BookPDF Available

Practical C++ Metaprogramming

Edouard Alligand and Joel Falcou
Practical C++
Modern Techniques for
Accelerated Development
Boston Farnham Sebastopol Tokyo
Beijing Boston Farnham Sebastopol Tokyo
Practical C++ Metaprogramming
by Edouard Alligand and Joel Falcou
Copyright © 2016 O’Reilly Media, Inc. All rights reserved.
Printed in the United States of America.
Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA
O’Reilly books may be purchased for educational, business, or sales promotional use.
Online editions are also available for most titles ( For
more information, contact our corporate/institutional sales department:
800-998-9938 or
Editors: Nan Barber and Brian Foster
Production Editor: Colleen Lobner
Copyeditor: Octal Publishing, Inc.
Proofreader: Rachel Head
Interior Designer: David Futato
Cover Designer: Randy Comer
Illustrator: Rebecca Demarest
September 2016: First Edition
Revision History for the First Edition
2016-09-13: First Release
The O’Reilly logo is a registered trademark of O’Reilly Media, Inc. Practical C++
Metaprogramming, the cover image, and related trade dress are trademarks of
O’Reilly Media, Inc.
While the publisher and the authors have used good faith efforts to ensure that the
information and instructions contained in this work are accurate, the publisher and
the authors disclaim all responsibility for errors or omissions, including without
limitation responsibility for damages resulting from the use of or reliance on this
work. Use of the information and instructions contained in this work is at your own
risk. If any code samples or other technology this work contains or describes is sub‐
ject to open source licenses or the intellectual property rights of others, it is your
responsibility to ensure that your use thereof complies with such licenses and/or
Table of Contents
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
A Misunderstood Technique 1
What Is Metaprogramming? 3
How to Get Started with Metaprogramming 6
Summary 8
2. C++ Metaprogramming in Practice. . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
A Typical Code Maintenance Assignment 9
Creating a Straightforward Interface 10
Generating Code Automatically 13
Making Values and Pointers Work Together 13
Putting It All Together 25
Summary 26
3. C++ Metaprogramming and Application Design. . . . . . . . . . . . . . . . 27
Compile-Time Versus Runtime Paradigms 27
Type Containers 30
Compile-Time Operations 31
Advanced Uses of Metaprogramming 40
Helper Functions and Libraries 42
Summary 43
Another arcane text about an overly complex language! C++ is
already difficult enough to master; why do people feel the need to
make it even more difficult?
C++’s power comes at a price, but with the latest revisions of the
language, the bar has been drastically lowered. The improvements in
C++11 and C++14 have had a positive impact in many areas, from
how you write a loop to how you can write templates.
We’ve had the idea of writing about template metaprogramming for
a long time, because we wanted to demonstrate how much easier it
has become. We also wanted to prove its usefulness and efficiency.
By that we mean that it’s not only a valid solution, but sometimes
the best solution.
Last but not least, even if you don’t use metaprogramming every day,
understanding its concepts will make you a better programmer: you
will learn to look at problems differently and increase your mastery
and understanding of the language.
A Journey of a Thousand Miles Begins with a
Single Step
Really mastering C++ metaprogramming is difficult and takes a lot
of time. You need to understand how compilers work to get around
their bugs and limitations. The feedback you can receive when you
have an error is more often than not arcane.
That is the bad news.
The good news is that you don’t need to master C++ metaprogram‐
ming, because you are standing on the shoulders of giants.
In this report, we will progressively expose you to the technique and
its practical applications, and give you a list of tools that you can use
to get right to it.
Then, depending on your tastes and your aspirations, you can
decide how deep down the rabbit hole you want to go.
Understanding Metaprogramming
Metaprogramming is a technique that can greatly increase your pro‐
ductivity when properly used. Improperly used, though it can result
in unmaintainable code and greatly increased development time.
Dismissing metaprogramming based on a preconceived notion or
dogma is counterproductive. Nevertheless, properly understanding
if the technique suits your needs is paramount for fruitful and
rewarding use.
An analogy we like to use is that you should see a metaprogram as
a robot you program to do a job for you. After you’ve programmed
the robot, it will be happy to do the task for you a thousand times,
without error. Additionally, the robot is faster than you and more
If you do something wrong, though, it might not be immediately
obvious where the problem is. Is it a problem in how you program‐
med the robot? Is it a bug in the robot? Or is your program correct
but the result unexpected?
That’s what makes metaprogramming more difficult: the feedback
isn’t immediate, and because you added an intermediary you’ve
added more variables to the equation.
That’s also why before using this technique you must ensure that
you know how to program the robot.
viii | Preface
Conventions Used in This Report
The following typographical conventions are used in this report:
Indicates new terms, URLs, email addresses, filenames, and file
Constant width
Used for program listings, as well as within paragraphs to refer
to program elements such as variable or function names, data‐
bases, data types, environment variables, statements, and key‐
This element signifies a general note.
This element indicates a warning or caution.
This report would probably not exist without the work of Aleksey
Gurtovoy and David Abrahams, authors of the Boost.MPL library
and the reference book C++ Template Metaprogramming (Addison-
Wesley Professional).
More recently, Eric Niebler and Peter Dimov paved the way to what
modern C++ template metaprogramming should look like. They
have been greatly influential in our work.
We would also like to thank all of the contributors to the Brigand
library and Louis Dionne for his metaprogramming library bench‐
Finally, we would like to thank Jon Kalb and Michael Caisse for their
reviews, as well as our families, friends, and coworkers, who have
been incredibly supportive.
Preface | ix
If you grabbed this report, it means that you have at least curios‐
ity about C++ metaprogramming, a topic that often generates out‐
right rejection.
Before we talk about template metaprogramming, let’s ask ourselves
a question: why do we violently reject some techniques, even before
studying them?
There are, of course, many valid reasons to reject something new,
because, let’s be frank, sometimes concepts are just plain nonsense
or totally irrelevant to the task at hand.
However, there is also a lot to be said about managing your own
psychology when accepting novelty, and recognizing our own men‐
tal barriers is the best way to prevent them from growing.
The purpose of this report is to demonstrate that understanding
C++ metaprogramming will make you a better C++ programmer, as
well as a better software engineer in general.
A Misunderstood Technique
Like every technique, we can overuse and misunderstand metaprog‐
ramming. The most common reproaches are that it makes code
more difficult to read and understand and that it has no real benefit.
As you progress along the path of software engineering, the techni‐
ques you learn are more and more advanced. You could opt to rely
solely on simple techniques and solve complex problems via a com‐
position of these techniques, but you will be missing an opportunity
to be more concise, more productive, and sometimes more efficient.
Imagine that you are given an array and that you need to fill it with
increasing integers. You could write the following function:
void f(int * p, size_t l)
for(size_t i = 0; i < l; ++i)
p[i] = i;
// ...
int my_array[5];
f(my_array, 5);
Or you could use the Standard Template Library (STL):
int my_array[5];
std::iota(my_array, my_array + 5, 0);
The generated assembly code may be equivalent, if not identical, yet
in the latter case, because you learned about an STL function, you
gained both in productivity and in information density. A program‐
mer who doesn’t know about the iota function just needs to look
it up.
What makes a good software engineer is not only the size of his
toolbox, but, most importantly, his ability to choose the right tool at
the right time.
C++ metaprogramming has, indeed, an unusual syntax, although it
has significantly improved in the past few years with the release of
C++11 and C++14.
The concepts behind C++ metaprogramming, on the other hand,
are extremely coherent and logical: it’s functional programming!
