Prerequisite: One must be familiar with coding in C++.
I coded my first C++ libraries 5 years back, I had just started college, and wanted to do something solid!, One fine evening, I was looking back at that code, It was so, poorly written, So I started fixing it. I googled C++ Convention, After many search results, I found out there are no good resources out there. So I took upon myself, To Write One!
These are coding practices one may follow to write readable good code.
File Types:
Any programing file, in the end, is nothing but a text file with a fancy extension. In the Book of Genesis, the mentioned naming style is:
.c for C.cpp for C++.h for headers (independent of C or C++)
But due to the opensource nature of GNU, A bunch of different assembler and compilers were made which came with their own extensions variations, such as, .cc, .C, .cxx, .c++ and .hh, .H, .hxx, .hpp, .h++.
TBH, Just chose one set of extensions, after all, even .txt can be compiled.
I use, (Since Most IDE recognize them):
.c for C.cpp for C++.h for C headers.hpp for C++ headers
makefile:
A makefile is a special file, containing shell commands, that you create and name makefile (no extensions)
It is executed using the make command. My Typical makefile looks like this:
CXX=g++
CXXFLAGS=-std=c++11 -w
build: main.cpp local_header.hpp includes/external_lib.hpp
@echo "\n\nmain.cpp: Building..."
$(CXX) $(CXXFLAGS) -o main.out main.cpp
run: main.out
@echo "\n\nmain.cpp: Executing..."
./main.out
clean: main.out
@echo "\n\nmain.cpp: Cleaning..."
rm main.out
test: test.cpp
@echo "\n\ntest.cpp: Building..."
$(CXX) $(CXXFLAGS) -o test.out test.cpp
@echo "\n\ntest.cpp: Executing..."
./test.out
@echo "\n\ntest.cpp: Cleaning..."
rm ./test.out
all: build run clean
$
$ make all
One can use something like GNU Automake or CMake instead of hand-written makefiles. These tools may add the overhead of other files, but these files are informative files. Use these tools, if they benefit your project.
NOTE: use -MMD flag to get a .d file, containing all dependencies.
PreProcessors:
The preprocessors are the directives, which instruct the compiler to preprocess the information before starting compilation. All preprocessors start with #, For example: #include<iosteam>
These are the coding practices, I use:
Double Quotes ("...") for including local headers, This rule increases readability of Code
3 Levels of includes (I picked this habit while working in python) with linespace between them.
- First, for core
g++ system files.
- Second, for externally added libraries.
- Third, for locally created files.
As mentioned here one should include local headers first โ then system headers. The reason is to guarantee that all local headers are self-sufficient
"self-sufficient" means, local headers are not depending on System or External libraries.
This although is a good way of checking, I believe this "check" is not something headers/compilers should do, well at least not at the file level.
In an ideal world, Each header would have its own test files to resolve this, until then, going against the natural order might not be that bad.
SO, According to the above rules, A good header would be.
#pragma once
#include<system_file_1>
#include<system_file_2> // 1 line-space afterwards
#include<external_lib1_1> // 1 line-space afterwards
#include<external_lib2_2> // club external libs together
#include<external_lib2_3> // 1 line-space afterwards
#include "local_file"
namespace ra {
...
}
Nomenclature:
File | namespace | Class | function:
- Nothing is fixed, one may use
camelCase, PascalCase or
snake_case.
- I personally find
snake_case to be more readable.
Template Parameter:
- same as above, no rules.
- I prefer
PascalCase here.
Constants | Macros:
UPPER_SNAKE_CASE is preferred everywhere.
Variables:
- Before IDE(s) were a thing, People just couldn't hover over variables to get info. So some old libraries use the following conventions:
m_<any case> for private memberst_<any case> for function parameters
- I personally, discard this rule.
Taboo of Naming:
Check this link
C++ allows for 2 types of comments:
Generally speaking, single-line comments provide more control over the appearance and are considered better.
As a python developer, I wholeheartedly disagree, This is how I prefer my code!
class custom_class {
private:
float variable {3.14};
void function();
}
void custom_class::function(){
}
Custom Exceptions:
Custom exceptions provide you the flexibility to add attributes and methods that are not part of a standard programming language. These are most helpful in Debugging when code is unable to compile.
Below is My Favorite Exception:
namespace ra
{
class external_exception : public std::logic_error
{
public:
external_exception(
std::string function = __builtin_FUNCTION()
) : std::logic_error("`" + function + "` called for external exception!"){};
};
}
void main (){ throw ra::external_exception(); }
terminate called after throwing an instance of 'ra::external_exception'
what(): `main` called for external exception!
Let me explain, In argument of constructor of exception it says:
std::string function = __builtin_FUNCTION()
__builtin_FUNCTION() return the source location, when called.
Since called in Argument, whenever it is initiated that is the source location, Essentially Making it a traceable Exception
Here We have inherited from logic_error, but other exceptions/errors could be found here
NOTE: In C++ Exceptions and Errors are the same, but in theory, Exceptions are something that could be handled, whereas errors should not be as they break the system.
Modularity
A Modular design is shown to improve the design process by allowing better re-usability, relatively low maintenance, workload handling, and easier debugging processes. Modularity in C++ can be achieved by the following
Template
As defined (here)[https://www.javatpoint.com/cpp-templates]: A C++ template is a powerful feature added to C++. It allows you to define the generic classes and generic functions and thus provides support for generic programming. Generic programming is a technique where generic types are used as parameters in algorithms so that they can work for a variety of data types
Templates can be represented in two ways:
template <typename TemplateKlass>
TemplateKlass get_maximum(TemplateKlass x, TemplateKlass y){
return (x > y)? x: y;
}
I am will not into details, As there are far better resources
Function as Argument:
We can pass a function in the argument of another function, Well Actually A function pointer is passed to be exact. This comes in handy in the situation, where you have multiple services for the same request, for example, Hashing, you can use so many third-party algorithms for hashing.
bool verify_hash(std::string (*hash_func)(std::string)){
return hashed_output == hash_func(input)
}
Inheritance
I picked up this in my Unity3D days, In Unity3D Almost everything is inherited and then used.
The plan is simple, Create a class with a bunch of public members all throwing error.
And Then Everyone would have to Inherit that class and override it.
For Example
class base_class{
public:
base_class(){}
base_class(float) {throw not_implemented_exception();}
virtual operator string() const = 0;
bool operator<(const base_class &other) const { throw not_implemented_exception(); }
template <typename TemplateKlass>
void some_func(TemplateKlass temp) {throw not_implemented_exception();}
}
class extended_class : public base_class{
private:
float m_temp;
public:
extended_class(){}
extended_class(float t_temp) : m_temp(t_temp)
operator std::string() {return to_string(t_temp);}
bool operator<(const extended_class &other) const { m_temp < other.m_temp }
template <typename TemplateKlass>
void some_func(TemplateKlass temp) {...}
}
Bonus: Icon Pack!
Yeah... I created my own icons too...
