Oracle Migration Workbench Reference Guide for Microsoft SQL Server and Sybase Adaptive Server Migrations Release 9.2.0 for Microsoft Windows 98/2000 and Microsoft Windows NT Part Number A97248-01 |
|
This chapter includes the following sections:
Microsoft SQL Server and Sybase Adaptive Server store triggers and stored procedures with the server. Oracle stores triggers and stored subprograms with the server. Oracle has three different kinds of stored subprograms, namely functions, stored procedures, and packages. For detailed discussion on all these objects, see the PL/SQL User's Guide and Reference, Release 1 (9.0.1).
The following topics are discussed in this section:
Microsoft SQL Server and Sybase Adaptive Server database triggers are AFTER triggers. This means that triggers are fired after the specific operation is performed. For example, the INSERT trigger fires after the rows are inserted into the database. If the trigger fails, the operation is rolled back.
Microsoft SQL Server and Sybase Adaptive Server allow INSERT, UPDATE, and DELETE triggers. Triggers typically need access to the before image and after image of the data that is being changed. Microsoft SQL Server and Sybase Adaptive Server achieve this with two temporary tables called INSERTED and DELETED. These two tables exist during the execution of the trigger. These tables and the table for which the trigger is written have the exact same structure. The DELETED table holds the before image of the rows that are undergoing change because of the INSERT/UPDATE/DELETE operation, and the INSERTED table holds the after image of these rows. If there is an error, the triggers can issue a rollback statement.
Most of the Microsoft SQL Server and Sybase Adaptive Server trigger code is written to enforce referential integrity. Microsoft SQL Server and Sybase Adaptive Server triggers are executed once per triggering SQL statement (such as INSERT, UPDATE, or DELETE). If you want some actions to be performed for each row that the SQL statement affects, you must code the actions using the INSERTED and DELETED tables.
Oracle has a rich set of triggers. Oracle also provides triggers that fire for events such as INSERT, UPDATE, and DELETE. You can also specify the number of times that the trigger action is to be executed. For example, once for every row affected by the triggering event (such as might be fired by an UPDATE statement that updates many rows), or once for the triggering statement (regardless of how many rows it affects).
A ROW trigger is fired each time that the table is affected by the triggering event. For example, if an UPDATE statement updates multiple rows of a table, a row trigger is fired once for each row affected by the UPDATE statement. A STATEMENT trigger is fired once on behalf of the triggering statement, regardless of the number of rows in the table that the triggering statement affects.
Oracle triggers can be defined as either BEFORE triggers or AFTER triggers. BEFORE triggers are used when the trigger action should determine whether the triggering statement should be allowed to complete. By using a BEFORE trigger, you can avoid unnecessary processing of the triggering statement and its eventual rollback in cases where an exception is raised.
As combinations, there are four different types of triggers in Oracle:
It is sometimes necessary to create a ROW trigger or a STATEMENT trigger to achieve the same functionality as the Microsoft SQL Server and Sybase Adaptive Server trigger. This occurs in the following cases:
In the following example, the group function AVG is used to calculate the average salary:
SELECT AVG(inserted.salary) FROM inserted a, deleted b WHERE a.id = b.id;
This would be converted to Oracle by creating an AFTER ROW trigger to insert all the updated values into a package, and an AFTER STATEMENT trigger to read from the package and calculate the average.
For examples of Oracle triggers, see the Oracle9i Application Developer's Guide - Fundamentals, Release 1 (9.0.1).
Stored procedures provide a powerful way to code the application logic that can be stored with the server. Microsoft SQL Server and Sybase Adaptive Server and Oracle all provide stored procedures.
The language used to code these objects is a database-specific procedural extension to SQL. In Oracle it is PL/SQL and in Microsoft SQL Server and Sybase Adaptive Server it is Transact SQL (T/SQL). These languages differ to a considerable extent. The individual SQL statements and the procedural constructs, such as if-then-else, are similar in both versions of the procedural SQL. Considerable differences can be found in the following areas discussed in this section:
This section also considers various components of typical Microsoft SQL Server and Sybase Adaptive Server stored procedures and suggests ways to design them in order to avoid conversion problems. By applying the standards described below to the coding, you can convert your stored procedures from Microsoft SQL Server and Sybase Adaptive Server to Oracle.
Different relational database management systems (RDBMSs) use different methods to send data to clients. For example, in Microsoft SQL Server and Sybase Adaptive Server the server sends data to the client in the form of a byte-stream. The client is responsible for retrieving all the data from the communication channel before sending another request to the server. In Oracle, the client can issue one or more SQL statements on the same network connection, and the system global area (SGA) stores all the data retrieved from the database. The server sends the data to the client as requested and the client sends a FETCH request on the connection whenever it is ready for the next set of results. This section discusses the different methods used to send data to clients under the following headings:
Microsoft SQL Server and Sybase Adaptive Server and Oracle can all send data to clients by means of output variables.
Many Microsoft SQL Server and Sybase Adaptive Server applications rely on the SQL Server-specific stream-based data return method called "result sets". Oracle is optimized to return data more efficiently when the data is requested using an ANSI-standard SQL SELECT statement, as compared to any proprietary stored procedure method. Therefore, the best design decision is to use stored procedures for data processing and SELECT statements for queries.
In Oracle, the use of cursor variables allows client programs to retrieve well-structured result sets.
To send even a single row back to the client from the stored procedure, Microsoft SQL Server and Sybase Adaptive Server can use result sets instead of an ANSI-standard method.
For example:
CREATE PROCEDURE get_emp_rec @empid INT AS SELECT fname, lname, loginid, addr, title, dept, mgrid FROM employee WHERE empid = @empid
The above procedure can be converted to an Oracle PL/SQL procedure as follows:
CREATE OR REPLACE PROCEDURE get_emp_rec (empid IN NUMBER, fname OUT VARCHAR2, lname OUT VARCHAR2, loginid OUT VARCHAR2, addr OUT VARCHAR2, title OUT VARCHAR2, dept OUT NUMBER, mgrid OUT NUMBER) AS BEGIN SELECT fname, lname, loginid, addr, title, dept, mgrid INTO fname, lname, loginid, addr, title, dept, mgrid FROM employee WHERE empid = empid; END;
Output variables are a structured way of sending data from server to client. Output variables allow the caller to see the results in a predictable manner, as the structure of the output variable is predefined. This method also allows encapsulation of behavior of the stored procedures.
