book

C++ Templates: The Complete Guide, 2nd Edition

by David Vandevoorde, Nicolai M. Josuttis, Douglas Gregor

September 2017

Intermediate to advanced

822 pages

26h 51m

English

Addison-Wesley Professional

Read now

Unlock full access

What You Should Know Before Reading This BookOverall Structure of the BookHow to Read This BookSome Remarks About Programming StyleThe C++11, C++14, and C++17 StandardsExample Code and Additional InformationFeedback

1.1 A First Look at Function Templates1.1.1 Defining the Template1.1.2 Using the Template1.1.3 Two-Phase Translation1.2 Template Argument Deduction1.3 Multiple Template Parameters1.3.1 Template Parameters for Return Types1.3.2 Deducing the Return Type1.3.3 Return Type as Common Type1.4 Default Template Arguments1.5 Overloading Function Templates1.6 But, Shouldn’t We …?1.6.1 Pass by Value or by Reference?1.6.2 Why Not inline?1.6.3 Why Not constexpr?1.7 Summary
2.1 Implementation of Class Template Stack2.1.1 Declaration of Class Templates2.1.2 Implementation of Member Functions2.2 Use of Class Template Stack2.3 Partial Usage of Class Templates2.3.1 Concepts2.4 Friends2.5 Specializations of Class Templates2.6 Partial Specialization2.7 Default Class Template Arguments2.8 Type Aliases2.9 Class Template Argument Deduction2.10 Templatized Aggregates2.11 Summary
3.1 Nontype Class Template Parameters3.2 Nontype Function Template Parameters3.3 Restrictions for Nontype Template Parameters3.4 Template Parameter Type auto3.5 Summary
4.1 Variadic Templates4.1.1 Variadic Templates by Example4.1.2 Overloading Variadic and Nonvariadic Templates4.1.3 Operator sizeof…4.2 Fold Expressions4.3 Application of Variadic Templates4.4 Variadic Class Templates and Variadic Expressions4.4.1 Variadic Expressions4.4.2 Variadic Indices4.4.3 Variadic Class Templates4.4.4 Variadic Deduction Guides4.4.5 Variadic Base Classes and using4.5 Summary
5.1 Keyword typename5.2 Zero Initialization5.3 Using this->5.4 Templates for Raw Arrays and String Literals5.5 Member Templates5.5.1 The .template Construct5.5.2 Generic Lambdas and Member Templates5.6 Variable Templates5.7 Template Template Parameters5.8 Summary
6.1 Perfect Forwarding6.2 Special Member Function Templates6.3 Disable Templates with enable_if<>6.4 Using enable_if<>6.5 Using Concepts to Simplify enable_if<> Expressions6.6 Summary
7.1 Passing by Value7.2 Passing by Reference7.2.1 Passing by Constant Reference7.2.2 Passing by Nonconstant Reference7.2.3 Passing by Forwarding Reference7.3 Using std::ref() and std::cref()7.4 Dealing with String Literals and Raw Arrays7.4.1 Special Implementations for String Literals and Raw Arrays7.5 Dealing with Return Values7.6 Recommended Template Parameter Declarations7.7 Summary
8.1 Template Metaprogramming8.2 Computing with constexpr8.3 Execution Path Selection with Partial Specialization8.4 SFINAE (Substitution Failure Is Not An Error)8.4.1 Expression SFINAE with decltype8.5 Compile-Time if8.6 Summary
9.1 The Inclusion Model9.1.1 Linker Errors9.1.2 Templates in Header Files9.2 Templates and inline9.3 Precompiled Headers9.4 Decoding the Error Novel9.5 Afternotes9.6 Summary
10.1 “Class Template” or “Template Class”?10.2 Substitution, Instantiation, and Specialization10.3 Declarations versus Definitions10.3.1 Complete versus Incomplete Types10.4 The One-Definition Rule10.5 Template Arguments versus Template Parameters10.6 Summary
