Generic Programming

A programming language paradigm where algorithms are written in terms of types to-be-specified-later and then instantiated when needed for specific types of the parameters

• One of many paradigms supported by C++, along with procedural programming (functions) and object-oriented programming
• Replaces the need for variant types (one variable can store many different types of data, but typically not at the same time)
• Replaces the need for the use of inheritance

C++ Templates

• A feature of the C++ programming language that supports generic programming
• Can write code that works with multiple data types without having to duplicate the code for each type.
• Templates can be applied to both functions and classes
• Code generated from templates can be heavily optimized
• Much of the power and flexibility of the C++ Standard Library comes from the use of C++ templates

Function Template

• template keyword
• Template parameters enclosed in angle brackets
• typename keyword used to declare the template parameter

Instantiating Function Templates

• Compiler determines T from the types of parameters in the call
• Instantiates, (i.e., generates) a separate function for each type
• Can also explicitly declare the template parameter value (typically a type)
• Can expand as much as needed to new types

Non-Type

• Template parameters do not have to be types
• When a template is instantiated with specific types or values, the compiler generates a specialized version of the template code for those types or values

Class Templates

Fibonacci

C++ Template Fibonacci Sequence

Template Disadvantages

• Compiler needs access to function or method definitions to instantiate, leading to use in include file
• Instantiation for each type may lead to code bloat
• Code may become difficult to understand
• Error messages when types with the wrong interface are used

C++ Concepts

• First introduced as "C++0x Concepts" in C++11, but were removed
• Concepts Lite were merged into C++20

C++ Concepts and Errors

Extension Points

• Lambda functions passed to std::function fields/data members
• Direct inheritance from a Base class
• Inheritance from Handler used by a Base class

Observations

• Lambda functions with capture are very flexible and quick to implement in code
• Lambda functions passed to function pointer parameters are very fast
• However, lambda functions, when passed to std::function (that allows a non-empty capture), are very slow

C++ Templates

• Using C++ Templates with lambda functions is very fast
• The compiler can inline lambda functions in templates
• No use of slow std::function
• How std::sort() works
• How about as extension points?

Template Function Extension

Template Class Extension

Template Class Extension

Speed of Parameter Types

• Choices:
• Function pointer
• std::function
• Template parameter
• Clang Comparison
• GCC Comparison

Extension Points

• Lambda functions to std::function fields/data members
• Direct inheritance from a Base class
• Inheritance from Handler used by a Base class
• C++Templates for classes with lambda functions

C++ Templates as Extensions Points

• Lots of flexibility
• Allows the use of the capture
• Code is (potentially) inlined
• All the advantages of lambdas to std::function, without the speed disadvantage
• One Drawback: Have to declare in the include file (.hpp) as the compiler needs access to the definitions

Extension Points

• Lambda functions to std::function fields/data members
• Direct inheritance from a Base class
• Inheritance from Handler used by a Base class
• C++Templates for classes with lambda functions