Output variables offer the following benefits:
If a third-party user interface product uses the result set capability of Microsoft SQL Server and Sybase Adaptive Server, make sure that the same functionality can be made available to the Oracle database. For example, PowerBuilder can use result sets to populate the data windows.
Although many client programs, such as Oracle Call Interface (OCI), precompilers, SQL*Module, and SQL*Plus, recognize cursor variables, most Open Database Connectivity (ODBC) drivers cannot recognize cursor variables. One solution when using ODBC drivers is to identify the code that produces the result set, and move this code online in the client program. The Oracle9i and Oracle8i ODBC Driver release 8.1.5.4.0 and later releases support result sets.
In the following example, an Microsoft SQL Server and Sybase Adaptive Server stored procedure returns a result set with multiple rows:
CREATE PROCEDURE make_loginid BEGIN update employee set loginid = substring(fname,1,1) + convert(varchar(7),empid) select fname, lname, loginid from employee END
This procedure sends all the qualifying rows to the client as a continuous data stream. To further process the rows, the client program must retrieve the rows one after another from the communication channel.
The following piece of the DB-Library/C code executes the above procedure and prints each row to the screen.
main() { /* Data structure dbproc is conceptually very similar to CDA data structure used in Oracle's OCI/C programs */ dbcmd(dbproc, "exec make_loginid"); /* The above command sets the command buffer with the transact-sql command that needs to be executed. */ dbsqlexec(dbproc); /* This command causes the parsing and execution of the SQL command on the server side. */ dbresults(dbproc); /* This command puts the result rows onto the communications channel. */ /*The following while loop retrieves the result rows one after the other by calling the function dbnextrow repeatedly. This implementation is cursor implementation through DB-Library functions. */ while (dbnextrow(dbproc) != NO_MORE_ROWS) { dbprrow(dbproc); /* This function prints the retrieved row to the standard output. */ }
You can migrate Microsoft SQL Server and Sybase Adaptive Server stored procedures to the Oracle PL/SQL stored procedures or packages in different ways, as follows:
Examples of these different Oracle solutions to the result set problem are presented below:
CREATE OR REPLACE PROCEDURE make_loginid AS BEGIN update employee set loginid = substr(lname,1,1) || substr(to_char(empid),1,7); END;
The following SELECT statement becomes part of the client code:
select fname, lname, loginid from employee
The following PL/SQL code shows how to migrate the make_loginid procedure to Oracle by using PL/SQL tables as output parameters:
CREATE OR REPLACE PACKAGE make_loginid_pkg IS BEGIN DECLARE EmpFnameTabType IS TABLE OF employee.fname %TYPE INDEX BY BINARY_INTEGER; DECLARE EmpLnameTabType IS TABLE OF employee.lname %TYPE INDEX BY BINARY_INTEGER; DECLARE EmpLoginidTabType IS TABLE OF employee.loginid %TYPE INDEX BY BINARY_INTEGER; emp_fname_tab EmpFnameTabType; emp_lname_tab EmpLnameTabType; emp_loginid_tab EmpLoginidTabType; PROCEDURE make_loginid (emp_fname_tab OUT EmpFnameTabType, emp_lname_tab OUT EmpLnameTabType, emp_loginid_tab OUT EmpLoginidTabType); END make_loginid_pkg;
The package body definition is as follows:
CREATE OR REPLACE PACKAGE BODY make_loginid_pkg IS BEGIN PROCEDURE make_loginid (emp_fname_tab OUT EmpFnameTabType, emp_lname_tab OUT EmpLnameTabType, emp_loginid_tab OUT EmpLoginidTabType) AS DECLARE i BINARY_INTEGER := 0; BEGIN update employee set loginid = substr(fname,1,1) || substr(to_char(empid),1,7); FOR emprec IN (select fname,lname,loginid from employee) LOOP i := i + 1; emp_fname_tab[i] = emprec.fname; emp_lname_tab[i] = emprec.lname; emp_loginid_tab[i] = emprec.loginid; END LOOP; END make_loginid; END make_loginid_pkg;
This procedure updates the PL/SQL tables with the data. This data is then available to the client after the execution of this packaged procedure.
The package definition is as follows:
CREATE OR REPLACE PACKAGE make_loginid_pkg IS BEGIN PROCEDURE update_loginid; PROCEDURE fetch_emprec done_flag IN OUT INTEGER, nrows IN OUT INTEGER, fname OUT VARCHAR2, lname OUT VARCHAR2, loginid OUT VARCHAR2); END make_loginid_pkg;
The package body definition is as follows:
CREATE OR REPLACE PACKAGE BODY make_loginid_pkg IS BEGIN CURSOR emprec IS select fname, lname, loginid from employee; PROCEDURE update_loginid IS BEGIN update employee set loginid = substr(fname,1,1) || substr(to_char(loginid),1,7); END update_loginid; PROCEDURE fetch_emprec done_flag IN OUT INTEGER, nrows IN OUT INTEGER, fname OUT VARCHAR2, lname OUT VARCHAR2, loginid OUT VARCHAR2) IS BEGIN IF NOT emprec%ISOPEN THEN OPEN emprec; nrows := 0; END IF; done_flag := 0; FETCH emprec INTO fname, lname, loginid; IF emprec%NOTFOUND THEN CLOSE emprec; done_flag := 1; ELSE nrows := nrows + 1; ENDIF; END fetch_emprec; END make_loginid_pkg;
Oracle allows you to define a cursor variable to return query results. This cursor variable is similar to the user-defined record type and array type. The cursor stored in the cursor variable is like any other cursor. It is a reference to a work area associated with a multi-row query. It denotes both the set of rows and a current row in that set. The cursor referred to in the cursor variable can be opened, fetched from, and closed just like any other cursor.