That’s why on the surface it might look arcane, but after you are
taught the underlying concepts it all makes sense. This is something
we will see in more depth in Chapter 3.
In this report, we want to expose you to C++ metaprogramming in a
way that is intelligible and practical.
2 | Chapter 1: Introduction
When you are finished, we hope that you that will agree with us that
it is both useful and accessible.
What Is Metaprogramming?
By definition, metaprogramming is the design of programs whose
input and output are programs themselves. Put another way, it’s
writing code whose job is to write code itself. It can be seen as the
ultimate level of abstraction, as code fragments are actually seen as
data and handled as such.
It might sound esoteric, but its actually a well-known practice. If
you’ve ever written a Bash script generating C files from a boiler‐
plate file, you’ve done metaprogramming. If you’ve ever written C
macros, you’ve done metaprogramming. In another sphere, you
could debate whether generating Java classes from a UML schema is
not actually just another form of metaprogramming.
In some way, you’ve probably done metaprogramming at various
points in your career without even knowing it.
The Early History of Metaprogramming
Throughout the history of computer science, various languages have
evolved to support different forms of metaprogramming. One of the
most ancient is the LISP family of languages, in which the program
itself was considered a piece of data and well-known LISP macros
were able to be used to extend the languages from within. Other lan‐
guages relied on deep reflection capabilities to handle such tasks
either during compile time or during runtime.
Outside of the LISP family, C and its preprocessor became a tool of
choice to generate code from a boilerplate template. Beyond classical
function-like macros, the technique known as X-macros was a very
interesting one. An X-macro is, in fact, a header file containing a list
of similar macro invocations—often called the components—which
can be included multiple times. Each inclusion is prefixed by the
redefinition of said macro to generate different code fragments for
the same list of components. A classic example is structure serializa‐
tion, wherein the X-macros will enumerate structure members, first
during definition and then during serialization:
// in components.h
PROCESS(float, x )
What Is Metaprogramming? | 3
1See the documentation for details.
2The original code is available (in German) at
PROCESS(float, y )
PROCESS(float, z )
PROCESS(float, weight )
// in particle.c
typedef struct
#define PROCESS(type, member) type member;
#include "components.h"
#undef PROCESS
} particle_t;
void save(particle_t const* p, unsigned char* data)
#define PROCESS(type, member) \
memmove(data, &(p->member), sizeof(p->member)); \
data += sizeof(p->member); \
#include "components.h"
#undef PROCESS
X-macros are a well-tested, pure C-style solution. Like a lot of C-
based solutions, they work quite well and deliver the performance
we expect. We could debate the elegance of this solution, but con‐
sider that a very similar yet more automated system is available
through the Boost.Preprocessor vertical repetition mechanism,
based on self-referencing macros.1
Enter C++ Templates
Then came C++ and its generic types and functions implemented
through the template mechanism. Templates were originally a very
simple feature, allowing code to be parameterized by types and inte‐
gral constants in such a way that more generic code can emerge
from a collection of existing variants of a given piece of code. It was
quickly discovered that by supporting partial specialization,
compile-time equivalents of recursion or conditional statements
were feasible. After a short while, Erwin Unruh came up with a very
interesting program2 that builds the list of every prime number
4 | Chapter 1: Introduction
3Substitution failure is not an error.
between 1 and an arbitrary limit. Quite mundane, isn’t it? Except
that this enumeration was done through warnings at compile time.
Let’s take a moment to ponder the scope of this discovery. It meant
that we could turn templates into a very crude and syntactically
impractical functional language, which later would actually be pro‐
ven by Todd Veldhuizen to be Turing-complete. If your computer
science courses need to be refreshed, this basically meant that, given
the necessary effort, any functions computable by a Turing machine
(i.e., a computer) could be turned into a compile-time equivalent by
using C++ templates. The era of C++ template metaprogramming
was coming.
C++ template metaprogramming is a technique based on the use
(and abuse) of C++ template properties to perform arbitrary com‐
putations at compile time. Even if templates are Turing-complete,
we barely need a fraction of this computational power. A classic ros‐
ter of applications of C++ template metaprogramming includes the
Complex constant computations
Programmatic type constructions
Code fragment generation and replication
Those applications are usually backed up by some libraries, like
Boost.MPL or Boost.Fusion, and a set of patterns including tag dis‐
patching, recursive inheritance, and SFINAE.3 All of those compo‐
nents thrived in the C++03 ecosystem and have been used by a large
number of other libraries and applications.
The main goal of those components was to provide compile-time
constructs with an STL look and feel. Boost.MPL is designed around
compile-time containers, algorithms, and iterators. In the same way,
Boost.Fusion provides algorithms to work both at compile time and
at runtime on tuple-like structures.
What Is Metaprogramming? | 5
For some reason, even with those familiar interfaces, metaprogram‐
ming tools continued to be used by experts and were often over‐
looked and considered unnecessarily complex. The compilation
time of metaprograms was also often criticized as hindering a nor‐
mal, runtime-based development process.
Most of the critiques you may have heard about template metaprog‐
ramming stem from this limitation—which no longer applies, as we
will see in the rest of this report.
How to Get Started with Metaprogramming
When making your first forays into metaprogramming, our advice
is to experiment on the side with algorithms and type manipula‐
tions, as we show in Chapters 2 and 3, and in actual projects, begin‐
ning with the simplest thing: static assertions.
Writing metaprograms to do compile-time checks is the safest and
simplest thing you can do when getting started. When you are
wrong, you will get a compilation-time error or a check will incor‐
rectly pass, but this will not affect the reliability or correctness of
your program in any way.
This will also get your mind ready for the day when concepts land
in C++.
Checking Integer Type
Some programs or libraries obfuscate the underlying integer type of
a variable. Having a compilation error if your assumption is wrong
is a great way to prevent difficult-to-track errors:
"invalid integer detected!");
If your obfuscated integer isn’t an unsigned 32-bit integer, your pro‐
gram will not compile and the message “invalid integer detected”
will be printed.
You might not care about the precise type of the integer—maybe just
the size is important. This check is very easy to write:
static_assert(sizeof(obfuscated_int) == 4,
"invalid integer size detected!");
6 | Chapter 1: Introduction
Checking the Memory Model
Is an integer the size of a pointer? Are you compiling on a 32-bit or
64-bit platform? You can have a compile-time check for this:
static_assert(sizeof(void *) == 8, "expected 64-bit platform");
In this case, the program will not compile if the targeted platform
isn’t 64-bit. This is a nice way to detect invalid compiler/platform
We can, however, do better than that and build a value based on the
platform without using macros. Why not use macros? A metapro‐
gram can be much more advanced than a macro, and the error out‐
put is generally more precise (i.e., you will get the line where you
have the error, whereas with preprocessor macros this is often not
the case).
Let’s assume that your program has a read buffer. You might want
the value of this read buffer to be different if you are compiling on a
32-bit platform or a 64-bit platform because on 32-bit platforms you
have less than 3 GB of user space available.
The following program will define a 100 MB buffer value on 32-bit
platforms and 1 GB on 64-bit platforms:
static const std::uint64_t default_buffer_size =
std::conditional<sizeof(void *) == 8,
std::integral_constant<std::uint64_t, 100 * 1024 * 1024>,
std::integral_constant<std::uint64_t, 1024 * 1024 * 1024>
Heres what the equivalent in macros would be:
#ifdef IS_MY_PLATFORM_64
static const std::uint64_t default_buffer_size
= 100 * 1024 * 1024;
static const std::uint64_t default_buffer_size
= 1024 * 1024 * 1024;
The macros will silently set the wrong value if you have a typo in the
macro value, if you forget a header, or if an exotic platform on
which you compile doesn’t have the value properly defined.
