[History of C++] Explanation on why the keyword `class` has no more reason to exist

Author: Chloé Lourseyre
Editor: Peter Fordham

Introduction to the new concept : History of C++

A few months back (at the start of this blog) I was thinking about interesting things you can find in C++, then I realized one thing: the keyword class doesn’t really have a good reason to exists!

This may seem a bit harsh said that way, but I will explain my statement later in the article.

Anyway, studying the reason for the word class to exist lead me to look into the history of the language. It was very interesting and I really enjoyed it.

So I decided to write a mini-series of articles about the history of C++ aiming to present concepts that may be outdated today, in C++20, and explain why some strange or debatable things were made that way in the past.

Sources

For this miniseries, I have three main sources:

• The Design and Evolution Of C++ (Bjarne Stroustrup)
• Sibling Rivalry: C and C++ (Bjarne Stroustrup)
• A History of C++: 1979−1991 (Bjarne Stroustrup)

I have a few things to say about them.

First, Sibling Rivalry: C and C++ and A History of C++: 1979-1991 are both fully available on Stroustrup’s website (just follow the links).

Second, I find it quite sad that I could only find sources from one author. Sure, Bjarne Stroustrup is probably the best individual to talk about his own work, but I would have like the insight of other authors (if you know any, please tell me in the comments).

Why the keyword class could simply not exist in C++20?

Today, in C++20, we have two very similar keywords that work almost exactly the same: class and struct.

There is one and only one difference between class and struct: by default, the members and the inheritance of struct are public, but in those of class are private.

Here are three reason as to why this tiny difference is not worth a different keyword¹:

• In practice, the default access modifier is almost never used, in my experience. Most developers don’t use default access modifier and prefer to specify it.
• In 2021, a good code is a clear code. Explicitly writing the access specifier is -in that regard- better than leaving it implicit. This may be arguable on solo or small projects, but when you start to develop with many peers, it is better to write a few more characters to be sure that the code is clear for everyone.
• Having two keywords is more ambiguous than having one. I have very often encountered developers that think there are more differences than that between class and struct, and they have sometimes even told me what they thought these additional differences were. If there was only one keyword, this confusion wouldn’t exist.

I can hear that some people have counter-arguments. The ones I heard a lot are:

• This is syntactic sugar².
• They actually use the implicit access specifier³.
• There is semantics behind the use of each keyword that go beyond the sole technical meaning⁴.

All in all, what I’m trying to say is that the language would practically be the same if the keyword class did not exist. So, in the mindset of C++20, we could ask ourselves “what is the purpose of adding a keyword that is neither needed nor useful?”.

I know one thing: that class is one of the oldest C++-specific keywords. Let’s dive into the history of C++ to understand why it exists.

History of the keyword class

Birth

The first official appearance of the keyword class was in Classes: An Abstract Data Type Facility for the C language (Bjarne Stroustrup, 1980), and was actually not talking about C++, but about what we call C with classes.

What is “C with classes”? I think I’ll dedicate a whole article to this subject so I’ll keep it short here. It is C++ immediate predecessor, started in 1979. The original goal was to add modularity to the C language, inspired by Simula⁵ classes. At first, it was a mere tool, but soon evolved to a whole language. However, since C with Classes was a mild success while needing continuous support, Bjarne decide to abandon it and create a new language, using his experience of C with Classes, and that aimed to be more popular. He called this new language C++.

The choice of the word “class” directly comes from Simula, and the fact that Stroustrup dislikes inventing new terminology.

You’ll find more about C with Classes in the book The Design and Evolution Of C++ (Bjarne Stroustrup), where a whole section is dedicated to it.

So the keyword class was actually born within the predecessor of C++. In terms of design, it’s one of the oldest concepts and even the motivation behind the creation of C with classes.

Original difference between struct and class

In C with Classes, struct and class and quite different.

Structures works just like in C, they are simple data structure, whereas it is within classes that the concept of methods are introduced.