There is a difference; since it is a PL/SQL variable, it can be passed into and out of procedures like any other PL/SQL variable. As a result, procedures that use cursor variables are reusable. You can see what the output of the procedure is by looking at the procedure definition. You can use the same procedure to return the results of a SELECT statement to a calling client program. Cursor variables can even be the return value of a function. The cursor variables preserve well-structured programming concepts while allowing the client routine to retrieve result sets.
Typically, the cursor would be declared in a client program (for example, OCI, precompilers, SQL*Module, or SQL*Plus) and then passed as an IN OUT parameter to the PL/SQL procedure. The procedure then opens the cursor based on a SELECT statement. The calling program performs the FETCHs from the cursor, including the possibility of using ARRAY FETCH to retrieve multiple rows in one network message, and closes the cursor when it is done.
... struct emp_record { char ename[11]; float sal; }emp_record; SQL_CURSOR c; EXEC SQL EXECUTE BEGIN emp_package.open_emp(:c,1); END; END-EXEC; ... /* fetch loop until done */ EXEC SQL FETCH :c INTO :emp_record; ... CLOSE :c; ...
CREATE OR REPLACE PACKAGE emp_package IS TYPE emp_part_rec IS RECORD (ename emp.ename%type, sal emp.sal%type); TYPE emp_cursor IS REF CURSOR RETURN emp_part_rec; PROCEDURE open_emp (c_emp IN OUT emp_cursor, select_type IN NUMBER); END emp_package; CREATE OR REPLACE PACKAGE BODY emp_package IS PROCEDURE open_emp (c_emp IN OUT emp_cursor, select_type IN NUMBER) IS BEGIN IF select_type=1 THEN OPEN c_emp FOR SELECT ename, sal FROM EMP WHERE COMM IS NOT NULL; ELSE OPEN c_emp FOR SELECT ename, sal FROM EMP; END IF; END open_emp; END emp_package;
Microsoft SQL Server and Sybase Adaptive Server stored procedures can return multiple different result sets to the calling routine.
For example, consider the following procedure:
CREATE PROCEDURE example_proc AS BEGIN SELECT empno, empname, empaddr FROM emp WHERE empno BETWEEN 1000 and 2000 SELECT empno, deptno, deptname FROM emp, dept WHERE emp.empno = dept.empno AND emp.empno BETWEEN 1000 and 2000 END
This procedure returns two different result sets. The client is responsible for processing the results. To convert Microsoft SQL Server and Sybase Adaptive Server multiple result sets to Oracle, pass one more cursor variable to the stored procedure to open a second cursor; the client program then looks at both cursor variables for data. However, it can be difficult to track all the result sets in a single procedure. It is recommended that you just use one result set, that is, one cursor variable per procedure, if possible.
Cursors allow row-by-row operations on a given result set. Microsoft SQL Server and Sybase Adaptive Server provide ANSI-standard SQL syntax to handle cursors. The additional DECLARE CURSOR, OPEN, FETCH, CLOSE, and DEALLOCATE CURSOR clauses are included in T/SQL. Using these statements you can achieve cursor manipulation in a stored procedure. After FETCHing the individual row of a result set, this current row can be modified with extensions provided with UPDATE and DELETE statements.
The UPDATE statement syntax is as follows:
update <table_name> set <column_name> = <expression> from <table1>, <table_name> where current of <cursor name> The DELETE statement syntax is as follows: delete from <table_name> where current of <cursor name> Microsoft SQL Server and Sybase Adaptive Server cursors map one-to-one with Oracle cursors.
In individual SQL statements, you should try to follow ANSI-standard SQL whenever possible. However, there are cases where you need to use database-specific SQL constructs, mostly for ease of use, simplicity of coding, and performance enhancement. For example, Microsoft SQL Server and Sybase Adaptive Server constructs such as the following are SQL Server-specific, and cannot be converted to Oracle without manual intervention:
update <table_name> set ... from <table1>, <table_name> where...
The manual intervention required to convert statements such as this can be seen in the following examples:
DELETE sales FROM sales, titles WHERE sales.title_id = titles.title_id AND titles.type = 'business'
DELETE FROM sales WHERE title_id IN (SELECT title_id FROM titles WHERE type = 'business' )
UPDATE titles SET price = price + author_royalty FROM titles, title_author WHERE titles.title.id = title_author.title_id
UPDATE titles O SET price = ( SELECT (O.price + I.author_royalty) FROM title_author I WHERE I.title_id = O.title_id) WHERE EXISTS (SELECT 1 FROM title_author WHERE title_author.title_id = O.title_id) ;
All the ANSI-standard SQL statements can be converted from one database to another using automatic conversion utilities.
In Microsoft SQL Server and Sybase Adaptive Server, transactions are explicit by definition. This implies that an individual SQL statement is not part of a logical transaction by default. A SQL statement belongs to a logical transaction if the transaction explicitly initiated by the user with a BEGIN TRANSACTION (or BEGIN TRAN) statement is still in effect. The logical transaction ends with a corresponding COMMIT TRANSACTION (or COMMIT TRAN) or ROLLBACK TRANSACTION (or ROLLBACK TRAN) statement. Each SQL statement that is not part of a logical transaction is committed on completion.
In Oracle, transactions are implicit as set by the ANSI standard. The implicit transaction model requires that each SQL statement is part of a logical transaction. A new logical transaction is automatically initiated when a COMMIT or ROLLBACK command is executed. This also implies that data changes from an individual SQL statement are not committed to the database after execution. The changes are committed to the database only when a COMMIT statement is run. The differences in the transaction models impact the coding of application procedures.
For client/server applications, it is recommended that you make the transaction-handling constructs part of the client procedures. The logical transaction is always defined by client users, and they should control it. This strategy is also more suitable for distributed transactions, where the two-phase commit operations are necessary. Making the transaction-handling statements a part of the client code serves a two-fold purpose; the server code is more portable, and the distributed transactions can be independent of the server code. Try to avoid using the BEGIN TRAN, ROLLBACK TRAN, and COMMIT TRAN statements in the stored procedures. In Microsoft SQL Server and Sybase Adaptive Server, transactions are explicit. In Oracle, transactions are implicit. If the transactions are handled by the client, the application code residing on the server can be independent of the transaction model.