Also, it is often very difficult to come up with good macros to detect
the correct platform (although Boost.Predef has now greatly
reduced the complexity of the task).
How to Get Started with Metaprogramming | 7
Things changed with the advent of C++11 and later C++14, where
new language features like variadic lambdas, constexpr functions,
and many more made the design of metaprograms easier. This
report will go over such changes and show you how you can now
build a metaprogramming toolbox, understand it, and use it with far
greater efficiency and elegance.
So, let’s dive in headfirst—we’ll start with a short story that demon‐
strates what you can do.
8 | Chapter 1: Introduction
C++ Metaprogramming
in Practice
Let’s imagine that you are responsible for the construction—from
the ground up—of a brand new module in a big weather prediction
system. Your task is to take care of the distribution of complex com‐
putations on a large computing grid, while another team has the
responsibility for the actual computation algorithms (in a library
created two decades previously).
We will see in this chapter what kinds of problems arise when you
try to interface two bricks that were created 20 years apart, examine
the typical approaches, and see if the template metaprogramming
approach brings any benefit.
A Typical Code Maintenance Assignment
After two years of development, your distributed weather system is
at last done! You’ve been very thorough in applying modern C++
principles all along, and took advantage of pass-by-value every‐
where you could. You are happy with the performance, the software
is now stable, and you’ve made the design as sound as possible given
the time you had.
But now, you need to interface with “the Thing,” aka “The Simula‐
tion Library of Awesomeness,” or SLA for short.
The SLA was designed in the 1990s by developers who have now
gone insane or missing. Every time you install the SLA on a system,
it is no longer possible to run any other kind of software without
having a team of senior system administrators perform a week-long
ritual to cleanse the machine.
Last but not least, the SLA only believes in one god, and that god
is e Great Opaque Pointer. All interfaces are made as incoherent as
possible to ensure that you join the writers in an unnamable crazy
laughter, ready to be one with The Great Opaque Pointer.
If you didn’t have several years of experience up your sleeve, you
would advocate a complete rewrite of the SLA—but you know
enough about software engineering to know that “total rewrite” is
another name for “suicide mission.
Are we dramatizing? Yes, we are. But let’s have a look at a function
of the SLA:
// we assume alpha and beta to be parameters to the mathematical
// model underlying the weather simulation algorithms--any
// resemblance to real algorithms is purely coincidental
void adjust_values(double * alpha1,
double * beta1,
double * alpha2,
double * beta2);
Now let’s have a look at how you designed your application:
class reading
/* stuff */
double alpha_value(location l, time t) const;
double beta_value(location l, time t) const;
/* other stuff */
Let us not try to determine what those alpha and beta values are,
whether the design makes sense, or what exactly adjust_values
does. What we really want to see is how we adapt two pieces of soft‐
ware that have very different logic.
Creating a Straightforward Interface
Interfacing your software with other software is part of your job. It is
easy to mock the lack of logic or cleanliness of a program that has
been running and maintained for 25 years, but at the end of the day,
it must work; no excuses.
10 | Chapter 2: C++ Metaprogramming in Practice
In this case, you might be tempted to take a pragmatic approach and
just interface functions as needed, with a wrapper like this:
std::tuple<double, double, double, double> get_adjusted_values(
const reading & r,
location l, time t1, time t2)
double alpha1 = r.alpha_value(l, t1);
double beta1 = r.beta_value(l, t1);
double alpha2 = r.alpha_value(l, t2);
double beta2 = r.beta_value(l, t2);
adjust_values(&alpha1, &beta1, &alpha2, &beta2);
return std::make_tuple(alpha1, beta1, alpha2, beta2);
The std::tuple<> pattern
You can see that we use a tuple to “return a bunch of
otherwise unrelated stuff.” This is a common pattern in
modern C++, and later you will see why using tuples
has some advantages when it comes to metaprogram‐
But if we look again at the manual approach, we can see a certain
number of issues:
It’s error prone because of the interface of the library we are
working with. Tests can catch some but not all of these bugs.
The code is very repetitive; for a couple of functions, it is doa‐
ble, but a hundred? Or a thousand?
How do you maintain the code? If there are changes in any of
the functions, maintenance costs will grow exponentially.
What if the names of the functions change? What if the object
changes? What if the methods change?
Creating a Straightforward Interface | 11
You could retort, “Fine, let’s make it generic,” as shown here:
template <typename Reading>
std::tuple<double, double, double, double> get_adjusted_values(
const Reading & r,
location l, time t1, time t2)
double alpha1 = r.alpha_value(l, t1);
double beta1 = r.beta_value(l, t1);
double alpha2 = r.alpha_value(l, t2);
double beta2 = r.beta_value(l, t2);
adjust_values(&alpha1, &beta1, &alpha2, &beta2);
return std::make_tuple(alpha1, beta1, alpha2, beta2);
Sure, it’s an improvement, but not a big improvement. To which
you will reply, “Fine, let’s make the methods generic!” as in this
template <typename AlphaValue, typename BetaValue>
std::tuple<double, double, double, double> get_adjusted_values(
AlphaValue alpha_value, BetaValue beta_value,
location l, time t1, time t2)
double alpha1 = alpha_value(l, t1);
double beta1 = beta_value(l, t1);
double alpha2 = alpha_value(l, t2);
double beta2 = beta_value(l, t2);
adjust_values(&alpha1, &beta1, &alpha2, &beta2);
return std::make_tuple(alpha1, beta1, alpha2, beta2);
And you would call the function as follows:
reading r;
// some code
auto res = get_adjusted_values(
[&r](double l, double t){ return r.alpha_value(l, t); },
[&r](double l, double t){ return r.beta_value(l, t); },
/* values */);
What we will see here is how we can push this principle of reusabil‐
ity and genericity much further, thanks to template metaprogram‐
12 | Chapter 2: C++ Metaprogramming in Practice
What we want to avoid writing is all the systematic code that takes
the results from our C++ methods, puts them in the correct form
for the C function, passes them to that C function, and gets the
results in a form compatible with our C++ framework.
We can call it the boilerplate.
With template metaprogramming techniques, we will make the
compiler work for us and avoid a lot of mistakes and tedious work.
Generating Code Automatically
You may be thinking, “I can write a Python script that will generate
the code for me.” This is indeed doable, if the wrapping code isn’t
too complex and you will not require a comprehensive C++ parsing.
It will increase the complexity of building and maintaining your
application, however, because in addition to requiring a compiler,
you will now require a scripting language, probably with a certain
set of libraries. This kind of solution is another form of automation.
You might also create an abstraction around the library, or at least a
facade. You’ll still have one problem left, though: you have to write
all of the tedious code.
But… computers are very good at repetitive tasks, so why not pro‐
gram the computer to write the facade for you? Wouldn’t that greatly
increase your productivity?
Why not give it a try? In other words, let’s write a program that will
generate the program. Let’s metaprogram!
Making Values and Pointers Work Together
If we look at the problem from a higher perspective, we see that we
have on one side methods working with values, and on the other
side functions working with pointers.
The typical C++ approach for a function that takes one parameter
and returns one parameter is straightforward:
template <typename ValueFunction, typename PointerFunction>
double magic_wand(ValueFunction vf,
PointerFunction pf,
double param)
double v = vf(param);
Generating Code Automatically | 13
return v;
We take a callable, vf, that accepts a double as a parameter and
returns a double as a parameter. Because we’re using a template, we
don’t need to be specific about what exactly vf is (it can be a func‐
tion, a functor, or a method bound to an object instance).
