ASNA's DataGate for SQL Server Classic (DSS for .NET) provides AVR programmers’ access to Microsoft SQL Server 2000 through native RPG file IO using familiar RPG operation codes such as CHAIN and SETLL. DSS for .NET allows you write one set of applications that works concurrently with both the iSeries and SQL Server databases.

You will not need to learn SQL to get effective access to SQL Server. The skills you use today for iSeries file IO will work exactly the same way for SQL Server via DSS for .NET. The same AVR programs that connect to SQL Server will seamlessly connect to the iSeries and ASNA’s DataGate engine.

DSS for .NET makes a SQL Server instance appear as a database similar to DataGate or iSeries. The characteristic features of DB2/400 materialize on SQL Server when seen through the eyes of DSS for .NET. With DSS for .NET, you can use the Library list, Physical and Logical files, even DB2 field types, like packed and zoned.

Some Restrictions Apply

Even though DSS for .NET tries to make SQL Server look like DB2/400, there are several features that cannot be implemented in a totally transparent fashion. The SQL Server 2000 database engine is of a different design and implementation than the iSeries and DataGate engines.

The intent of this document is to help you become aware of the items that will most likely affect your application and the process of using SQL Server as your database. Through it all, remember the goal you are seeking is to create applications that take advantage of the many features of SQL Server and to create applications that can run with either SQL Server or DB2/400 as the underlying database engine.

In the next sections we’ll deal with all of the issues that will need your attention, but the following items are probably the ones with the most impact for many of you.

No Multi-Format Logical Files:

DSS for .NET implements a physical file through the use of a native SQL Server table. A logical file is implemented through a native view. SQL Server Views are single formatted in nature, so there is no support for multi-format logical data files. You will have to eliminate any reference to multi-format logical files in your application.

Print files, although they are typically multi-formatted, are fully supported in DSS for .NET, however.

Single Member Files:

SQL Server doesn’t have the concept of members of a table/view. DSS for .NET makes it appear as though each file had one (and only one) member; the member name is exactly the same as the file name. For the member name, you can use the exact name or the special values *FIRST and *FILE. You should be careful when using the Copy Data and Copy Library tools of the DataGate Database Manager to copy a file from DataGate or iSeries, because they default to *SAME for the target member name; if the source member name was differently than the file name the copy will fail.

If your application depends on the existence of multiple members per file, you will want to re-architect it to provide an alternative method for handling the logic involving these files.

Logical Field Restrictions:

There are two restrictions on the usage of logical fields when they change the name or the type of their corresponding base physical field. When the field is retyped, most typically because the field is a concatenation or substring of the physical, then the field becomes read only. A logical field, whose name has changed from its physical base field, can’t be used as a key field in the logical file.

No Qtemp:

Version 5.0 of DSS for .NET does not support the special library QTEMP.

Unlocking Records:

This is probably the most demanding area of application adaptation between the differences of implementation between iSeries method of record locking and that of the SQL Server database engine. The problem arises in two areas: Using the *nolock keyword on the read operations and on the implementation of the UNLOCK command.

DSS for .NET uses SQL Server ‘Server Cursors’ to implement file access. When a file is opened for update, it is not possible to tell SQL Server to not lock the record on a read, so a read with *nolock has no effect for files opened for Update. There are two methods to resolve this problem:

1. Declare a second instance of the file marked as input only and use it wherever the NoLock option would have been given on a read/chain.

2. Retain the *nolock and follow the read/chain with an UNLOCK command. Be aware this method imposes some restrictions as stated in the next paragraph.

The UNLOCK command leaves the cursor in a no-position state, meaning you can’t perform a subsequent read (next/previous) without repositioning the file with a SETLL, SETGT or CHAIN.

The Process

In a nutshell, the process of upgrading your application is composed of the following steps:

1. Get a decent size computer.

2. Move your data from the iSeries or DataGate Engine to SQL Server.

3. If necessary, modify your application to adapt to any new restrictions.

4. Optimize your application to take advantage of the new AVR op-codes.

1. Computer Size

SQL Server is very demanding on computer resources. We recommend you get a computer with at least the capacity of the cheapest machine you can get from Dell, but with twice the amount of memory. The machine doesn’t have to be a Dell computer, but use the Dell machine as a reference to configure your box.

One more thing, we highly recommend that you use Windows XP Professional or Windows 2000 workstation for your development machine.