Oracle PL/SQL checks each SQL statement for errors before proceeding with the next statement. If an error occurs, control immediately jumps to an exception handler. This avoids you having to check the status of every SQL statement. For example, if a SELECT statement does not find any rows in the database, an exception is raised, and the code to deal with this error is executed.
In Microsoft SQL Server and Sybase Adaptive Server, you need not check for errors after each SQL statement. Control is passed to the next statement, irrespective of the error conditions generated by the previous statement. It is your responsibility to check for errors after the execution of each SQL statement. Failure to do so may result in erroneous results.
In Oracle, to simulate the behavior of Microsoft SQL Server and Sybase Adaptive Server and to pass the control to the next statement regardless of the status of execution of the previous SQL statement, you must enclose each SQL statement in an equivalent PL/SQL block. This block must deal with all possible exceptions for that SQL statement. This coding style is required only to simulate Microsoft SQL Server and Sybase Adaptive Server behavior. An Oracle PL/SQL procedure ideally has only one exception block, and all error conditions are handled in that block.
Consider the following code in an Microsoft SQL Server and Sybase Adaptive Server stored procedure:
begin select @x = col1 from table1 where col2 = @y select @z = col3 from table2 where col4 = @x end
In this code example, if the first SELECT statement does not return any rows, the value of @x could be UNDEFINED. If the control is passed on to the next statement without raising an exception, the second statement returns incorrect results because it requires the value of @x to be set by an earlier statement. In a similar situation, Oracle PL/SQL raises a NO_DATA_FOUND exception if the first statement fails.
The Microsoft SQL Server and Sybase Adaptive Server RAISERROR statement does not return to the calling routine. The error code and message is passed to the client, and the execution of the stored procedure continues further. The Oracle RAISE_APPLICATION_ERROR statement returns to the calling routine. As a standard, a RETURN statement must appear after the RAISERROR statement in Microsoft SQL Server and Sybase Adaptive Server, so that it can be converted to the Oracle RAISE_APPLICATION_ERROR statement.
Microsoft SQL Server and Sybase Adaptive Server allow you to customize the error messages using a system table. The system procedures allow the user to add error messages to the system. Adding error messages to the Microsoft SQL Server and Sybase Adaptive Server system table is not desirable because there is no equivalent on the Oracle system. This can be avoided by maintaining a user-defined error messages table, located in the centralized database. Standard routines can be written to add the error message to the table and retrieve it whenever necessary. This method serves a two-fold purpose: it ensures that the system is more portable across different types of database servers, and it gives the administrator centralized control over the error messages.
This section provides information about data types under the following headings:
T/SQL local variables can be any server data type except TEXT and IMAGE. PL/SQL local variables can be any server data type including the following:
PL/SQL local variables can also be either of the following composite data types allowed by PL/SQL:
See the Data Types section in Chapter 2 for a list of Microsoft SQL Server and Sybase Adaptive Server data types and their equivalent Oracle data types.
Microsoft SQL Server and Sybase Adaptive Server do not have composite data types
This section compares the following Microsoft SQL Server and Sybase Adaptive Server and Oracle schema objects:
Each schema object is compared in separate tables based on create, drop, execute and alter, where applicable. The tables are divided into the following four sections
Some tables are followed by a recommendations section that contains important information about conversion implications.
This section provides the following tables for the schema object Procedure :
Recommendations:
Functionally identical parts can be identified in the T/SQL procedure and PL/SQL procedure structure. Therefore, you can automate the conversion of most of the constructs from Microsoft SQL Server and Sybase Adaptive Server to Oracle.
OR REPLACE keywords in an Oracle CREATE PROCEDURE statement provide an elegant way of recreating the procedure. In Microsoft SQL Server and Sybase Adaptive Server, the procedure must be dropped explicitly before replacing it.
Recommendations:
The above statement does not have any effect on the conversion process. This information is provided for reference only.
This section provides the following tables for the schema object Function:
This section provides the following tables for the schema object Package:
This section provides the following tables for the schema object Package Body:
This section provides information about the Microsoft SQL Server and Sybase Adaptive Server constructs and equivalent Oracle constructs generated by the Migration Workbench. The conversions of the following constructs are discussed in detail:
Listed is the syntax for the Microsoft SQL Server and Sybase Adaptive Server constructs and their Oracle equivalents, as well as comments about conversion considerations.
The procedures in the Oracle column are the direct output of the Migration Workbench. These PL/SQL procedures have more lines of code compared to the source Microsoft SQL Server and Sybase Adaptive Server procedures because these PL/SQL procedures are converted to emulate Microsoft SQL Server and Sybase Adaptive Server functionality. The PL/SQL procedures written from scratch for the same functionality in Oracle would be much more compact. The PL/SQL procedures generated by the Migration Workbench indicate the manual conversion required by adding appropriate commands. In general, the Migration Workbench deals with the Microsoft SQL Server and Sybase Adaptive Server T/SQL constructs in one of the following ways:
Comments
An Microsoft SQL Server and Sybase Adaptive Server stored procedure can be converted to a stored procedure, a function, or a package in Oracle. The output depends upon the option used when running the Migration Workbench.
The Migration Workbench automatically adds what is necessary to simulate Microsoft SQL Server and Sybase Adaptive Server functionality. In the example in Table 3-16 above, the Migration Workbench added the following three variables:
StoO_selcnt INTEGER; StoO_error INTEGER; StoO_rowcnt INTEGER;
These variables are needed in the EXCEPTION clause in the PL/SQL procedures that must be added for each SQL statement to emulate Microsoft SQL Server and Sybase Adaptive Server functionality. See the SELECT Statement topic in this section for clarification of the purpose of these variables.
Comments
Parameter passing is almost the same in Microsoft SQL Server and Sybase Adaptive Server and Oracle. By default, all the parameters are INPUT parameters, if not specified otherwise.