11.1 Callables11.1.1 Supporting Function Objects11.1.2 Dealing with Member Functions and Additional Arguments11.1.3 Wrapping Function Calls11.2 Other Utilities to Implement Generic Libraries11.2.1 Type Traits11.2.2 std::addressof()11.2.3 std::declval()11.3 Perfect Forwarding Temporaries11.4 References as Template Parameters11.5 Defer Evaluations11.6 Things to Consider When Writing Generic Libraries11.7 Summary
12.1 Parameterized Declarations12.1.1 Virtual Member Functions12.1.2 Linkage of Templates12.1.3 Primary Templates12.2 Template Parameters12.2.1 Type Parameters12.2.2 Nontype Parameters12.2.3 Template Template Parameters12.2.4 Template Parameter Packs12.2.5 Default Template Arguments12.3 Template Arguments12.3.1 Function Template Arguments12.3.2 Type Arguments12.3.3 Nontype Arguments12.3.4 Template Template Arguments12.3.5 Equivalence12.4 Variadic Templates12.4.1 Pack Expansions12.4.2 Where Can Pack Expansions Occur?12.4.3 Function Parameter Packs12.4.4 Multiple and Nested Pack Expansions12.4.5 Zero-Length Pack Expansions12.4.6 Fold Expressions12.5 Friends12.5.1 Friend Classes of Class Templates12.5.2 Friend Functions of Class Templates12.5.3 Friend Templates12.6 Afternotes
13.1 Name Taxonomy13.2 Looking Up Names13.2.1 Argument-Dependent Lookup13.2.2 Argument-Dependent Lookup of Friend Declarations13.2.3 Injected Class Names13.2.4 Current Instantiations13.3 Parsing Templates13.3.1 Context Sensitivity in Nontemplates13.3.2 Dependent Names of Types13.3.3 Dependent Names of Templates13.3.4 Dependent Names in Using Declarations13.3.5 ADL and Explicit Template Arguments13.3.6 Dependent Expressions13.3.7 Compiler Errors13.4 Inheritance and Class Templates13.4.1 Nondependent Base Classes13.4.2 Dependent Base Classes13.5 Afternotes
14.1 On-Demand Instantiation14.2 Lazy Instantiation14.2.1 Partial and Full Instantiation14.2.2 Instantiated Components14.3 The C++ Instantiation Model14.3.1 Two-Phase Lookup14.3.2 Points of Instantiation14.3.3 The Inclusion Model14.4 Implementation Schemes14.4.1 Greedy Instantiation14.4.2 Queried Instantiation14.4.3 Iterated Instantiation14.5 Explicit Instantiation14.5.1 Manual Instantiation14.5.2 Explicit Instantiation Declarations14.6 Compile-Time if Statements14.7 In the Standard Library14.8 Afternotes
15.1 The Deduction Process15.2 Deduced Contexts15.3 Special Deduction Situations15.4 Initializer Lists15.5 Parameter Packs15.5.1 Literal Operator Templates15.6 Rvalue References15.6.1 Reference Collapsing Rules15.6.2 Forwarding References15.6.3 Perfect Forwarding15.6.4 Deduction Surprises15.7 SFINAE (Substitution Failure Is Not An Error)15.7.1 Immediate Context15.8 Limitations of Deduction15.8.1 Allowable Argument Conversions15.8.2 Class Template Arguments15.8.3 Default Call Arguments15.8.4 Exception Specifications15.9 Explicit Function Template Arguments15.10 Deduction from Initializers and Expressions15.10.1 The auto Type Specifier15.10.2 Expressing the Type of an Expression with decltype15.10.3 decltype(auto)15.10.4 Special Situations for auto Deduction15.10.5 Structured Bindings15.10.6 Generic Lambdas15.11 Alias Templates15.12 Class Template Argument Deduction15.12.1 Deduction Guides15.12.2 Implicit Deduction Guides15.12.3 Other Subtleties15.13 Afternotes
16.1 When “Generic Code” Doesn’t Quite Cut It16.1.1 Transparent Customization16.1.2 Semantic Transparency16.2 Overloading Function Templates16.2.1 Signatures16.2.2 Partial Ordering of Overloaded Function Templates16.2.3 Formal Ordering Rules16.2.4 Templates and Nontemplates16.2.5 Variadic Function Templates16.3 Explicit Specialization16.3.1 Full Class Template Specialization16.3.2 Full Function Template Specialization16.3.3 Full Variable Template Specialization16.3.4 Full Member Specialization16.4 Partial Class Template Specialization16.5 Partial Variable Template Specialization16.6 Afternotes