2. Move Your Data

Create a Database Name:

The first step after installing DataGate for SQL Server on your machine is to create a Database Name that points to your SQL Server. You can use the Database Wizard or the Work with Database Names option of the Database menu of Database Manager.

Under typical SQL Server installations, there is only one instance of the server installed on the machine and that instance typically contains the following databases:

· Master

· Model

· Msdb

· Northwind

· Pubs

· Tempdb

Each one of these databases will be presented through DSS for .NET as individual libraries. Since all databases are shown under a common database name, there is no use for the ‘Label’ parameter of a database name. It is conventional to use a label of “SQL”, but the name is actually irrelevant.

The following figure shows a Database Name called Ontario pointing to the SQL Server engine in My Computer.

You can use *DOMAIN for your user name or you can supply one of the login names recognized by SQL Server with its corresponding password.

Open the Database:

Test your Database Name by opening the database with Database Manager. You should see something like the following:

When you open a database for the first time, Database Manager prompts you for your viewing preferences for the database with a dialog like the following.

Notice in particular the checkbox Show system objects. It controls whether to or not to display a set of SQL Server objects called sysxxxx. For iSeries databases, this option shows or hides objects like QSYS.

Prepare Files and Copy Libraries:

The next step in moving your data is to prepare a library (or libraries) on the source machine with all the files you wish to copy to the target SQL Server.

After you have prepared the library, use the Copy Library option from the Tool menu to move the data.

You can use one of your existing libraries, but it may require some clean up before it is ready. To check your data, use one of the options below:

1. You can do a visual inspection of the files in the library to verify/modify them to comply with the check list.

2. Alternatively issue the copy command and wait for it to tell you whether there are any problems.

Visual Inspection Using a Checklist:

Use the following checklist to ensure a DataGate or iSeries data file will be successfully copied to SQL Server:

The file must have:

· only one member

· the member name the same as the file name

· only one format

· maximum of 16 key fields

· maximum key length of 900 bytes

· maximum of 1024 fields

· maximum record length of 8060 bytes

Logical files have extra restrictions. Make sure that the key fields are not substrings or concatenations of physical fields and that the name of the logical field is the same as that of its base field.

Issue the Copy Command and Note if there are any Problems

If you choose the second option, uncheck the Copy Data Records in the Copy Library dialog to speed up the process. Take note of any problem and after fixing it, delete the target library and issue the Copy Library command again, this time copying the data records. You can view the result of the copy in the Message Log window.

Copy Library:

Once you have modified your files to comply with the limitations stated above, you can use the Copy Library menu option.

All files stored under SQL Server must reside inside a ‘Library’ just like files on an iSeries. DataGate Engine has allowed files to be placed at the ‘root’ level of the database and under subfolders inside a library. If you have files in places other than first level libraries, you’ll have to adapt your application (or the library list) to point to the new locations.

When a physical file is copied to SQL Server, a table is created with the name of the physical file. For logical files, a view is created. In either case, if the file is keyed, an index is created on the table; for logical files the index is created on the primary base table.

When a view is encountered through DSS for .NET, it is reported as being a new kind of logical file, called SQL Logical. That is why when a logical file is copied, regardless of whether a logical file is a Simple Logical or a Join Logical, it will appear on the SQL Server database as a SQL Logical file.

Field Reference Files

Field Reference Files are used on the iSeries to collect definitions of named field types that are used in the creation of data files. SQL Server has a built-in data dictionary where user types can be defined. DSS for .NET surfaces this data dictionary through the special file *FieldRef. There is one *FieldRef file per library.

If you use a field reference file as part of your data file definitions, you should first import those fields into the data dictionary of the DSS for .NET library.

Follow these steps to populate the dictionary:

1. Open the source database.

2. Drag the field reference file to the Work Definition Area.

3. Open the target database.

4. Create the target library.

5. Drag the field reference file to the target database taking care of specifying *FieldRef as the file name to be created. DSS for .NET will notice the special name and add the fields to the data dictionary.

After you have populated the data dictionary, you can view the type definitions by clicking on the *FieldRef file in the target library.

If you don’t see the *FieldRef file in the library, ensure that the option to Show System Objects is checked ON in the options dialog.

