Last chapter we learned how computers do everything in binary… but that’s just a bunch of numbers. It seems like a far cry from the complex information we store in computers these days. How does something like a game, or a song, or a slide show get turned into a bunch of numbers? This is a big topic that we’ll discuss later on, but the basic idea is not too difficult.

Let’s start with eight bytes. Remember, computers work with bytes rather than bits for convenience. To keep things short, we’ll write things in hexadecimal.

00 00 00 00 00 00 00 00

Right now the bytes are all zero, but we want to store information in these bytes. Eight bytes is 64 bits. That means we can store values from 0 to 264 − 1. But can we really use all eight bytes? Real computers have much more memory than just eight bytes; they can have hundreds of thousands of bytes. With thousands of bytes around, how do you tell the computer that we only want to use eight? One way would be to agree that the first byte says how many bytes we’re grouping together:

08 00 00 00 00 00 00 00

We have seven bytes leftover to actually store information, so we can store values up to 256 − 1. Let’s store a nice big number in there. How about 4,240,876,579,651,596? That fills out the last seven bytes nicely:

08 0F 11 0E 06 11 00 0C

But what if we want to store something other than numbers? We could store words by agreeing that each byte will be one letter. A is 00, B is 01, C is 02, etc. Then we could store up to seven letters with these bytes. Let’s store the word “program”:

08 0F 11 0E 06 11 00 0C

Huh. Now we have a problem. How does the computer know whether we are storing the word “program” or the value 4,240,876,579,651,596? Well let’s make another agreement. After the number of bytes, we’ll devote the second byte to describing what kind of information we’re storing. For example we could agree that 00 means it’s a number, 01 means it’s letters, 02 could be a picture, etc. We only have six bytes left, but at least now we can tell the difference between “turing” and 20,976,906,013,958:

08 00 13 14 11 08 0D 06
08 01 13 14 11 08 0D 06

You see? By grouping bytes in different ways and agreeing on the meanings of different bytes, we can store a variety of information just using numbers. These agreements have a technical term:

A format groups bytes together and assigns specific meanings to certain bytes and byte values.

Size Matters

As I mentioned before, computers can have many, many bytes. To deal with this, we have easier ways to refer to the huge numbers of bytes that come into play when it comes to data. Bytes use the standard metric prefixes:

Terabytes are becoming more common these days, but it will probably be a few years before the next prefix (peta) becomes common. There are some complications, unfortunately. As you might recall from the previous chapter, humans really like decimal but computers like binary. So while it is natural for people to use kilo, mega, and giga, etc. to mean 10001, 10002, and 10003, it is common for computer scientists to use powers of 1024 = 210. In other words, sometimes bytes are measured like this:

It is confusing. And it gets worse. Recall how the formatting of the data in the previous section reduced the number of bytes we could use to store information. Even though we had eight bytes, we could only use six after the formatting was done. Computer components are often advertised as storing such-and-such gigabytes of data—but is that using powers of 10 or powers of 2? Is that before formatting or after formatting? One often needs to read the fine print when it comes to measuring data.

Limitations

We’ve taken a look now at how computers compute and how they store information, but so far every example has boiled down to integers and counting. But aren’t there other kinds of math? Aren’t there are other kinds of information? What about numbers like π? What about information like music and paintings? How can we represent such things using a computer?

The truth, surprisingly, is that we can’t—because computers deal exclusively with discrete data.

Discrete data can be counted.

For example, take the number of students in a class. It doesn’t make sense to have only part of a student in a class, you can only change the number of students by whole numbers. Letters are similarly discrete, since a letter must be one of the letters of the alphabet; a letter cannot be somewhere between A and B. In contrast, lots of things we perceive in real life are continuous.

Continuous data can only be measured.

For example, your height is continuous since you can’t suddenly switch from being one height to another. The light that we see and the sound waves we hear are also continuous since they can transition smoothly from one state to another. A good test for something being discrete versus continuous is whether the precision of the measurement matters. While you can count the number of students in a class and be 100% sure of your answer, your height depends on what tool you use to measure it and how precisely you record the value. If you use a more precise tool on continuous data, your answer can change.

As you might expect, these limitations on data translate over to the types of information we can store with data.

The words “data” and “information” are used almost interchangeably these days, but there’s a subtle difference between them. To put it simply, information has meaning whereas data are just cold, hard facts.

In a computer, the bits and bytes are data. The interpretations of those data create information.

Discrete data can only produce digital information.

Only continuous data can produce analog information.

Continuous vs. Discrete

This surprising distinction is often misunderstood by those unfamiliar with computers. The distinction implies that if we want to store any continuous data in our computer, or have the computer produce any continuous data, we need to be able to convert between digital and analog information. Technically, this is impossible. Instead we make do by approximating analog information with digital information and then later reconstructing analog information from the digital approximation.

Imagine a perfect circle. No matter how far you zoom in to the edge, it’s still a perfectly smooth curve. The circle represents continuous data. Now imagine that I asked you to make a copy of that circle, but you can only use squares. The squares represent discrete data. You could try to fill in the outline of the circle with squares. The result might have gaps or corners, but if the squares are small enough, the result will look a bit like a circle. If I then asked you to recreate the circle based on your copy, you could try to trace around the edges of the squares. The quality of your result would depend on how well your squares approximated the original circle.

Like fitting a square peg into a round hole, the process is by definition imperfect. A huge amount of effort is expended by computer scientists and engineers to improve our digital approximations of continuous data.

Possibly because of the complexity that formats can require to store information, formats are sometimes thought of as “codes”. Just like our DNA encodes the information for our living bodies, formats encode the information we store in our computers.

Information can be encoded to store it in formatted data. The data can be decoded to get the information back again.

Exercises

  1. So far, all of the calculations we’ve toyed around with have used positive integers. Those are easy because it’s rather obvious how to represent a positive integer using bits and bytes. Come up with a format for storing negative integers as well as positive ones. Then try to make a format for storing numbers with a decimal point e.g. 3.14 or −1.0007.
  2. Come up with three examples of digital information and three examples of analog information.
  3. Computer images are usually displayed using additive color. That means that each color is represented as a mixture of the three primary additive colors: red, green, and blue. Create a format for storing a color in a computer using the additive model. What choices do you get to make when designing this format? Do any of your choices impose limitations on the colors stored using your format? Can you get around these limitations?