Show Lecture.DivideAndConquer as a slide show.

CS253 Divide And Conquer

Julius Caesar, thinking “Veni, vidi, vici”

Horror Stories

• Don’t just start typing—you have to design your code.
• Programming tasks tend to be complex.
• Don’t do it all in main(), even if it works.
• If you fear writing functions, change your major now. 🍟
• If a program works, that’s just the start, not the end. Maintainability matters.
• In the real world (and in CS253), programs live for a while. Code from long ago will come back to haunt you.
• Also, modularity will enable you to actually write correct code.

Divide and Conquer

• The concept of “divide and conquer” comes from warfare.
• Don’t try to defeat all of your enemies at once.
• Divide your enemies into partisan groups that won’t help each other.
• Then, defeat them separately, as smaller groups.

Programming

• In programming, we use similar divide and conquer techniques.
• Huge problems can be broken into several large problems.
• Large problems can be broken into several small problems.
• Solve each small problem separately.
• We have monkey brains that are not designed to create complex algorithms. Make it easy on your 🐒 🧠 by giving it smaller problems to deal with.

Programming

• Typically, when programming, we break down complex problems by creating functions (or perhaps objects).
• Avoid the giant MAIN() FROM 🔥🔥🔥🔥.
• Instead, break down the problem into smaller problems, one per function.
• If those functions are still too big, break them into yet smaller functions.
• Rule of thumb: if a function doesn’t fit on the screen, then it’s too big—break it up.
• However, don’t fear small functions: it’s fine to have a one-line function bool is_valid_input(char).

Example

Consider the program grep:

• It takes a regular expression pattern and a number of filename arguments.
• It looks for the pattern in all the files (or standard input).
• It displays the lines that match the pattern.
• Depending on the options given:
  • Display the lines that don’t match.
  • Maybe the whole line, or maybe just the matching part.
  • Maybe with line numbers.
  • Maybe just the first matching line.

How not to do it

    int main(int argc, char *argv[]) {
        parse options;
        if (no arguments)
            for (each line in standard input)
        	if (line matches the pattern or maybe if it doesn’t) {
        	    display the line number, if required;
        	    display the line according to the options;
        	    if (only list the first match)
        		break;
        	}
        else
            for (int i=1; i<argc; i++)
        	for (each line in that file)
        	    if (line matches the pattern or maybe if it doesn’t) {
        		display the line number, if required;
        		display the line according to the options;
        		if (only list the first match)
        		    break;
        	    }
    }

Not DRY, and would really be much longer. grep has over 40 options.

How to do it

    int main(int argc, char *argv[]) {
        parse_options(argv);
        if (no arguments)
            process_stream(cin);
        else
            for (int i=1; i<argc; i++)
        	process_file(argv[i]);
    }

There—that’s manageable.

Break it down

    void process_file(const string &filename) {
        ifstream in(filename);
        if (!in) {
            complain;
            exit(1);
        }
        process_stream(in);
    }

No duplication of code: process_file() uses process_stream().

Break it down

    void process_stream(istream &in) {
        for (string line; getline(in, line); )
            if (match_pattern(line)) {
        	display_line(line);
        	if (only list the first match)
        	    break;
            }
    }

• Each function does one easily-described task.
• process_stream():
  • doesn’t know anything about argv
  • doesn’t care how match_pattern() works
  • doesn’t care how display_line() works
  • doesn’t even know what a regular expression is.

Objections

This will make my program longer!

So? Are you running out of disk space?

This will slow my program down!

Really? I’ll bet that you can’t measure the difference, with all this file I/O and regular expression matching going on.

I can understand it just fine if it’s all in main()! Can’t you ?

Aren’t you masculine. Yes, it’s possible to do that way, but, …, why ? Save your effort for the tasks that need  to be difficult.