At that time, there was a real difference between structures and classes, thus the distinction⁶.

Into C++

The two of the greatest features of early C++ were virtual functions and functions overloading.

In addition to that, namespaces rules where introduced to define on how scopes names would behave. Among those rules:

• Names are private unless they are explicitly declared public.
• A class is a scope (implying that classes nest properly).
• C structure names don’t nest (even when they are lexically nested).

These rules make structures and classes behave differently in terms of scopes and names.

For instance, this was legal:

struct outer {
    struct inner {
        int i;
    };
};

struct inner a = { 1 };

But replacing struct with class provoked a compilation error.

In later C++, the code above doesn’t compile (it needs to be outer::inter a = {1};).

“Fusion” with the keyword struct

It’s difficult to say when this occurred specifically, because none of the sources I found clearly state “This is when structures and classes became the same concept, but we can investigate.

According to The C++ Programming Language – Reference Manual (Bjarne Stroustrup, 1984), the first published document specifically about C++:

(listing derived types)

classes containing a sequence of objects of various types, a set of functions for manipulating these objects, and a set of restrictions on the access of these objects and function;

structures which are classes without access restrictions;

Bjarne Stroustrup, The C++ Programming Language – Reference Manual, §4.4 Derived types

Moreover, if we take a look at the feedback Stroustrup gives about virtual functions and the object layout model (concepts introduced in 1986):

At his point, the object model becomes real in the sense that an object is more than the simple aggregation of the data members of a class. […] Then why did I not at this point choose to make structs and classes different notions?

My intent was to have a single concept: a single set of layout rules, a single set of lookup rules, a single set of resolution rules, etc. […]

Bjarne Stroustrup, The Design and Evolution of C++, §3.5.1 The Object Layout Model

Even though it seems that structures couldn’t hold private members at that time (the private keyword didn’t even exist then!), we can safely say this was the moment where structures and classes were “fused”.

It was at the creation of C++ that the structures from C and the classes from Simula merged together.

But when did they actually become the same?

But technically, structs and classes became what they are today the moment the keyword private was invented. It seems that happened the same time the keyword protected was introduced, for Release 1.2 in 1987.

From then to now

Despite all that, despite the fact that class is technically useless, there is a lot more to it.

The keyword class has acquired semantics.

Indeed, nowadays, writing the keyword class means that you are implementing a class that is not a data bucket, whereas the keyword struct is mainly used for data buckets and similar data structures. In their technicity, these words do not differ, but because of usage, because these keywords have history, they acquired more meaning than the sole technical meaning.

Go see Jonathan Boccara’s very relevant article: The real difference between struct and class – Fluent C++ (fluentcpp.com) for more details. This article is inspired from the Core Guidelines.

The fact that nowadays’ class has lived more than forty years makes it very different from the class from 1980 and the class that it would be if it were introduced in 2020.

But the question that immediately pops up in my head is the following: should we continue to use class this way? Should we stick to the semantics it acquired or should we seek evolution towards a more modern meaning? The answer is simple: it’s up to each of us. We, C++ devs, are the ones that make the language evolve, every day, in every single line of code.

The Core Guidelines tells us how we should use each feature today, but maybe tomorrow someone (you?) will come up with a better, safer, clearer way to code. Understanding what were structs and classes in the past and what they are today is the first step to define what they will be tomorrow.

Wrapping up

The best way to summarize the history of these two keyword is this way: “The structures from C and the classes from Simula merged together at the creation of C++.”, but we can also say that thanks to that, despite representing the same feature, the both have different meaning.

This article is not a pamphlet against class and I will not conclude this article with an half-educated half-authoritative argument like I often do⁷. Instead, I will tell you that I realized how important it is to contextualize articles like the ones I publish on this blog with the version of C++ that is used and the articles and books that are used as examples and inspirations.

