1 A Model of I/O

I/O is one of the least standardized components of modern programming languages. It seems to be one the hardest things to get “right” when designing programming languages.

In many early programming languages, the input and output statements were “fundamental” statement types, like assignment or loops, built into the syntax of the language. Later languages tended to treat input and output as an offshoot of the language’s standard library of pre-written functions.

But even then, there was considerable variation on how those libraries were designed. Designers would try to find a few unifying ideas to base their libraries upon, only to find it difficult to apply them uniformly.

Designers would try to find a few unifying ideas to base their libraries upon, only to find it difficult to apply them uniformly.

C++ bases its I/O model on the >> and << operators. These take the form:

input-stream >> variable-to-read-into

and

output-stream << variable-to-be-written

(If you think of the >> and << as arrows, the indicate the direction in which the data is “moving” during the I/O operation.)

This is a very flexible model, because

1. On the left, we can have all manner of different I/O devices or destinations. We could have the standard streams cin and cout, or a file, or a network connection, or …

2. On the right, we can have almost any data type. For the simple data types predefined in the C++ language, the I/O operators are already defined. And C++ allows (encourages) programmers who invent their own data types to create their own code for use when applyng << and >> on those types.

2 Standard Input and Output

Almost all modern programming languages recognize the notion of a “standard input” that, by default, reads from the keyboard, and a “standard output” that, by default, writes to the screen.

In C++, these are cin and cout.

3 Input operators vs. input functions

Suppose that I want to read in a pair of integers into variables named k and m, and that the actual input coming up from the keyboard/file/whatever looks like

21      42

I can do that with the code

int k, m;
input >> k >> m;

This would put the value 21 into k and 42 into m. But notice that something interesting happens between the input of the two integers – we skipped over a number of blanks between the ‘1’ at the end of the first number and the ‘4’ at the beginning of the next number.

By convention, >> always starts by skipping over any whitespace characters starting from the position where we last ended reading. The “actual” input is processed starting at the first non-whitespace character that is located.