The callable pf accepts a pointer to a double as a parameter and
updates the value. We then return the updated value.
We called that function magic_wand because its the magic wand that
makes your type problem go away!
But the problem is that we have more than one function and more
than one parameter. We therefore need to somehow guess the type
of the function, manipulate the type to correctly extract values,
pass a pointer to these values to the PointerFunction, and return
the result.
If you pause to think about it, you’ll quickly realize that we need two
Type manipulation
Being able to work on an arbitrary number of parameters and
iterate on them
In other words, we’d like to write C++ that modifies types and not
values. Template metaprogramming is the perfect tool for compile-
time type manipulations.
Let us take a look at a general case. How could we write a program
that takes a double and transforms it into a pointer to a double?
Type Manipulation 101
Since C++11, the standard library has come with a fair number of
functions to manipulate types. For example, if youd like to trans‐
form a double into a double *, you can do this:
#include <type_traits>
// double_ptr_type will be double *
using double_ptr_type = std::add_pointer<double>::type;
14 | Chapter 2: C++ Metaprogramming in Practice
And vice versa:
#include <type_traits>
// double_type will be double
using double_type = std::remove_pointer<double *>::type;
// note that removing a pointer from a nonpointer type is safe
// the type of double_too_type is double
using double_too_type = std::remove_pointer<double>::type;
These kinds of type manipulations (adding and removing pointers,
references, and constness) are basic building blocks and extremely
useful when dealing with type constraints. For example, your tem‐
plate parameter might have to be a const reference when you
actually need a value. With these tools you can ensure that your type
is exactly what you need.
A Generic Function Translator
The generic version of the magic wand can take an arbitrary num‐
ber of functions, concatenate the results into a structure, pass point‐
ers to these results to our legacy C function that will apply the
weather model, and return its output.
In other words, in pseudocode, we want something like this:
MagicListOfValues generic_magic_want(OldCFunction old_f,
ListOfFunctions functions,
ListOfParameters params)
MagicListOfValues values;
/* wait, something is wrong, we can't do this
for(auto f : functions)
return values;
The only problem is that we can’t do that.
Making Values and Pointers Work Together | 15
Why? The first problem is that we need a collection of values, but
those values might have heterogeneous types. Granted, in our exam‐
ple we return doubles and we could use a vector.
The other problem is a performance issue—why resize the collection
at runtime when you know exactly its size at compile time? And why
use the heap when you can use the stack?
That’s why we like tuples. Tuples allow for heterogeneous types to be
stored, their size is fixed at compile time, and they can avoid a lot of
dynamic memory allocation.
That raises some questions, though. How do we build these tuples
based on the parameters of our legacy C function? How do we iter‐
ate on a tuple? How do we work on the list of functions? How do we
pass parameters?
Extracting the C Function’s Parameters
The first step of the process is, for a given function F, to build a tuple
matching the parameters.
We will use the pattern matching algorithms of partial template spe‐
cialization to do that:
template <typename F>
struct make_tuple_of_params;
template <typename Ret, typename... Args>
struct make_tuple_of_params<Ret (Args...)>
using type = std::tuple<Args...>;
// convenience function
template <typename F>
using make_tuple_of_params_t =
typename make_tuple_of_params<F>::type;
The Magic ... Operator
In C++11, the semantics of the ... operator have been
changed and greatly extended to enable us to say to the
compiler, “I expect a list of types of arbitrary length.” It
has no relationship anymore with the old C ellipsis
operator. This operator is a pillar of modern C++ tem‐
plate metaprogramming.
16 | Chapter 2: C++ Metaprogramming in Practice
With our new function, we can therefore do the following:
template <typename F>
void magic_wand(F f)
// if F is in the form void(double *, double *)
// make_tuple_of_params is std::tuple<double *, double *>
make_tuple_of_params_t<F> params;
// ...
We now have a tuple of params we can load with the results of our
C++ functions and pass to the C function. The only problem is that
the C function is in the form void(double *, double *, double
*, double *), and we work on values.
We will therefore modify our make_tuple_of_params functor
template <typename Ret, typename... Args>
struct make_tuple_of_derefed_params<Ret (Args...)>
using type = std::tuple<std::remove_ptr_t<Args>...>;
Hey! What’s Going On with the ... Operator?!
When we wrote Args..., we somehow expanded the list of param‐
eters inside our std::tuple. That’s one of the most straightforward
uses of the operator.
The general behavior of the ... operator is to replicate the code
fragment on its left for every type in the parameter pack. In this
case, the remove_ptr_t will be carried along.
For example, if your arguments are:
int i, double d, std::string s
expansion with std::tuple<Args...> will yield:
std::tuple<int, double, std::string>
and expansion with std::( tuple <std::( add_(
pointer_t<Args>...> will yield:
std::tuple<int *, double *, std::string *>
Making Values and Pointers Work Together | 17
Now the function works as follows:
template <typename F>
void magic_wand(F f)
// if F is in the form void(double *, double *)
// make_tuple_of_params is std::tuple<double, double>
make_tuple_of_derefed_params<F> params;
// ...
We just need to load up the results!
Getting a List of Functions and Parameters
Now that we can extract the contents of the C functions parameters,
we need to assemble them in objects that we can manipulate easily
in C++.
Indeed, you might be tempted to write this:
template <typename Functions, typename Params>
void magic_wand(/* stuff */, Functions... f, Params... p)
// stuff
After all, you have a list of functions and a list of parameters, and
you want both of them. The only problem is, how can the compiler
know when the first list ends and the second list begins?
Again, tuples come to the rescue:
template <typename... Functions, typename... Params>
void magic_wand(/* stuff */,
const std::tuple<Functions...> & f,
const std::tuple<Params...> & p1,
const std::tuple<Params...> & p2)
// stuff
This enables the compiler to know that multiple tuples of arbitrary
and unrelated lengths are expected. You could, of course, make a
tuple of tuples if you expect more than two sets of parameters, but
there’s no need to make our example more complex than it needs
to be.
18 | Chapter 2: C++ Metaprogramming in Practice
Performance Warning
Although compilers are getting very good at removing
unnecessary copies, and rvalue references help with
moving objects, be mindful of what you put inside
your tuples and how many of them you create.
Passing the values, in our example, becomes the following:
magic_wand(/* stuff */,
// our C++ functions
[&r](double l, double t){ return r.alpha_value(l, t); },
[&r](double l, double t){ return r.beta_value(l, t); }),
// first set of params
std::make_tuple(l, t1),
// second set of params
std::make_tuple(l, t2));
Which means that inside the body of the magic_wand function, we
will have tuples containing the functions we need to call and the
parameters we need to pass to them.
Filling the Values for the C Function
We’ve progressed, but we have not arrived. On one hand we have
tuples of values to pass to the C function; on the other hand, we
have a tuple of functions and parameters.
We now want to fill the tuple of values with the results, which means
calling every function inside the tuple and passing the correct
template <typename LegacyFunction,
typename... Functions,
typename... Params>
auto magic_wand(
LegacyFunction legacy,
const std::tuple<Functions...> & functions,
const std::tuple<Params...> & params1,
const std::tuple<Params...> & params2)
make_tuple_of_derefed_params_t<LegacyFunction> params = {
/* we would like to do
for(auto f : functions)
for(auto f : functions)
Making Values and Pointers Work Together | 19
// rest of the code
Returning auto
In C++14 you don’t need to be explicit about the
return type of a function; the type can be determined
at compile time contextually. Using auto in this case
greatly simplifies the writing of generic functions.