17.1 Relaxed typename Rules17.2 Generalized Nontype Template Parameters17.3 Partial Specialization of Function Templates17.4 Named Template Arguments17.5 Overloaded Class Templates17.6 Deduction for Nonfinal Pack Expansions17.7 Regularization of void17.8 Type Checking for Templates17.9 Reflective Metaprogramming17.10 Pack Facilities17.11 Modules
18.1 Dynamic Polymorphism18.2 Static Polymorphism18.3 Dynamic versus Static Polymorphism18.4 Using Concepts18.5 New Forms of Design Patterns18.6 Generic Programming18.7 Afternotes
19.1 An Example: Accumulating a Sequence19.1.1 Fixed Traits19.1.2 Value Traits19.1.3 Parameterized Traits19.2 Traits versus Policies and Policy Classes19.2.1 Traits and Policies: What’s the Difference?19.2.2 Member Templates versus Template Template Parameters19.2.3 Combining Multiple Policies and/or Traits19.2.4 Accumulation with General Iterators19.3 Type Functions19.3.1 Element Types19.3.2 Transformation Traits19.3.3 Predicate Traits19.3.4 Result Type Traits19.4 SFINAE-Based Traits19.4.1 SFINAE Out Function Overloads19.4.2 SFINAE Out Partial Specializations19.4.3 Using Generic Lambdas for SFINAE19.4.4 SFINAE-Friendly Traits19.5 IsConvertibleT19.6 Detecting Members19.6.1 Detecting Member Types19.6.2 Detecting Arbitrary Member Types19.6.3 Detecting Nontype Members19.6.4 Using Generic Lambdas to Detect Members19.7 Other Traits Techniques19.7.1 If-Then-Else19.7.2 Detecting Nonthrowing Operations19.7.3 Traits Convenience19.8 Type Classification19.8.1 Determining Fundamental Types19.8.2 Determining Compound Types19.8.3 Identifying Function Types19.8.4 Determining Class Types19.8.5 Determining Enumeration Types19.9 Policy Traits19.9.1 Read-Only Parameter Types19.10 In the Standard Library19.11 Afternotes
20.1 Algorithm Specialization20.2 Tag Dispatching20.3 Enabling/Disabling Function Templates20.3.1 Providing Multiple Specializations20.3.2 Where Does the EnableIf Go?20.3.3 Compile-Time if20.3.4 Concepts20.4 Class Specialization20.4.1 Enabling/Disabling Class Templates20.4.2 Tag Dispatching for Class Templates20.5 Instantiation-Safe Templates20.6 In the Standard Library20.7 Afternotes
21.1 The Empty Base Class Optimization (EBCO)21.1.1 Layout Principles21.1.2 Members as Base Classes21.2 The Curiously Recurring Template Pattern (CRTP)21.2.1 The Barton-Nackman Trick21.2.2 Operator Implementations21.2.3 Facades21.3 Mixins21.3.1 Curious Mixins21.3.2 Parameterized Virtuality21.4 Named Template Arguments21.5 Afternotes
22.1 Function Objects, Pointers, and std::function<>22.2 Generalized Function Pointers22.3 Bridge Interface22.4 Type Erasure22.5 Optional Bridging22.6 Performance Considerations22.7 Afternotes
23.1 The State of Modern C++ Metaprogramming23.1.1 Value Metaprogramming23.1.2 Type Metaprogramming23.1.3 Hybrid Metaprogramming23.1.4 Hybrid Metaprogramming for Unit Types23.2 The Dimensions of Reflective Metaprogramming23.3 The Cost of Recursive Instantiation23.3.1 Tracking All Instantiations23.4 Computational Completeness23.5 Recursive Instantiation versus Recursive Template Arguments23.6 Enumeration Values versus Static Constants23.7 Afternotes
24.1 Anatomy of a Typelist24.2 Typelist Algorithms24.2.1 Indexing24.2.2 Finding the Best Match24.2.3 Appending to a Typelist24.2.4 Reversing a Typelist24.2.5 Transforming a Typelist24.2.6 Accumulating Typelists24.2.7 Insertion Sort24.3 Nontype Typelists24.3.1 Deducible Nontype Parameters24.4 Optimizing Algorithms with Pack Expansions24.5 Cons-style Typelists24.6 Afternotes
