C++ is a multi-paradigm programming language. This means that there is no single style that is always the right one to use on C++ programs. It all depends on the task at hand. For example, it is completely fine to use C++ only as a structured language, since it has all elements of C and a few more that make structured programming even easier in C++.

The main style of use of the C++ programming language is the classic object oriented style, since it is a style that is easier to reconcile with business practices that have been developed in the previous years.

However, not every project can be done with class object oriented methods. Sometimes it is more productive to use ideas developed in other programming traditions, such as symbolic, functional, and dynamic programming. C++ can also do this, but it needs a little additional thought from the part of the program designer.

The book Advanced C++ Programming Styles and Idioms discusses several idioms that can be successfully used with the C++ programming language.

Many of these idioms are not mainstream, and some of them may not be even acceptable in your organization. Despite this, these idioms could just as well  be the tool of choice for specific problem domains.

Dynamic Features in C++
Among these idioms, are techniques that have been used for a long time in dynamic languages such as Smalltalk and Lisp, such as dynamic dispatching, symbolic computing, and related techniques.

The author, James Coplien, describes how to implement object orientation based on exemplar objects, instead of class based objects. In examplar-based programming, one uses objects as the foundation for creating objects, instead of using classes — as is the normal practice in C++. One of the keys for doing this is to use factory methods to create new objects, instead of using the new keyword, which by definition requires knowledge about the specific class of the created object.

Another strategy is to use Envelope objects, which provide some of the functionality of symbols in a language such as Lisp. Using Envelope objects, one can have automatic memory management through reference counting, and a clean separation between a value and the label that contains it.

Coplien describes how to create more dynamic objects by avoid the classic hierarchy-based approach for classes. He proposes two main solutions for this problem: in the first solution, one creates a shallow hierarchy, where each class has all possible methods that could be implemented by subclasses. In this approach, each object is always crated using the interface of the parent class, and implements only the methods that are of their interest. Subclasses rely on polymorphism to receive the correct messages, and send an error whenever they don’t respond to a particular message.

This technique has the great advantage that clients don’t need to know exactly what class is being called. Also, since the differences between objects is now based only on what methods they respond to, the system is now based on generic methods, rather than on specific classes. Any class can override any method of the top class, which makes it easy to supply dynamic behavior.

The drawback of the previous approach however is that, unless we known all methods that are of interest, the system is subject to recompilation whenever a new method is added to the parent class. To avoid this problem, one can go one step further and create a dynamic dispatch system. In such a system, there is just one method to which an operation descriptor is passed, followed by a number of parameters. In this way, easy concrete subclass can add as many operations as necessary, without any change to the parent class. While such a system may not perform as quickly as a statically defined dispatch method, it is much more flexible, and allows the system to evolve without recompilation.

Finally, the books discuss how to load code dynamically, and therefore avoid linking. The approach works on most systems, however it is machine and compiler-dependent, so it should be used with care when portability is of importance for the system.

Further Reading

Check Advanced C++ Programming Styles and Idioms at Amazon.

Some time ago I wrote about the sam text editor, a UNIX text editor that incorporates many features that are cherished by UNIX users, such as regular expressions and programmability.

It turns out that many of these features were already present on the standard UNIX editor: ed. For many, it seems crazy that one might want to edit text in an editor that is not screen oriented, and is capable only of showing individual lines. However, sometimes this is not only OK, it is even desirable.

The most common case for me is when I am on a UNIX command line. Frequently there is a small change that needs to be performed to a file in order to continue. For example, this happens most of the time when I am compiling code with gcc or some other compiler that I might be using at the moment. After a compilation error happens, the compiler will say exactly what file is the culprit, and at what line the problem occurred.

The common UNIX user will then start their preferred screen-oriented editor to make the change. This can be one of vi, emacs, gedit, eclipse, or whatever editor you may be into. Then, after the editor starts (which, depending on the program, may take a few seconds — or minutes if it is eclipse), you need to locate the line reported by the compiler, and make the necessary change.

The point is that you don’t need to have access to the whole file to make this simple edit. You just need to see one line of the file. An editor such as emacs or vi will do the work, but it can only make it more confusing, by displaying all the other lines that you don’t need.

With ed, you can just type ed followed by the name of the file: ed will start immediately, since there is no UI to wait for. Finally you type the number of the line followed by enter, and the desired line is displayed. It is now just a matter of using a single command to change the line (either s or c, depending on the kind of changes you want to perform).

Another situation where ed is really useful is when you just want to add lines, either to the end or the beginning of a file. That’s another case when it doesn’t matter that you have access to a screen-oriented editor: you just need to enter text into a very specific point of the file. What you can very well do in this case is to forgo the screen editor and use ed to add as much text as you want to the beginning (using 1 followed by i) or end (using the “a” command).

So, you see that using ed has a lot to do with the way UNIX systems work. You always want to use the tool that is most efficient for the task at hand. If the task is simple, the best way to do it is using a simple tool such as ed.

When I am writing, one of my constant quests has been to find a good editor where I can not only write but also organize my ideas. As strange as it might think, my best solution up to now has been not a text editor, but a presentation software. Yes, that is true, I use power point to write text.

One of the things that many users don’t know is that power point can also be a decent text editor and, in some respects, even better than Microsoft Word — just change the font used to “Times Roman” 10pts and you’re ready. There are some reasons why I think it is easier to edit large documents using power point than Word.

First, it provides a clear distinction between pages. Each page is a “slide”. You can create as many pages as you want, and they are independent of each other. You can move then around, and the content will also move around.
It is just like having old and trusted pages of paper.

This solves a major organization problem. When I edit a long text document, sooner than later I have problems with keeping information around. Where to put summaries, plots, notes, research, and many other things I need?

Clearly, one can create separate files for each document, and use a complicated directory structure to keep things organized. Whatever you do, at the end it is always confusing to keep your many ideas in a logical shape.

For example, if you have a new idea you would later want to incorporate into a document, where are you going to write it down? Simple: create a new slide to store it. An individual slide then becomes something hybrid between a file and a page of text. But you don’t have to go through the hassle of creating a new file, everything is stored in the presentation. When you run out of space in one page, create another one. It is as simple as that.
In the end, you are going to have as many sections in your document as you want, and they can be organized easily. You can move then around in the sorting order, and delete the ones you don’t want.

For example, I use the structure of power point to do several things. If I am creating a blog, I use each slide of the presentation to store one post. The slide contains roughly the size of a blog post, so I know that I should keep just one page per post, and the number of pages is the number of posts.

When writing a book, I use a power point file for the whole project, and each slide is a page. Then I use the summary tools to determine what I’m going to write in each page. It is the easiest way I have find to organize my writing — no more papers or simple notes left around.