In template metaprogramming, there is no iterative construct. You
can’t iterate on your list of types by using for. You can, however, use
recursion to apply a callable on every member of the tuple. This
approach has been used since 2003 to great effect, but it has the dis‐
advantage of generating a huge amount of intermediate types and
therefore increases compilation time.
Whenever you can, you should use the ... operator to apply a calla‐
ble to every member of a list. This is faster, it doesn’t generate all the
unneeded intermediate types, and the code is often more concise.
How can we use the ... operator for that? Here, we will create a
sequence that matches the size of the tuple in order to apply a func‐
tor to each member:
template <typename F, typename Params, std::size_t... I>
auto dispatch_params(F f,
Params & params,
return f(std::get<I>(params)...);
What happens here is the following:
template <typename F, typename Params, std::size_t... I>
auto dispatch_params(F f,
Params & params,
// not real C++ code
return f(std::get<0>(params),
20 | Chapter 2: C++ Metaprogramming in Practice
std::get<N>(params)); // where N is the last index
The advantage is that all of the work is done by the compiler and its
much faster than recursion (or macros).
The trick is to create an index sequence—whose sole purpose is to
give us an index on which to apply the ... operator—of the right
size. This is done as follows:
static const std::size_t params_count = sizeof...(Params);
Compile-Time Size of a List
At compile time, when you need to know how many
elements you have in your list, you use sizeof...().
Note that in this case we stored that into a static
const variable, but it would actually be better to use a
std::integral_constant. You will learn more about
this in Chapter 3.
We are getting very close to solving our problem; that is, automating
the generation of facade code to adapt the simulation library to our
distributed system.
But the problem is not fully solved yet because we need to somehow
“iterate” on the functions. We will modify our dispatch function so
that it accepts the tuple of functions as a parameter and takes an
index, as demonstrated here:
template <std::size_t FunctionIndex,
typename FunctionsTuple,
typename Params,
std::size_t... I>
auto dispatch_params(FunctionsTuple & functions,
Params & params,
return (std::get<FunctionIndex>(functions))
And we will use the same index_sequence trick to call dis
patch_params on every function of the tuple:
template <typename FunctionsTuple,
std::size_t... I,
Making Values and Pointers Work Together | 21
typename Params,
typename ParamsSeq>
auto dispatch_functions(FunctionsTuple & functions,
Params & params,
ParamsSeq params_seq)
return std::make_tuple(dispatch_params<I>(functions,
The previous code enables us to aggregate the result of the succes‐
sive calls to each element of the tuple into a single tuple.
The final code thus becomes:
template <typename LegacyFunction,
typename... Functions,
typename... Params>
auto magic_wand(
LegacyFunction legacy,
const std::tuple<Functions...> & functions,
const std::tuple<Params...> & params1,
const std::tuple<Params...> & params2)
static const std::size_t functions_count =
static const std::size_t params_count = sizeof...(Params);
make_tuple_of_derefed_params_t<LegacyFunction> params =
/* rest of the code */
As you can see, the logic of our function makes generalization to an
arbitrary list of parameters possible.
Calling the Legacy C Function
We now have loaded in a tuple the results of our C++ method calls.
Now we want to pass a pointer to these values to the C function.
22 | Chapter 2: C++ Metaprogramming in Practice
With all the concepts we have seen so far, we know how to solve
that problem.
We need to determine the size of our results tuple, which we can
do by calling the std::tuple_size function (which is compile-
time) and do exactly what we’ve done previously to pass all of the
template <typename F, typename Tuple, std::size_t... I>
void dispatch_to_c(F f, Tuple & t, std::index_sequence<I...>)
The only twist is that we will take the address to the tuple member
because the C function requires a pointer to the value to update. It is
safe because std::get<> returns a reference to the tuple value.
Here is the completed function:
template <typename LegacyFunction,
typename... Functions,
typename... Params>
auto magic_wand(
LegacyFunction legacy,
const std::tuple<Functions...> & functions,
const std::tuple<Params...> & params1,
const std::tuple<Params...> & params2)
static const std::size_t functions_count =
static const std::size_t params_count = sizeof...(Params);
using tuple_type =
tuple_type t =
static const std::size_t t_count =
Making Values and Pointers Work Together | 23
return params;
Simplifying the Code
Wouldn’t it be nice if we didn’t need to specify the type of the result
of the tuple concatenation? After all, the compiler knows which kind
of tuple it’s going to be. But in that case, how could we compute the
size of the resulting tuple?
We can use the decltype directive to access the type of a variable:
auto val = /* something */;
decltype(val) // get type of val
This simplifies the code and removes the need for the
make_tuples_of_params_t functor, as shown here:
template <typename LegacyFunction,
typename... Functions,
typename... Params>
auto magic_wand(LegacyFunction legacy,
const std::tuple<Functions...> & functions,
const std::tuple<Params...> & params1,
const std::tuple<Params...> & params2)
static const std::size_t functions_count =
static const std::size_t params_count =
auto params = std::tuple_cat(
static constexpr auto t_count =
24 | Chapter 2: C++ Metaprogramming in Practice
return params;
You could also improve the efficiency of the code by using rvalue
references and ensuring that you use perfect forwarding semantics.
Putting It All Together
How can we use what we’ve built to finalize facade generation?
For clarity, we will use an explicit return type, but we could use
auto. Using an explicit return type has the advantage of generating a
compilation error if your type conversions are incorrect.
Another important reason for this decision is that we can consider
get_adjusted_values as a public API function. Using an auto
return type makes the function more difficult to use because its
return type isn’t clear. Your users aren’t compilers!
Let’s have a look at the code:
template <typename Reading>
std::tuple<double, double, double, double>
get_adjusted_values(Reading & r,
location l,
time t1,
time t2)
return magic_wand(adjust_values,
[&r](double l, double t)
return r.alpha_value(l, t);
[&r](double l, double t)
return r.beta_value(l, t);
std::make_tuple(l, t1),
std::make_tuple(l, t2));
The power of this new function is that if the legacy C function or the
C++ object changes, there will be little to no code rewriting to
be done.
Writing the wrappers will also be extremely straightforward, safe,
and productive: just call the magic_wand function with the required
Putting It All Together | 25
values. You can make it even more generic by wrapping the parame‐
ters in other functors and deducing the right types as needed.
And guess what? Its also possible to write code to generate all the
wrappers for you based on the function profiles. We’ve seen in this
chapter all of the building blocks to achieve that.
Did we accomplish our mission? We’d like to believe that, yes,
we did.
With the use of a couple of template metaprogramming tricks, we
managed to drastically reduce the amount of code required to get
the job done. That’s the immediate benefit of automating code gen‐
eration. Less code means fewer errors, less testing, less maintenance,
and potentially better performance.
This is the strength of metaprogramming. You spend more time
carefully thinking about a small number of advanced functions, so
you don’t need to waste your time on many trivial functions.
Now that you have been exposed to template metaprogramming,
you probably have many questions. How can I check that my
parameters are correct? How can I get meaningful error messages
if I do something wrong? How can I store a pure list of types,
without values?
More importantly, can these techniques be made reusable?
Let’s take it from the beginning…
26 | Chapter 2: C++ Metaprogramming in Practice
C++ Metaprogramming
and Application Design
Chapter 2 gave you a taste of how powerful type manipulation can
be and how it can make code more compact, generic, and elegant,
and less error prone.
As in all automation systems, metaprogramming requires some
tools to avoid rewriting the same code over and over. This chapter
will go over the basic components of such a toolbox. We also will try
to extract axioms (small, self-contained elements of knowledge) to
apply in our everyday jobs as metaprogram developers.