I think it s important to understand history to be able to judge the practices we have today. Do we do something out of habit or is there a real advantage to it? It s a question that needs to be asked every day, or else we’ll end up writing outdated code in an outdated mindset.

How C++ developers think⁸ evolves from decade to decade. Each eon, developers have different mindsets, different goals, different issues, a different education and so on… I don’t blame how people coded in the past, but I do blame those who code now like we coded a decade ago, and in the future I hope to see my peers blaming me when I code in “old C++”.

Thanks for reading and see you next week!

Author: Chloé Lourseyre
Editor: Peter Fordham

Addenda

To go further

Here are two thoughts that are a bit off-topic.

Namespaces in C

Something that wasn’t inherited from C into C++ was the struct namespace.

Indeed, in C, the namespace containing the struct names isn’t the same as the gobal C namespace. In C struct foo and foo aren’t refereeing to the same object. In C++, assuming that foo is a structure, struct foo and foo are the same name.

There is a way, in C, to link these two namespace, using typedef. To learn more about that, read this article: How to use the typedef struct in C (educative.io)

The class keyword in templates

Maybe have you already seen this syntax:

template <class C_>
void foo(C_ arg)
{
    // ...
}

What does class mean in this context?

It actually means nothing more that “a type, whatever which one”.

This is a bit confusing, because we may think that, as type, C_ is supposed to be a class. But it is not. Later, the typename keyword was introduced to lift this confusion, but we can still use class if we want to.

Annotations

1. Since struct and class are so similar, I choose to consider class to be the keyword in excess, simply because struct exists in C and not class, and that it is the process of the keyword class that brought them both so close.

2. This argument is not true. By essence, syntactic sugar is supposed to make the code easier to read or write. Since struct/class is just a substitute of a keyword for another, there is no gain in clarity whatsoever. The only reason the code may seem easier to read is because of the semantics these keyword hold, but not the syntax itself.

3. Yes, I know, some people actually use the implicit. I do too. But what you forget when you say such a thing is that most of the software industry doesn’t think like you nor write code like you. The statement of the struct/class duplicity comes from an empiric fact. Our own individual practices can never be an argument against that fact.

4. While this is factually true, the reasoning is upside-down. It’s because of their history that they have different semantic meaning. If they were created today, from nothing, they would not have that semantics, only their duplicity. I’ll talk about that near the end of the article.

5. Simula is the name of two simulation programming languages, Simula I and Simula 67, developed in the 1960s at the Norwegian Computing Center in Oslo, and is considered the first object-oriented programming language. Simula is very unknown amongst the community of developers, but has greatly influenced other famous languages. Simula-type objects are reimplemented in Object Pascal, Java, C#, and, of course, C++.

6. At that time, you could emulate any structure with a class, but it was still interesting, all the more in the mindset of then, to make the distinction.

7. I tend to always agree with the C++ Core Guidelines (isocpp.github.io), even if I am always trying to be critical as of our habits and practices. But keep in mind that the guidelines we have today may be different from the one we’ll have tomorrow.

8. I think this statement is actually true for every languages, but C++ is the perfect archetype as it is one of the oldest most-used language on the market, today in 2021.

Sources

In order of appearance:

• The Design and Evolution Of C++ (Bjarne Stroustrup), 1994
• Sibling Rivalry: C and C++ (Bjarne Stroustrup), 2002
• A History of C++: 1979−1991 (Bjarne Stroustrup), 1993
• The C++ Programming Language – Reference Manual (Bjarne Stroustrup), 1984
• The real difference between struct and class – Fluent C++ (Jonathan Boccara), 2017
• C++ Core Guidelines (isocpp.github.io)

Off-topic: Feedspot’s Top 30 C++ Programming Blogs and Websites

Recently, Feedspot made a top-list of the best C++ programming blogs and website, and Belay the C++ ended up in 9th position!

You can find the top-list here: Top 30 C++ Programming Blogs and Websites You Must Follow in 2021 (feedspot.com)