3. Adapt the Application

Enable DataGate for SQL Server Warnings:

To assist in finding areas in code that will potentially cause an error, a new IDE option has been added to ASNA Visual RPG to give warnings for invalid operations under DSS for .NET. This option is located within the “Compiler” Tab of the Project >Settings menu option.

Select the Enable DataGate for SQL Server Warnings option so that a check mark appears in the check box and select OK.

Change your DCLDISKFILE Statements:

Depending upon your application, the following are some changes you will/may need to make to your DCLDISKFILE statements.

*Arrival processing:

Arrival sequence processing must be changed to indexed processing. SQL Server doesn’t have the concept of a relative record number (RRN).

The simplest method for handling this would be to employ the use of a key field in the file. If the file is being processed only in consecutive fashion (i.e. you don’t use random op-codes like SETLL or CHAIN on the file) then you can choose any field to be the key and specify RANDOM(*NO) in the DCLDISKFILE. If you do process the file randomly, then you will need to add a field to the file to simulate the RRN sequence. The easiest way to achieve this is by adding an identity field called RecNum of type integer.

Format names:

In DSS for .NET, the record format name is always “R” + Filename. If the record format name is anything else, and you’ve specified this record format name on I/O operations, a rename format will need to be specified on the DCLDISKFILE statement as follows. RNMFMT(OldFormatName) where OldFormatName is the name of the existing format. No changes will need to be made to the actual I/O statements.

Query Files:

If your application makes use of the Open Query File capability of DataGate or iSeries, you will have to take care of a couple of things.

§ The string passed in the QrySelect must comply with the syntax of SQL Server, for example you should use the word ‘and’ not the symbol ‘&’.

§ If you are changing the order of the file by providing a value in the QryKeyFlds parameter, then you must also specify RANDOM(*NO) in the DCLDISKFILE to state that you will be accessing this file only consecutively.

Check Format ID:

If an application is going to run against both DSS for .NET and DataGate 400, and your file currently contains binary fields, you will need to consider changing the field type.

Unlocking Records:

The behavior of DSS for .NET when the file is opened for update is similar to DG/400, but with two significant differences:

1. Updating a record does not release the lock on the record.

2. Explicitly unlocking a record causes the ‘current record position’ to be lost.

These differences bear the following considerations:

Loops involving SetLL-SetGT and Read-ReadE-ReadPE should be re-coded to use the Range operations.

The most demanding change is the one requiring segments of code involving CHAIN-UPDATE combinations to be studied and possibly modified.

· If the CHAIN-UPDATE happens in a tight loop, then at the end of the loop, an UNLOCK should be issued to release the last record updated. Notice however that the record position will be lost after the UNLOCK.

· If the CHAIN-UPDATE is sprinkled throughout the code, then each case has to be closely studied.

4. Optimize Your Applications with New Op-Codes

To optimize SQL Server processing and to enhance client/server performance with all supported database engines (iSeries/400, DataGate and SQL Server) with dynamic Network Blocking, AVR 4.0 introduces three new op-codes; SetRange, ReadRange and DeleteRange, as described below:

New I/O op-codes provide better performance in two ways:

1. When working with large SQL Server files, DSS for .NET is able to optimize the record set selection operation.

2. Network Blocking becomes dynamic, in that records outside of the range are not transmitted to the client.

SetRange

Use the SETRANGE op-code in place of SETLL and SETGT when you’re doing SETLL/READE and SETGT/READPE loops.

ReadRange

Use the READRANGE op-code in place of the CHAIN op-code when you’re doing CHAIN/READE loops.

DeleteRange

Use the DELETERANGE in place of DELETE loops.

Error Handling

Strong consideration should be given to Visual RPG Error Indicators when performing I/O operations as they could be turned on now for reasons not present before. For instance, when the Chain opcode returned an error, the cause was commonly assumed to be that the record was busy. With DSS for .NET, error indicators can be turned on for various reasons such as the use of the *NOLOCK option on I/O operations. (See Locking Considerations in the DG400 vs. DSS for .NET document).

New Error Conditions under DSS for .NET:

The following situations will cause the error indicator to be set on.

1. Arrival Random Access.

When a file is opened for *Arrival access, commands such as Chain, Setll, etc, will cause an error condition.

2. Use of the *NOLOCK option is not supported under DSS for .NET.

Any I/O operations that make use of the *NOLOCK option when working with files opened for update will cause an error condition.