Compile-Time Versus Runtime Paradigms
C++ runtime code is based on the fact that users can define func‐
tions that will operate on values represented by some data type
(either native or user defined) to produce new values or side effects.
A C++ program is then the orchestration of said function calls to
advance the application’s goal at runtime.
If those notions of functions, values, and types defined by runtime
behavior are pretty straightforward, things become blurry when
speaking about their compile-time equivalents.
1If you think it could be useful, feel free to write such a proposal.
2If you’re an astute reader with a more theoretical background, you might have recog‐
nized a crude implementation of Church numerals.
Values at Compile Time
Compile-time computations need to operate on values defined as
valid at compile time. Here’s what this notion of compile-time values
Types, which are entities defined at compile time
Integral constants, defined by using the constexpr keyword or
passed to the template class as template parameters.
Unfortunately, types and integral constants are two completely dif‐
ferent beasts in C++. Moreover, there is currently no way in
C++11/14/17 to specify a template parameter to be either a type or
an integral constant.1 To be able to work with both kinds of values
(types and integral constants), we need a way to make both kinds of
values homogeneous.
The standard way to do so, which stems from Boost.MPL, is to
wrap the value in a type. We can do this by using
std::integral_constant, which we can implement roughly as
template<typename T, T Value>
struct integral_constant
using type = T;
static constexpr T value = Value;
This structure is a simple box containing a value. This box’s type is
dependent on both the value and the value type, making it unambig‐
uous.2 We can later retrieve either the value type through the ::type
internal type definition or the numerical value through the ::value
Because integral constants can be turned easily into types, we can
consider that the only required flavor of compile-time values is type.
This is such a strong proposition that we will define this as one of
our basic axioms in metaprogramming.
28 | Chapter 3: C++ Metaprogramming and Application Design
Meta-Axiom #1
Types are first-class values inside compile-time pro‐
Functions at Compile Time
We can view runtime functions as mappings between values of some
types and results of a given type (which might be void). In a perfect
world, such functions would behave the same way as their mathe‐
matical, pure equivalents, but in some cases we might need to trig‐
ger side effects (like I/O or memory access).
Functions at compile time live in a world where no side effects
can occur. They are, by their very nature, pure functions living in a
small functional language embedded in C++. As per our first axiom,
the only interesting values in metaprogramming are types. Thus,
compile-time functions are components that map types to other
This statement looks familiar. It is basically the same definition as
that for runtime functions, with “values” replaced by “types.” The
question now is, how can we specify such a component when it’s
clear C++ syntax won’t let us write return float somewhere?
Again, we take advantage of the pioneering work of Boost.MPL by
reusing its notion of the metafunction. Quoting from the documen‐
tation, a metafunction is “a class or a class template that represents a
function invocable at compile-time.” Such classes or class templates
follow a simple protocol. The inputs of the metafunctions are passed
as template parameters and the returned type is provided as an
internal type definition.
A simple metafunction can be written as follows:
template<class T>
struct as_constref
using type = T const&;
Compile-Time Versus Runtime Paradigms | 29
3Let’s ignore the potential issue of adding a reference to a reference type.
As its name implies, this metafunction turns a type into a reference
to a constant value of said type.3 Invoking a metafunction is just a
matter of accessing its internal ::type, as demonstrated here:
using cref = as_constref<float>::type;
This principle has already leaked from MPL into the standard. The
type_traits header provides a large number of such metafunc‐
tions, supporting for analyzing, creating, or modifying types based
on their properties.
Pro Tip
Most of the basic needs for type manipulation are pro‐
vided by type_traits. We strongly advise any
metaprogrammer-in-training to become highly famil‐
iar with this standard component.
Type Containers
C++ runtime development relies on the notion of containers to
express complex data manipulations. Such containers can be defined
as data structures holding a variable number of values and following
a given schema of storage (contiguous cells, linked cells, and so on).
We can then apply operations and algorithms to containers to mod‐
ify, query, remove, or insert values. The STL provides pre-made
containers, like list, set, and vector.
How can we end up with a similar concept at compile time? Obvi‐
ously, we cannot request memory to be allocated to store our values.
Moreover, our “values” actually being types, such storage makes lit‐
tle sense. The logical leap we need to make is to understand that
containers are also values, which happen to contain zero or more
other values; if we apply our systematic “values are types” motto,
this means that compile-time containers must be types that contain
zero or more other types. But how can a type contain another type?
There are multiple solutions to this issue.
30 | Chapter 3: C++ Metaprogramming and Application Design
The first idea could be to have a compile-time container be a type
with a variable number of internal using statements, as in the fol‐
lowing example:
struct list_of_ints
static constexpr std::size_t size = 4;
using element0 = char;
using element1 = short;
using element2 = int;
using element3 = long;
There are a few issues with this solution, though. First, there is no
way to add or remove types without having to construct a new type.
Then, accessing a given type is complex because it requires us to be
able to map an integral constant to a type name.
Another idea is to use variadic templates to store types as the param‐
eter pack of a variadic type. Our list_of_ints then becomes the
template<typename... Values> struct meta_list {};
using list_of_ints = meta_list<chr,short,int,long>;
This solution has neither of the aforementioned drawbacks. Opera‐
tions on this meta_list can be carried out by using the intrinsic
properties of the parameter pack, because no name mapping is
required. Insertion and removal of elements is intuitive; we just
need to play with the contents of the parameter pack.
Those properties of variadic templates define our second axiom of
metaprogramming: the fact that any variadic template structure is
de facto a compile-time container.
Meta-Axiom #2
Any template class accepting a variable number of type
parameters can be considered a type container.
Compile-Time Operations
We now have defined type containers as arbitrary template classes
with at least a template parameter pack parameter. Operations on
such containers are defined by using the intrinsic C++ support for
template parameter packs.
Compile-Time Operations | 31
We can do all of the following:
Retrieve information about the pack.
Expand or shrink the pack’s contents.
Rewrap the parameter pack.
Apply operations on the parameter packs contents.
Using Pack-Intrinsic Information
Let’s try to make a simple metafunction that operates on a type con‐
tainer by writing a way to access a container’s size:
template<class List> struct size;
template<template<class...> class List, class... Elements>
struct size<List<Elements...>>
: std::integral_constant<std::size_t, sizeof...(Elements)>
Let’s go over this snippet. First, we declare a size structure that
takes exactly one template parameter. At this point, the nature of
this parameter is unknown; thus, we can’t give size a proper defini‐
tion. Then, we partially specialize size for all of the types of the
form List<Elements...>. The syntax is a bit daunting, so lets
decompose it. The template parameters of this specialization com‐
prise the following:
A template template parameter awaiting a template parameter
pack as an argument
A template parameter pack
From those two parameters, we specialize size<Elements...>. We
can write this as Elements..., which will trigger an expansion of
every type in the pack, which is exactly what List requires in its
own parameters. This technique of describing the variadic structure
of a type container so that an algorithm can be specified will be our
main tool from now on.
Take a look at how we can use this compile-time algorithm and how
the compiler interprets this call. Consider the following as we try to
evaluate the size of std::tuple<int,float,void>:
constexpr auto s = size<std::tuple<int,float,void>>::value;
32 | Chapter 3: C++ Metaprogramming and Application Design
By the definition of std::tuple, this call will match the
size<List<Elements...>> specialization. Rather trivially, List will
be substituted by std::tuple and Elements will be substituted by
the parameter pack {int, float, void}. When it is there, the
sizeof... operator will be called and will return 3. size will then
inherit publicly from std::integral_constant<std::size_t,3>
and forward its internal value constant. We could have used any
kind of variadic structure instead of a tuple, and the process would
have been similar.
