This document summarizes the design aspects of the [ini4j] project. First let me quote something every library designer should keep in mind:
"Perfection (in design) is achieved not when there is nothing more to add, but rather when there is nothing more to take away." - Antoine de Saint-Exupéry
One of the most important aspects was that the [ini4j] is not a complex configuration management system, but principally a library that allows you to manage files in .ini format.
We could say that we regard the java.util.Properties object as a pattern in designing. The Properties object works as a file containing pairs of name and value. It can be read from a stream or be written to a stream (lately even in XML format). This way it is understandable why the Ini object doesn't need to belong to a file: you can read it from InputStream (and Reader) or write it to OutputStream (and Writer).
The programming interface had to be designed as simple as possible, firstly for the sake of easier understanding and usage, secondly in order to avoid needless complexity. The handling of the .ini files in any application is just a tool, not the goal.
As a result of this approach the Ini object is in fact a Map composed of sections. The section object is a Map composed of String values assigned to String keys.
Another very important design aspect was that if a standard API exists, that must be used to handle the .ini file. Fortunately, ever since the introduction of version 1.4, there is a Preferences API, which is actually a simple tree structure of name/value pairs. This is ideal for the .ini format's top level API, because the programmer using the library doesn't have to learn a new API, but can handle the .ini files using the Preferences API included in the standard JDK!
So the .ini file is in fact a one-level Preferences tree. On the only level the nodes of the tree are the sections of the .ini file, and the leaf nodes are the section's name/value pairs. Due to the Preferences API the conversion of values to primitive types or vice versa is solved (getInt(), setInt(), etc).
The designer of a library always seeks the favour of the potential users, the programmers in this case, so comfort is at least as important an aspect as simplicity. An API should be as comfortable as possible. The Java Beans-style programming interface was born on these terms, that is, to allow the programmer to define completely freely how (as what type) he'd wish to access the values of the section as a java interface.
When handling large .ini files, it is not practical to read the whole file into the memory. So there might be a need for a sequential parser that calls the processing routine as a callback while parsing sections or name/value pairs inside sections. The JAXP SAX parser can be regarded as a model.
Along these lines, the corresponding document object of the .ini file must be built based on this sequential parser. Thus the model of the Ini object is the JAXP DOM document.
A bit of history: an early ancestor of the [ini4j] library was an exercise assigned as homework at a Java programmer course. Its purpose was to prepare the exposition of JAXP SAX/DOM API. The assignment was to write a sequential parser for the .ini file, and then to use it to build an object in the memory.