castor

Hello Ralf,

Does it really need to replace StringBuffer? What is the advantage of a Writer implementation (like StringWriter) in compare with StringBuffer?

I will replace PrintStream by Writer.

We have various properties to configure output that are independed of the schema object:

• org.castor.ddlgen.CharFormat=SENSITIVE, UPPER and LOWER
• org.castor.ddlgen.Newline=\n
• org.castor.ddlgen.Indention=\t

These properties are accessed at various places all around ddlgen at the moment.

The idea is that these properties are set only once at the new Writer and do not need to be accessed elsewhere. This has the following advantages:

• improved performance as the properties don't need to be accessed for every object to output
• functionallity to format genertaed ddl is concentrated in one class: the new Writer
• all the toDDL(), toDropDDL(), toCreateDDL() methods get much shorter

I thought of the following interface for the new Writer (not complete):

• write(String) outputs String as is
• writeln(String) calls write(String) followed by newline()
• newline() output newline and indention of next line
• indent() increases indention
• unindent decreases indention

More write and writeln methods for other data types may be added on demand. A further improvement could be to offer write(String, Object[]) methods that internally use MessageFormat. This would enable us to use a pattern based approach for DDL generation. These patterns may sometimes be much easier to read and maintain.

In addition to the introduction of the new Writer it will be required to pass a instance of the Writer to every method where DDL gets generated. Therefore the parameterless toCreate() method have to be changed to toCreateDDL(DDLWriter). This also applies to other such methods.

Hope this explains the idea behind this issue a bit better then the summary alone.

I created the DDLWriter which encapsulates almost common task for writer. It will replace the PrintStream and StringBuffer.

Thanks for your work. When looking at the writer code you provided I recognized the following:

• you are using java 5 features but we like ddlgen to be compliant with java 1.4
• could you please take a bit more care on formating of javadoc and code
• isn't it possible for you to prepare patch files as that would be required for changes to available classes
• why are the methods for objects called writeObject() instead of write(). The compiler should be able to choose the right one
• the writer does not handle char formating: SENSITIVE, UPPER and LOWER
• after thinking again about when to output indent characters I came to the conclusion that it would be better to output them at the first write to a new line and not directly after writing the newline as that may cause wrong indention in following situation.

w.writeln("line1");
w.indent();
w.writeln("line2");
w.unindent();
w.writeln("line3");

In that case line 3 will be indented even if it should not be indented.

Hi Ralf,
I changed all signature createxxx() to createxxx(DDLWriter) and also fix the tester. By the way, I cannot run the test suite. Can you help me?

Which test suite ? XML or JDO ?

The DDLGen test suite.

I'll review your patch and run DDLGEN test suite

Final patch