Adding and Removing Pack Elements
The next natural step is to try to modify the elements inside a
type container. We can do this by using the structural description
of a parameter pack. As an example, let’s try to write
push_back<List,Element>, which inserts a new element at the end
of a given type container.
The implementation starts in a now-familiar way:
template<class List, class New> struct push_back;
As for size, we declare a push_back structure with the desired type
interface but no definition. The next step is to specialize this type so
that it can match type containers and proceed:
template<template<class...> class List,
class... Elements, class New>
struct push_back<List<Elements...>, New>
using type = List<Elements...,New>;
As compile-time metaprogramming has no concept of values, our
only way to add an element to an existing type container is to
rebuild a new one. The algorithm is pretty simple: expand the exist‐
ing parameter pack inside the container and add one more element
at the end. By the definitions of List and ..., this builds a new valid
type where the New element has been inserted at the end.
Can you infer the implementation for push_front?
Compile-Time Operations | 33
Removal of existing elements in a type container follows a similar
reasoning but relies on the recursive structure of the parameter
pack. Fear not! As we said earlier, recursion in template metaprog‐
ramming is usually ill advised, but here we will only exploit the
structure of the parameter pack and we won’t do any loops. Lets
begin with the bare-bones code for a hypothetical remove_front
template<class List> struct remove_front;
template<template<class...> class List, class... Elements>
struct remove_front<List<Elements...>>
using type = List</* what goes here??? */>;
As you can see, we haven’t diverged much from what we’ve seen so
far. Now, let’s think about how we can remove the first type of an
arbitrary parameter pack so that we can complete our implementa‐
tion. Let’s enumerate the cases:
Contains at least one element (the head) and a potentially empty
pack of other types (the tail). In this case, we can write it as
List<Head, Tail...>
This is empty. In this case, it can be written as List<>.
If we know that a head type exists, we can remove it. If the list is
empty, the job is already done. The code then reflects this process:
template<class List> struct remove_front;
template<template<class...> class List
, class Head, class... Elements>
struct remove_front<List<Head,Elements...>>
using type = List<Elements...>;
template<template<class...> class List>
struct remove_front<List<>>
using type = List<>;
34 | Chapter 3: C++ Metaprogramming and Application Design
This introspection of the recursive nature of the parameter pack is
another tool in our belt. It has some limitations, given that decom‐
posing a pack into a list of heads and a single tail type is more com‐
plex, but it helps us build basic blocks that we can reuse in more
complex contexts.
Pack Rewrapping
So far, we’ve dealt mostly with accessing and mutating the parameter
pack. Other algorithms might need to work with the enclosing type
As an example, let’s write a metafunction that turns an arbitrary type
container into a std::tuple. How can we do that? Because the dif‐
ference between std::tuple<T...> and List<T...> is the enclosing
template type, we can just change it, as shown here:
template<class List> struct as_tuple;
template<template<class..> class List, class... Elements>
struct as_tuple<List<Elements...>>
using type = std::tuple<Elements...>;
But wait: there’s more! Changing the type container to tuple or
variant or anything else can actually be generalized by passing the
new container type as a parameter. Let’s generalize as_tuple into
struct rename;
template<template<class..> class Container
, template<class..> class List
, class... Elements
struct rename<Container, List<Elements...>>
using type = Container<Elements...>;
The code is rather similar. We use the fact that a template template
parameter can be passed naturally to provide rename with its actual
target. A sample call can then be as follows:
using my_variant = rename<boost::variant
, std::tuple<int,short>
Compile-Time Operations | 35
This technique was explained by Peter Dimov in his
blog in 2015 and instigated a lot of discussion around
similar techniques.
Container Transformations
These tools—rewrapping, iteration, and type introspection for type
containers—lead us to the final and most interesting metaprograms:
container transformations. Such transformations, directly inspired
by the STL algorithms, will help introduce the concept of structured
Concatenating containers
A first example of transformation is the concatenation of two exist‐
ing type containers. Considering any two lists L1<T1...> and
L2<T2...>, we wish to obtain a new list equivalent
to L1<T1...,T2...>.
The first intuition we might have coming from our runtime experi‐
ence is to find a way to “loop” over types as we repeatedly call
push_back. Even if it’s a correct implementation, we need to fight
this compulsion of thinking with loops. Loops over types will
require a linear number of intermediate types to be computed, lead‐
ing to unsustainable compilation times. The correct way of handling
this use case is to find a natural way to exploit the variadic nature of
our containers.
In fact, we can look at append as a kind of rewrapping in which we
push into a given variadic structure more types than it contained
before. A sample implementation can then be as follows:
template<typename L1, typename L2> struct append;
tempate< template<class...> class L1, typename... T1
, template<class...> class L2, typename... T2
struct append< L1<T1...>, L2<T2...> >
using type = L1<T1...,T2...>;
36 | Chapter 3: C++ Metaprogramming and Application Design
After the usual declaration, we define append as awaiting two differ‐
ent variadic structures filled with two distinct parameter packs. Note
that, as with regular specialization on nonvariadic templates, we can
use multiple parameter packs as long as they are wrapped properly
in the specialization. We now have access to all of the elements
required. The result is computed as the first variadic type instanti‐
ated with both parameter packs expanded.
Pro Tip
Dealing with compile-time containers requires no
loops. Try to express your algorithm as much as possi‐
ble as a direct manipulation of parameter packs.
Toward a compile-time transform
The append algorithm was rather straightforward. Let’s now hop
to a more complex example: a compile-time equivalent to
std::transform. Let’s first state what the interface of such a meta‐
program could be. In the runtime world, std::transform calls a
callable object over each and every value of the target container and
fills another container with the results. Again, this must be trans‐
posed to a metafunction that will iterate over types inside a parame‐
ter pack, apply an arbitrary metafunction, and generate a new
parameter pack to be returned.
Even if “iterating over the contents of a parameter pack using ...” is
a well-known exercise, we need to find a way to pass an arbitrary
metafunction to our compile-time transform variant. A runtime
callable object is an object providing an overload for the so-called
function call operator—usually denoted operator(). Usually those
objects are regular functions, but they can also be anonymous func‐
tions (aka lambda functions) or full-fledged user-defined classes
providing such an interface.
Generalizing metafunctions
In the compile-time world, we can pass metafunctions directly by
having our transform metaprogram await a template template
parameter. This is a valid solution, but as for runtime functions, we
might want to bind arbitrary parameters of existing metafunctions
to maximize code reuse.
Compile-Time Operations | 37
Let’s introduce the Boost.MPL notion of the metafunction class. A
metafunction class is a structure, which might or might not be a
template, that contains an internal template structure named apply.
This internal metafunction will deal with actually computing our
new type. In a way, this apply is the equivalent of the generalized
operator() of callable objects. As an example, let’s turn
std::remove_ptr into a metafunction class:
struct remove_ptr
template<typename T> struct apply
using type = typename std::remove_ptr<T>::type;
How can we use this so-called metafunction class? Its a bit different
than with metafunctions:
using no_ptr = remove_ptr::apply<int*>::type;
Note the requirement of accessing the internal apply template struc‐
ture. Wrapping this so that the end user is shielded from complexity
is tricky.
Note how the metafunction class is no longer a template but relies
on its internal apply to do its bidding. If you’re an astute reader, you
will see that we can generalize this to convert any metafunction into
a metafunction class. Let’s introduce the lambda metafunction:
template<template<class...> class MetaFunction>
struct lambda
struct type
template<typename Args...> struct apply
using type = typename MetaFunction<Args...>::type;
This lambda structure is indeed a metafunction because it contains
an internal type to be retrieved. This type structure is using pack
expansion to adapt the template template parameter of lambda so
that its usage is correct. Notice also that, like runtime lambda func‐
tions, this internal type is actually anonymous.