25.1 Basic Tuple Design25.1.1 Storage25.1.2 Construction25.2 Basic Tuple Operations25.2.1 Comparison25.2.2 Output25.3 Tuple Algorithms25.3.1 Tuples as Typelists25.3.2 Adding to and Removing from a Tuple25.3.3 Reversing a Tuple25.3.4 Index Lists25.3.5 Reversal with Index Lists25.3.6 Shuffle and Select25.4 Expanding Tuples25.5 Optimizing Tuple25.5.1 Tuples and the EBCO25.5.2 Constant-time get()25.6 Tuple Subscript25.7 Afternotes
26.1 Storage26.2 Design26.3 Value Query and Extraction26.4 Element Initialization, Assignment and Destruction26.4.1 Initialization26.4.2 Destruction26.4.3 Assignment26.5 Visitors26.5.1 Visit Result Type26.5.2 Common Result Type26.6 Variant Initialization and Assignment26.7 Afternotes
27.1 Temporaries and Split Loops27.2 Encoding Expressions in Template Arguments27.2.1 Operands of the Expression Templates27.2.2 The Array Type27.2.3 The Operators27.2.4 Review27.2.5 Expression Templates Assignments27.3 Performance and Limitations of Expression Templates27.4 Afternotes
28.1 Shallow Instantiation28.2 Static Assertions28.3 Archetypes28.4 Tracers28.5 Oracles28.6 Afternotes
A The One-Definition RuleA.1 Translation UnitsA.2 Declarations and DefinitionsA.3 The One-Definition Rule in DetailA.3.1 One-per-Program ConstraintsA.3.2 One-per-Translation Unit ConstraintsA.3.3 Cross-Translation Unit Equivalence Constraints
B.1 Traditional Lvalues and RvaluesB.1.1 Lvalue-to-Rvalue ConversionsB.2 Value Categories Since C++11B.2.1 Temporary MaterializationB.3 Checking Value Categories with decltypeB.4 Reference Types
C.1 When Does Overload Resolution Kick In?C.2 Simplified Overload ResolutionC.2.1 The Implied Argument for Member FunctionsC.2.2 Refining the Perfect MatchC.3 Overloading DetailsC.3.1 Prefer Nontemplates or More Specialized TemplatesC.3.2 Conversion SequencesC.3.3 Pointer ConversionsC.3.4 Initializer ListsC.3.5 Functors and Surrogate FunctionsC.3.6 Other Overloading Contexts
D.1 Using Type TraitsD.1.1 std::integral_constant and std::bool_constantD.1.2 Things You Should Know When Using TraitsD.2 Primary and Composite Type CategoriesD.2.1 Testing for the Primary Type CategoryD.2.2 Test for Composite Type CategoriesD.3 Type Properties and OperationsD.3.1 Other Type PropertiesD.3.2 Test for Specific OperationsD.3.3 Relationships Between TypesD.4 Type ConstructionD.5 Other TraitsD.6 Combining Type TraitsD.7 Other Utilities
E.1 Using ConceptsE.2 Defining ConceptsE.3 Overloading on ConstraintsE.3.1 Constraint SubsumptionE.3.2 Constraints and Tag DispatchingE.4 Concept TipsE.4.1 Testing ConceptsE.4.2 Concept GranularityE.4.3 Binary Compatibility
ForumsBooks and Web Sites

Content preview from C++ Templates: The Complete Guide, 2nd Edition

Chapter 23 Metaprogramming

Metaprogramming consists of “programming a program.” In other words, we lay out code that the programming system executes to generate new code that implements the functionality we really want. Usually, the term metaprogramming implies a reflexive attribute: The metaprogramming component is part of the program for which it generates a bit of code (i.e., an additional or different bit of the program).

Why would metaprogramming be desirable? As with most other programming techniques, the goal is to achieve more functionality with less effort, where effort can be measured as code size, maintenance cost, and so forth. What characterizes metaprogramming is that some user-defined computation happens at translation time. The ...