The value of the INPUT parameter cannot be changed from within the PL/SQL procedure. Thus, an INPUT parameter cannot be assigned any values nor can it be passed to another procedure as an OUT parameter. In Oracle, only IN parameters can be assigned a default value.
The @ sign in a parameter name declaration is removed in Oracle.
In Oracle, the parameter data type definition does not include length/size.
Microsoft SQL Server and Sybase Adaptive Server data types are converted to Oracle base data types. For example, all Microsoft SQL Server and Sybase Adaptive Server numeric data types are converted to NUMBER and all alphanumeric data types are converted to VARCHAR2 and CHAR in Oracle.
Comments
Microsoft SQL Server and Sybase Adaptive Server and Oracle follow similar rules for declaring local variables.
The Migration Workbench overrides the scope rule for variable declarations. As a result, all the local variables are defined at the top of the procedure body in Oracle.
Comments
IF statements in Microsoft SQL Server and Sybase Adaptive Server and Oracle are nearly the same except in the following two cases:
If EXISTS(...) in Microsoft SQL Server and Sybase Adaptive Server does not have an equivalent PL/SQL construct. Therefore, it is converted to a SELECT INTO WHERE EXISTS clause and an IF statement as shown in Example 3 above.
IF (SELECT... ) with comparison does not have an equivalent PL/SQL construct. Therefore, it is converted to a SELECT INTO...WHERE... clause, as shown in Example 4 above.
Comments
A RETURN statement is used to return a single value back to the calling program and works the same in both databases. Microsoft SQL Server and Sybase Adaptive Server can return only the numeric data type, while Oracle can return any of the server data types or the PL/SQL data types.
In a PL/SQL procedure, a RETURN statement can only return the control back to the calling program without returning any data. For this reason, the value is commented out if the Microsoft SQL Server and Sybase Adaptive Server procedure is converted to a PL/SQL procedure, but not commented out if converted to a PL/SQL function. The Migration Workbench does this automatically.
Comments
Microsoft SQL Server and Sybase Adaptive Server use RAISERROR to notify the client program of any error that occurred. This statement does not end the execution of the procedure, and the control is passed to the next statement.
PL/SQL provides similar functionality with RAISE_APPLICATION_ERROR statements.However, it ends the execution of the stored subprogram and returns the control to the calling program. It is equivalent to a combination of RAISERROR and a RETURN statement.
The Migration Workbench copies the error code and error message from a RAISERROR statement and places them in the RAISE_APPLICATION_ERROR statement appended to the error message.
Comments
The EXECUTE statement is used to execute another stored procedure from within a procedure. In PL/SQL, the procedure is called by its name within the PL/SQL block. If a procedure is converted to a PL/SQL function, make sure to assign the RETURN value to a variable when calling it (see the call to RecordExists in Table 3-22 above).
The Migration Workbench converts the parameter-calling convention to be either positional, named, or mixed. For information on parameter-calling conventions, see the Schema Objects section in this chapter.
Comments
The Migration Workbench can convert most WHILE constructs. However, the CONTINUE within a WHILE loop in Microsoft SQL Server and Sybase Adaptive Server does not have a direct equivalent in PL/SQL. It is simulated using the GOTO statement with a label. Because the Migration Workbench is a single-pass parser, it adds a label statement at the very beginning of every WHILE loop (see Example 2 in Table 3-23 above).
Comments
The GOTO <label> statement is converted automatically. No manual changes are required.
Comments
@@rowcount is converted to StoO_rowcnt, which takes its value from the PL/SQL cursor attribute SQL%ROWCOUNT.
@@error is converted to StoO_error, which contains the value returned by the SQLCODE function. The value returned by SQLCODE should only be assigned within an exception block; otherwise, it returns a value of zero. This requires that the Migration Workbench add a local exception block around every SQL statement and a few PL/SQL statements. Other global variables are converted with a warning message. These may need to be converted manually.
Comments
Assignment in Microsoft SQL Server and Sybase Adaptive Server is done using the SELECT statement as illustrated in Table 3-26.
PL/SQL assigns values to a variable as follows:
It uses the assignment statement to assign the value of a variable or an expression to a local variable. It assigns a value from a database using the SELECT..INTO clause. This requires that the SQL returns only one row, or a NULL value is assigned to the variable as can be seen in the following example:
SELECT empno INTO empno FROM employee WHERE ename = 'JOE RICHARDS'
Comments
Because of the differences in their architectures, Microsoft SQL Server and Sybase Adaptive Server stored procedures return data to the client program in a different way than Oracle.
Microsoft SQL Server and Sybase Adaptive Server and Oracle can all pass data to the client using output parameters in the stored procedures. Microsoft SQL Server and Sybase Adaptive Server use another method known as result sets to transfer the data from the server to client. The examples discussed here do not return multiple rows to the client.
In Example 1, the procedure returns a single row result set to the client which is converted to a PL/SQL procedure that returns a single row using the output parameters.
Example 1:
A SELECT statement is converted into a SELECT...INTO clause and the extra parameter "i_oval1" is added to the procedure definition. Since the Migration Workbench does not currently look up the data types on the Oracle server, it sets the default data type to VAR1CHAR2.
In Microsoft SQL Server and Sybase Adaptive Server, if the SELECT statement that assigns value to a variable returns more than one value, the last value that is returned is assigned to the variable.
Example 2:
The second example illustrates fetching data into a local variable. Since this is straightforward, the Migration Workbench handles it successfully.
Note: Microsoft SQL Server-specific SQL statements should be converted manually. The Migration Workbench handles ANSI-standard SQL statements only. |
Comments
The Microsoft SQL Server and Sybase Adaptive Server SELECT statement with a subquery as part of the SELECT list cannot be converted to PL/SQL using the Migration Workbench. Manual changes are needed to convert this type of SELECT statement.
The Migration Workbench writes appropriate comments in the output PL/SQL procedures and the subqueries are omitted.
Comments
T/SQL allows GROUP BY statements where the column used in the GROUP BY clause does not need to be part of the SELECT list. PL/SQL does not allow this type of GROUP BY clause.