38 | Chapter 3: C++ Metaprogramming and Application Design
Implementing transform
We now have a proper protocol to pass metafunctions to our
compile-time transform. Let’s write a unary transform that works
on type containers:
template<typename List, typename F> struct transform;
template<template<class...> class List,
struct transform<List<Elems...>,F>
using call = typename F::template apply<T>::type;
using type = List< call<Elems>... >;
This code is both similar to what we wrote earlier and a bit more
complex. It begins as usual by declaring and defining transform as
acting on container types using a parameter pack. The actual code
performs iterations over elements of the container using the clas‐
sic ... approach. The addition we need to make is to call the meta‐
function class F over each type. We do this by taking advantage of
the fact that ... will unpack and apply the code fragment on its left
for every type in the pack. For clarity, we use an intermediate tem‐
plate using a statement to hold the actual metafunction class appli‐
cation to a single type.
Now, as an example, let’s call std::remove_ptr on a list of types:
using no_pointers = transform< meta_list<int*,float**, double>
, lambda<std::remove_ptr>
Note the abstraction power of algorithms being transposed to the
compile-time world. Here we used a high-level metafunction to
apply a well-known pattern of computation on a container of types.
Observe also how the lambda construct can help us make the use
and reuse of existing metafunctions easier.
Pro Tip
Metafunctions follow similar rules to those for func‐
tions: they can be composed, bound, or turned into
various similar yet different interfaces. The transition
between metafunctions and metafunction classes is
only the tip of the iceberg.
Compile-Time Operations | 39
Advanced Uses of Metaprogramming
With a bit of imagination and knowledge, you can do things much
more advanced than performing compile-time checks with template
metaprogramming. The purpose of this section is just to give you an
idea of what is possible.
A Revisited Command Pattern
The command pattern is a behavioral design pattern in which you
encapsulate all the information required to execute a command into
an object or structure. It’s a great pattern, which in C++ is often
written with runtime polymorphism.
Putting aside the tendency of runtime polymorphism to induce the
“factory of factories” antipattern, there can be a nonnegligible per‐
formance cost induced by vtables because they prevent the compiler
from aggressively optimizing and inlining code.
The “Factory of Factories” Antipattern
This antipattern can happen in object-oriented pro‐
gramming when you spend more time writing code
to manage abstractions than you do writing code to
solve problems.
From a strictly software design point of view, it also forces you to
relate objects together just because they will go through the same
function at some point in time.
If generic programming has taught us anything, its that you don’t
need to create a relation between objects just to make them use
a function.
All you need to do is make the objects share common properties:
struct first_command
std::string operator()(int) { /* something */ }
struct second_command
std::string operator()(int) { /* something */ }
40 | Chapter 3: C++ Metaprogramming and Application Design
4To be fair, C++11 introduced a series of enhancements that allow the programmer to
ensure at compile time that she hasn’t forgotten to implement a virtual function. That
doesn’t eliminate the risk of invalid dynamic casts, though.
And have a function that accepts a command:
template <typename Command>
void execute_command(const Command & c, int param)
To which you will retort, “How do I transmit those commands
through a structure, given that I know only at runtime which com‐
mand to run?”
There are two ways to do it: manually, by using an unrestricted
union, or by using a variant such as Boost.Variant. Template meta‐
programming comes to the rescue because you can safely list the
types of the commands in a type list and build the variant (or the
union) from that list.
Not only will the code be more concise and more efficient, but it will
also be less error prone: at compile time you will get an error if you
forgot to implement a function, and the “pure virtual function call”
is therefore impossible.4
Compile-Time Serialization
What do we mean by compile-time serialization? When you want to
serialize an object, there are a lot of things you already know at com‐
pile time—and remember, everything you do at compile time
doesn’t need to be done any more at runtime.
That means much faster serialization and more efficient memory
For example, when you want to serialize a std::uint64_t, you
know exactly how much memory you need, whereas when you seri‐
alize a std::vector<std::uint64_t>, you must read the size of the
vector at runtime to know how much memory you need to allocate.
Recursively, it means that if you serialize a structure that is made up
strictly of integers, you are able, at compile time, to know exactly
how much memory you need, which means you can allocate the
required intermediate buffers at compile time.
Advanced Uses of Metaprogramming | 41
With template metaprogramming, you can branch, at compile time,
the right code for serialization. This means that for every structure
for which you are able to exactly compute the memory require‐
ments, you will avoid dynamic memory allocation altogether, yield‐
ing great performance improvements and reduced memory usage.
Helper Functions and Libraries
Must you reinvent the wheel and write all your own basic functions,
like we’ve seen in this chapter? Fortunately, no. Since C++11, a great
number of helper functions have been included in the standard, and
we strongly encourage you to use them whenever possible.
The standard isn’t yet fully featured when it comes to metaprogram‐
ming; for example, it lacks an official “list of types” type, algorithms,
and more advanced metafunctions.
Fortunately, there are libraries that prevent you from needing to
reinvent the wheel and that will work with all major compilers. This
will save you the sweat of working around compilers’ idiosyncrasies
and enable you to focus on writing metaprograms.
Boost comes with two libraries to help you with template metaprog‐
MPL, written by Aleksey Gurtovoy and David Abrahams
A complete C++03 template metaprogramming toolbox that
comes with containers, algorithms, and iterators. Unless you are
stuck with a C++03 compiler, we would recommend against
using this library.
Hana, written by Louis Dionne
A new metaprogramming paradigm, which makes heavy use of
lambdas. Hana is notoriously demanding of the compiler.
The authors of this report are also the authors of Brigand, a
C++11/14 metaprogramming library that nicely fills the gap
between Boost.MPL and Boost.Hana.
We strongly encourage you to use existing libraries because they will
help you structure your code and give you ideas of what you can do
with metaprogramming.
42 | Chapter 3: C++ Metaprogramming and Application Design
In this chapter, we took a journey into the land of types within C++
and we saw that they can be manipulated like runtime values.
We defined the notion of type values and saw how such a notion can
lead to the definition of type containers; that is, types containing
other types. We saw how the expressiveness of parameter packs can
lead to a no-recursion way of designing metaprograms. The small
yet functional subset of classic container operators we defined
showed the variety of techniques usable to design such metapro‐
grams in a systemic way.
We hope we reached our goal of giving you a taste for metaprogram‐
ming and proving that it isn’t just some arcane technique that
should not be used outside of research institutes. Whether you want
to check out the libraries discussed herein, write your own first met‐
aprogram, or revisit some code you wrote recently, we have only one
hope: that by reading this report you learned something that will
make you a better programmer.
Summary | 43
About the Authors
Edouard Alligand is the founder and CEO of Quasardb, an
advanced, distributed hyper scalable database. He has more than 15
years of professional experience in software engineering. Edouard
combines an excellent knowledge of low-level programming with a
love for template metaprogramming, and likes to come up with
uncompromising solutions to seemingly impossible problems. He
lives in Paris, France.
Joel Falcou is CTO of NumScale, an Associate Professor at the Uni‐
versity of Paris-Sud, and a researcher at the Laboratoire de Recher‐
che d’Informatique in Orsay, France. He is a member of the C++
Standards Committee and the author of Boost.SIMD and NT2. Joel’s
research focuses on studying generative programming idioms and
techniques to design tools for parallel software development.
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.