select box and choose the System option.
Click the OK button on the nested SQLDeveloper Properties dialog box. Then, click the Apply button on the SQLDeveloper Properties button and the OK button. You will see a workable SQL Developer interface when you launch the program through your modified shortcut.
Protocol adapter error
One of the errors that defeats a lot of new users who install the Oracle Database on the Windows operating system is a two-step event. The first step occurs when you try to connect to the database and it raises the following error:
SQL*Plus: Release 18.0.0.0.0 - Production on Fri Jan 7 21:00:42 2022 Version 18.4.0.0.0 Copyright (c) 1982, 2018, Oracle. All rights reserved. ERROR: ORA-12541: TNS:no listener |
The second step may occur after you get the “no listener” error when you try to start the Oracle listener and it fails to start. The Oracle listener control command is:
lsnrctl start |
When it returns the following error:
LSNRCTL FOR 64-bit Windows: Version 18.0.0.0.0 - Production ON 07-JAN-2022 21:02:20 Copyright (c) 1991, 2018, Oracle. ALL rights reserved. Starting tnslsnr: please wait... Unable TO OpenSCManager: err=5 TNS-12560: TNS:protocol adapter error TNS-00530: Protocol adapter error |
The problem is generally in two configuration files. They are the listener.ora and tnsnames.ora files. This typically occurs when the developer fails to set the localhost in the Windows operating system hosts configuration file. The chain of events that causes these errors can be avoided when the user puts the following two lines:
127.0.0.1 localhost ::1 localhost |
in the following hosts file:
C:\Windows\system32\drivers\etc\hosts |
You can typically avoid these errors when you configure the hosts configuration file correctly before installing the Oracle Database. That’s because the Oracle database installation will use localhost keyword instead of the current, and typically DHCP assigned, IP address.
The loss of connectivity errors typically occur when the IP address changes after the installation. DHCP IP addresses often change as machines disconnect and reconnect to a network.
You can fix a DHCP IP installation of an Oracle database by editing the listener.ora and tnsnames.ora files. You replace the IP addresses with the localhost keyword.
The listener.ora and tnsnames.ora files look like the following for an Oracle Database 21c Express Edition (provided you installed them in a C:\app\username directory:
listener.ora
# listener.ora Network Configuration File: C:\app\username\product\21.0.0\dbhomeXE\NETWORK\ADMIN\listener.ora # Generated by Oracle configuration tools. DEFAULT_SERVICE_LISTENER = XE SID_LIST_LISTENER = (SID_LIST = (SID_DESC = (SID_NAME = CLRExtProc) (ORACLE_HOME = C:\app\username\product\21.0.0\dbhomeXE) (PROGRAM = extproc) (ENVS = "EXTPROC_DLLS=ONLY:C:\app\username\product\21.0.0\dbhomeXE\bin\oraclr21.dll") ) ) LISTENER = (DESCRIPTION_LIST = (DESCRIPTION = (ADDRESS = (PROTOCOL = TCP)(HOST = localhost)(PORT = 1521)) (ADDRESS = (PROTOCOL = IPC)(KEY = EXTPROC1521)) ) ) |
tnsnames.ora
# tnsnames.ora Network Configuration File: C:\app\mclaughlinm\product\21.0.0\dbhomeXE\NETWORK\ADMIN\tnsnames.ora # Generated by Oracle configuration tools. XE = (DESCRIPTION = (ADDRESS = (PROTOCOL = TCP)(HOST = localhost)(PORT = 1521)) (CONNECT_DATA = (SERVER = DEDICATED) (SERVICE_NAME = XE) ) ) LISTENER_XE = (ADDRESS = (PROTOCOL = TCP)(HOST = localhost)(PORT = 1521)) ORACLR_CONNECTION_DATA = (DESCRIPTION = (ADDRESS_LIST = (ADDRESS = (PROTOCOL = IPC)(KEY = EXTPROC1521)) ) (CONNECT_DATA = (SID = CLRExtProc) (PRESENTATION = RO) ) ) |
As always, I hope this helps those looking for a solution to something that can take more time than it should to fix.