The Migration Workbench converts this type of SELECT statement to PL/SQL. However, the equivalent PL/SQL statement returns an error in Oracle.
Comments
The Migration Workbench can convert Microsoft SQL Server-specific column aliases to the equivalent Oracle format. No manual changes are required.
Comments
An UPDATE with a FROM clause cannot be converted. Instead, the Migration Workbench provides a comment indicating that manual conversion is required.
There are two ways to convert UPDATE with a FROM statements, and these are illustrated below.
Method 1:
Use the subquery in the SET clause if columns are being updated to values coming from a different table. For example, consider the following T/SQL statement:
UPDATE titles SET pub_id = publishers.pub_id FROM titles, publishers WHERE titles.title like 'C%' AND publishers.pub_name = 'new age'
Convert this statement to the following PL/SQL statement in Oracle :
UPDATE titles SET pub_id ( SELECT a.pub_id FROM publishers a WHERE publishers.pub_name = 'new age' ) WHERE titles.title like 'C%'
Method 2:
Use the subquery in the WHERE clause for all other UPDATE...FROM statements. For example, consider the following T/SQL statement:
UPDATE shippint_parts SET qty = 0 FROM shipping_parts sp, suppliers s WHERE sp.supplier_num = s.supplier_num AND s.location = "USA"
Convert this statement to the following PL/SQL statement in Oracle:
UPDATE shipping_parts SET qty = 0 WHERE supplier_num IN ( SELECT supplier_num FROM suppliers WHERE location = 'USA')
Comments
A DELETE with FROM..FROM clause must be converted manually.
While converting DELETE with FROM..FROM clause, remove the second FROM clause. For example consider the following T/SQL statement:
DELETE FROM sales FROM sales,titles WHERE sales.title_id = titles.title_id AND titles.type = 'business'
Convert the above statement to the following PL/SQL statement in Oracle:
DELETE FROM sales WHERE title_id IN (SELECT title_id FROM titles WHERE type = 'business' )
Comments
Temporary tables are supported by Oracle9i and Oracle8i. The Migration Workbench utilizes this feature in Oracle9i and Oracle8i.
Also, SELECT..INTO..#TEMPTAB is converted to an INSERT statement. You must make manual changes to ensure that rows are unique to a particular session and all the rows for that session are deleted at the end of the operation. This requires that you add an extra column to the table definition and the value of USERENV('session_id') for all the rows inserted. At the end, delete all rows for that session_id. If many procedures use the same temp table in the same session, SEQUENCE can be used to make sure that the rows are unique to a particular session_id/SEQUENCE combination.
Command Option -M
Comments
Convert an Microsoft SQL Server and Sybase Adaptive Server procedure that returns a multi-row result set to a PL/SQL packaged function by selecting the appropriate parse option in the property sheet for a stored procedure.
The T/SQL SELECT statement is converted to a cursor and a cursor variable is added as an OUT parameter to return the data back to the calling program. Use the cursor referenced by the cursor variable to fetch the result rows.
For more details on how Result Sets are handled by the Migration Workbench, see T/SQL and PL/SQL Language Elements section in this chapter.
Comments
Microsoft SQL Server and Sybase Adaptive Server introduced cursors in T/SQL. Syntactical conversion of cursors from Microsoft SQL Server and Sybase Adaptive Server to Oracle is very straightforward.
Comments
The Migration Workbench does a one-to-one mapping when converting Microsoft SQL Server and Sybase Adaptive Server transaction commands to their Oracle equivalents. For more details about how transactions are handled in Oracle, see the Transaction-Handling Semantics topic later in this chapter.
Note: Make sure that the functionality remains the same, as the transaction models may differ in Microsoft SQL Server and Sybase Adaptive Server and Oracle. |
T/SQL is the Microsoft SQL Server and Sybase Adaptive Server procedural SQL language and PL/SQL is the Oracle procedural SQL language. This section discusses the following T/SQL and PL/SQL language elements:
Microsoft SQL Server and Sybase Adaptive Server
Microsoft SQL Server and Sybase Adaptive Server offer two different transaction models: the ANSI-standard implicit transaction model and the explicit transaction model.
Microsoft SQL Server and Sybase Adaptive Server provide options to support ANSI-standard transactions. These options can be set or un-set using the SET command.
The following SET command sets the implicit transaction mode:
set chained on
The following SET command sets the isolation level to the desired level:
set transaction isolation level {1|3}
isolation level 1 prevents dirty reads. Isolation level 2 prevents un-repeatable reads. Isolation level 3 prevents phantoms. Isolation level 3 is required by ANSI standards. For Microsoft SQL Server and Sybase Adaptive Server, the default is isolation level 1.
To implement isolation level 3, Microsoft SQL Server and Sybase Adaptive Server apply HOLDLOCK to all the tables taking part in the transaction. In Microsoft SQL Server and Sybase Adaptive Server, HOLDLOCK, along with page-level locks, can block users for a considerable length of time, causing poor response time.
If the Microsoft SQL Server and Sybase Adaptive Server application implements ANSI-standard chained (implicit) transactions with isolation level 3, the application migrates smoothly to Oracle because Oracle implements the ANSI-standard implicit transaction model, which ensures repeatable reads.
In a non-ANSI standard application, Microsoft SQL Server and Sybase Adaptive Server transactions are explicit. A logical transaction has to be explicitly started with the statement BEGIN TRANSACTION. The transaction is committed with a COMMIT TRANSACTION or rolled back with a ROLLBACK TRANSACTION statement. The transactions can be named. For example, the following statement starts a transaction named
account_tran. BEGIN TRANSACTION account_tran
The explicit transaction mode allows nested transactions. However, the nesting is only syntactical. Only outermost BEGIN TRANSACTION and COMMIT TRANSACTION statements actually create and commit the transaction. This could be confusing as the inner COMMIT TRANSACTION does not actually commit.
