- an organized collection of links to topic articles
- about learning modern, idiomatic C++
- (roughly) ordered by increasing level of C++ proficiency
(top to bottom)
- many articles are based on slide decks taken
from an undergraduate C++ crash course
You should know how to use a
command line
and should have a basic understanding of fundamental
imperative programming concepts, like
- variables
(
int i = 0, var x = 0, … )
- conditional branching
(
if, then, else, … )
- loops
(
for, while, repeat, … )
- functions / subroutines
(
int main(), fn main, … )
It certainly helps if you are already somewhat familiar with other
programming languages like Python or Java.
You should be aware however, that C++ is very different
from Java or C# despite the superficial syntactic similarities.
Predictable Resource Control
"direct map to hardware": CPU, memory, …
Zero-Overhead Abstractions
"you only pay for what you use"
- multi-paradigm: value-oriented, procedural, generic, functional, object-oriented
-
value semantics
by default
- custom types can behave exactly like built-in types
- deterministic object lifetime
- precise memory control down to individual bits
ISO Standard C++
| C++98 |
1998 |
the original standard |
|
| C++11 |
2011 |
almost a new language |
modern C++ |
| C++14 |
2014 |
some improvements |
| C++17 |
2017 |
new features & library extensions |
| C++20 |
2020 |
game-changing new features & libraries |
C++ vs. …
C
|
not a strict subset of C++ |
| stronger type system |
weak type system |
| high-level abstractions |
only low-level memory abstraction |
| typed compile-time programming |
only untyped macros for compile-time programming |
powerful custom types (classes) |
only data aggregation (struct) |
| use compiler as correctness checker |
get the code to compile quickly |
if it compiles, it should be correct |
debugging is the real work |
Java
| multi-paradigm: value-oriented, procedural, generic, functional, object-oriented |
mainly object-oriented, some functional and generic aspects |
|
value semantics
by default for fundamental & custom types |
value semantics
only for primitives (int, float,… ) |
| optional reference semantics for all types |
baked-in reference semantics for class types |
| precise control over memory (de-)allocation; no garbage collection |
garbage collector; can degrade performance |
| deterministic & controllable object lifetime |
no predictable object lifetime control |
| ⇒ memory frugal |
⇒ high memory consumption |
| aggressive inlining can eliminate slow function calls |
performance degradation due to un-devirtualizable, non-inlinable methods |
Python
| almost always faster |
almost always slower (in practice around 25-50 times) |
| complex syntax and tons of features can
be intimidating to newcomers |
simple syntax; usually easy to comprehend |
| statically typed |
dynamically typed |
| many types of bugs can be caught at compile time |
many types of bugs will only manifest at runtime |
| suited for safety-critical large-scale systems |
hard to build reliable large-scale systems |
| even simple, small-scope tasks can quickly require an
expert knowledge of various arcane corner cases & quirks |
tends to be more beginner-friendly and
small scripts are usually quickly written |
| fairly small standard library but extensive ecosystem with
libraries for nearly everything |
batteries included philosophy with tons of libraries
only one import away |
Where it does and doesn't shine
- if memory consumption is critical
- if runtime is critical
- if energy consumption is critical
- OS level programming
- embedded systems
- large-scale, massively parallel systems
- source-code portability across architectures
- "scripting" & rapid prototyping
- GUIs with trivial internal logic
Comments…