The following example illustrates the nested transactions:
BEGIN TRANSACTION /* T/SQL Statements */ BEGIN TRANSACTION /* T/SQL Statements */ BEGIN TRANSACTION account_tran /* T/SQL Statements */ IF SUCCESS COMMIT TRANSACTION account_tran ELSE ROLLBACK TRANSACTION account_tran END IF /* T/SQL Statements */ IF SUCCESS COMMIT TRANSACTION ELSE ROLLBACK TRANSACTION END IF /* T/SQL Statements */ COMMIT TRANSACTION
When BEGIN TRANSACTION and COMMIT TRANSACTION statements are nested, the outermost pair creates and commits the transaction while the inner pairs only keep track of nesting levels. The transaction is not committed until the outermost COMMIT TRANSACTION statement is executed. Normally the nesting of the transaction occurs when stored procedures containing BEGIN TRANSACTION /COMMIT TRANSACTION statements call other procedures with transaction-handling statements. The global variable @@trancount keeps track of the number of currently active transactions for the current user. If you have more than one open transaction, you need to ROLLBACK, then COMMIT.
The named and unnamed inner COMMIT TRANSACTION statements have no effect. The inner ROLLBACK TRANSACTION statements without the name roll back the statements to the outermost BEGIN TRANSACTION statement and the current transaction is canceled. The named inner ROLLBACK TRANSACTION statements cancel the respective named transactions.
Oracle
Oracle applies ANSI-standard implicit transaction methods. A logical transaction begins with the first executable SQL statement after a COMMIT, ROLLBACK, or connection to the database. A transaction ends with a COMMIT, ROLLBACK, or disconnection from the database. An implicit COMMIT statement is issued before and after each DDL statement. The implicit transaction model prevents artificial nesting of transactions because only one logical transaction per session can be in effect. The user can set SAVEPOINT in a transaction and roll back a partial transaction to the SAVEPOINT.
For example:
UPDATE test_table SET col1='value_1'; SAVEPOINT first_sp; UPDATE test_table SET col1='value_2'; ROLLBACK TO SAVEPOINT first_sp; COMMIT; /* col1 is 'value_1'*/
Logical transactions are handled differently in Microsoft SQL Server and Sybase Adaptive Server and Oracle. In Microsoft SQL Server and Sybase Adaptive Server, transactions are explicit by default. Oracle implements ANSI-standard implicit transactions. This prevents a direct conversion from T/SQL transaction-handling statements to PL/SQL transaction-handling statements.
Also, Microsoft SQL Server and Sybase Adaptive Server require that transactions in stored procedures be allowed to nest, whereas Oracle does not support transaction nesting.
The following table compares Microsoft SQL Server and Sybase Adaptive Server to Oracle transaction-handling statements:
At the time of conversion, the Migration Workbench cannot determine the nest level of the current transaction-handling statement. The variable @@trancount is a runtime environment variable.
Table 3-38 shows the currently implemented Microsoft SQL Server and Sybase Adaptive Server to Oracle conversion strategy for the transaction-handling statements
Because of the difference in the way the two databases handle transactions, you may want to consider some reorganization of the transactions.
Try to design client/server applications so that the transaction-handling statements are part of the client code rather than the stored procedure code. This strategy should work because the logical transactions are almost always designed by the user and should be controlled by the user.
For the conversion of stored procedures, consider setting a SAVEPOINT at the beginning of the procedures, and roll back only to the SAVEPOINT. In Microsoft SQL Server and Sybase Adaptive Server, make the changes so that at least the outermost transaction is controlled in the client application.
Microsoft SQL Server and Sybase Adaptive Server
In Microsoft SQL Server and Sybase Adaptive Server, you must check for errors after each SQL statement because control is passed to the next statement regardless of any error conditions generated by the previous statement. The client ERROR_HANDLER routine is invoked as a call-back routine if any server error occurs, and the error conditions can be handled in the call back routine.
Stored procedures use the RAISERROR statement to notify the client of any error condition. This statement does not cause the control to return to the calling routine.
Microsoft SQL Server and Sybase Adaptive Server allow you to customize the error messages using a system table. The system procedures allow the user to add error messages to this table.
Oracle
In Oracle, each SQL statement is automatically checked for errors before proceeding with the next statement. If an error occurs, control immediately jumps to an exception handler if one exists. This frees you from needing to check the status of every SQL statement. For example, if a SELECT statement does not find any row in the database, an exception is raised. The corresponding exception handler part of the block should include the code to deal with this error. The built-in RAISE_APPLICATION_ERROR procedure notifies the client of the server error condition and returns immediately to the calling routine.
Oracle places an implicit SAVEPOINT at the beginning of a procedure. The built-in RAISE_APPLICATION_ERROR procedure rolls back to this SAVEPOINT or the last committed transaction within the procedure. The control is returned to the calling routine.
The Oracle RAISE_APPLICATION_ERROR statement allows the user to customize the error message. If an exception is raised, SQLCODE is returned automatically by PL/SQL to the caller. It keeps propagating until it is handled.
Recommendations
To simulate Microsoft SQL Server and Sybase Adaptive Server behavior in Oracle, you must enclose each SQL statement in an equivalent PL/SQL block. This block must deal with the exceptions that need to be trapped for the SQL statement.
See the T/SQL Versus PL/SQL Constructs section in this chapter for more information about the extra code required to simulate Microsoft SQL Server and Sybase Adaptive Server behavior.
If the RAISERROR statement in an Microsoft SQL Server and Sybase Adaptive Server stored procedure is immediately followed by the RETURN statement, these two statements can be converted to the Oracle RAISE_APPLICATION_ERROR statement.
You can customize error messages with the help of a user-defined table. You can write standard routines to add and retrieve error messages to this table. This method serves a two-fold purpose: it ensures that the system is portable, and it gives the administrator centralized control over the error messages.
Microsoft SQL Server and Sybase Adaptive Server
The following global variables are particularly useful in the conversion process:
@@error:
The server error code indicating the execution status of the most recently executed T/SQL statement. For code examples, see the @@Rowcount and @@Error Variables topic.
@@identity:
Returns the last identity value generated by the statement. It does not revert to a previous setting due to ROLLBACKS or other transactions.
@@rowcount:
The number of rows affected by the most recently executed T/SQL statement. For code examples, see the @@Rowcount and @@Error Variables topic.
@@servername:
The name of the local Microsoft SQL Server and Sybase Adaptive Server server.
@@sqlstatus:
The status information resulting from the last FETCH statements.
@@tranchained:
The current transaction mode of the T/SQL procedure. If @@tranchained returns 1, the TL/SQL procedure is in chained, or implicit transaction mode.
@@trancount:
Keeps track of the nesting level for the nested transactions for the current user.
@@transtate:
The current state of the transaction.
Oracle
SQLCODE:
The server error code indicating the execution status of the most recently executed PL/SQL statement.
SQL%ROWCOUNT:
The variable attached to the implicit cursor associated with each SQL statement executed from within the PL/SQL procedures. This variable contains the number of rows affected by the execution of the SQL statement attached to the implicit cursor.
Recommendations:
The @@error variable has a direct equivalent in Oracle, and that is the SQLCODE function. The SQLCODE function returns the server error code.
The SQL%ROWCOUNT variable in Oracle is functionally equivalent to @@rowcount.
There are many more special global variables available with PL/SQL. Not all those variables are listed here. There are more special global variables available in T/SQL also. Not all those variables are listed here because they do not play a major role in the conversion process.
See the Data Manipulation Language section in Chapter 2 for a discussion of Microsoft SQL Server and Sybase Adaptive Server and Oracle operators.
See the Data Manipulation Language section in Chapter 2 for a discussion of built-in functions in Microsoft SQL Server and Sybase Adaptive Server and Oracle.
Microsoft SQL Server and Sybase Adaptive Server stored procedures can return data to the client by means of a Result Set. A SELECT statement that does not assign values to the local variables sends the data to the client in the form of byte-stream.
In a case where a third-party user interface product uses the result set capability of Microsoft SQL Server and Sybase Adaptive Server, consult with the vendor to make sure that the same functionality is available for the Oracle database.
The following example procedure sends the data out as a result set. More appropriately, an OUTPUT parameter holding the value "YES" or "NO" (depending upon the evaluation of <condition>) or a function returning "YES" or "NO" should have been used.
CREATE PROCEDURE x AS BEGIN ... ... IF <condition> THEN SELECT "YES" ELSE SELECT "NO" END
Avoid Microsoft SQL Server and Sybase Adaptive Server stored procedures that return multiple result sets to the calling routine.
The following procedure returns two different result sets, which the client is responsible for processing:
CREATE PROCEDURE example_proc AS BEGIN SELECT empno,empname, empaddr FROM emp WHERE empno BETWEEN 1000 and 2000 SELECT empno,deptno, deptname FROM emp, dept WHERE empno.empno = dept.empno AND emp.empno BETWEEN 1000 and 2000 END
Recommendations
Some alternatives to simulating the result set in PL/SQL procedures are presented below:
Method 1
A T/SQL procedure with a result set may need some manual changes after conversion to an Oracle package with a member function. The problems are described in detail below.
For example, consider the following T/SQL procedure:
CREATE PROC test_proc AS BEGIN T/SQL block1 T/SQL block2 SELECT statement corresponding to the result set END
This procedure executes two T/SQL blocks before executing the SELECT statement associated with the result set. The procedure is converted to an Oracle package as follows:
CREATE OR REPLACE PACKAGE BODY test_proc_pkg AS BEGIN FUNCTION test_proc; END; CREATE OR REPLACE PACKAGE BODY test_proc_pkg AS BEGIN cursor declaration for the SELECT statement associated with the result set in the source T/SQL procedure; FUNCTION test_proc RETURN INTEGER AS BEGIN PL/SQL version of T/SQL block1; PL/SQL version of T/SQL block2; FETCH loop for the cursor declared in the package body; END; END;
The two T/SQL blocks in the source T/SQL procedure are executed only once when the procedure is called, and the result set is sent to the client.
In Oracle client, to simulate the fetching of the result set, the TEST_PROC_PKG.TEST_PROC function must be called repeatedly until all the rows from the cursor are fetched. The two PL/SQL blocks in the function are executed with each call to the function. This behavior differs from that in the source application.
You must manually separate the code associated with the FETCH loop for the cursor for the result set from the remaining code in the procedure. Changes to the client have to be made so that the rest of the procedure's code is called in accurate sequence with the repeated calls to the function returning rows from the result set.
The final Oracle package should be as follows:
CREATE OR REPLACE PACKAGE BODY test_proc_pkg AS BEGIN PROCEDURE proc1; FUNCTION test_proc; END; CREATE OR REPLACE PACKAGE BODY test_proc_pkg AS BEGIN cursor declaration for the SELECT statement associated with the result set in the source T/SQL procedure; PROCEDURE proc1 AS BEGIN PL/SQL version of T/SQL block1; PL/SQL version of T/SQL block2; END; FUNCTION test_proc RETURN INTEGER AS BEGIN FETCH loop for the cursor declared in the package body; END; END;
The client should call the TEST_PROC_PKG.PROC1 procedure before repeatedly calling the TEST_PROC.PKG.TEXT_PROC function in order to achieve functionality similar to the source T/SQL procedure.
The variables that are common to these two parts should be either declared globally within the package body or should be passed as parameters to the procedure and the function.
Microsoft SQL Server and Sybase Adaptive Server allow DDL constructs to be part of the stored procedures. Oracle allows DDL statements as part of the dynamic SQL. Oracle issues an implicit COMMIT statement after each DDL statement.
Most of the T/SQL DDL constructs give syntax errors. You must remove the DDL statements from the T/SQL source to convert the T/SQL procedure to PL/SQL using the Migration Workbench.
The following DDL statements are ignored by the Migration Workbench. The statements appear commented in the output with a message "statement ignored."
CREATE TABLE DROP TABLE CREATE VIEW DROP VIEW CREATE INDEX DROP INDEX
|
Copyright © 1998, 2002 Oracle Corporation. All Rights Reserved. |
|