In two of your cases this operator is pretty shit because at some point you’ll probably want to offset the access (this isn’t a knock at you but at the feature).
This operator would only really be relevant to the last case which rarely comes up outside of infrastructure/library work (building a tool building tool) and usually those code bases are cautious to adopt new features too quickly anyways for portability.
I’ve done serious C++ work (not much in the past decade though) - while references are absolutely amazing and you essentially want to pass by const ref by default I think well written maintainable C++ should maybe have a dozen heap objects tops. C++ is a better language if you generally forget that pointers and bare arrays exist.
Just again - I think you’re right and the fact that your list is only three things long (and arguably two of them would be misuses) is a pretty clear sign that this is an incredibly niche feature.
Mostly because at the lowest level of computing (machine code and CPU instructions), pointers are the only method (that I know of) of any kind of indirection.
At the lowest level, there are 2 types of references:
CPU registers
memory addresses (pointers)
Every higher level language feature for memory management (references, objects, safe pointers, garbage collection, etc) is just an abstraction over raw pointers
Pointers themselves are really just abstractions over raw integers, whose sole purpose is to index into RAM
With that in mind, pointers to pointers are a natural consequence of any kind of nested object hierarchy (linked lists, trees, objects with references to other objects, etc)
The only other kind of indirection would be self-modifying machine code (like a Wheeler Jump). But the computing world at large has nixed that idea for a multitude of reasons
Yes, you can do crazy shit if you try hard enough, but every reasonable programmer would access foo->child->b als foo->child->b and not via that crazy LISPy expression.
By question was: Why would you have a pointer to a memory address that itself only holds a pointer somewhere else?
I’ve not developed anything in C/C++, so I don’t know practical uses for a double pointer, aside from multidimensional arrays, or arrays of pointers
My point was that, conceptually, pointers to pointers is how most complex data structures work. Even if the C representation of said code doesn’t have a int** somewhere
Why do you even have pointers to pointers?
many reasons, for example
In two of your cases this operator is pretty shit because at some point you’ll probably want to offset the access (this isn’t a knock at you but at the feature).
This operator would only really be relevant to the last case which rarely comes up outside of infrastructure/library work (building a tool building tool) and usually those code bases are cautious to adopt new features too quickly anyways for portability.
I’ve done serious C++ work (not much in the past decade though) - while references are absolutely amazing and you essentially want to pass by const ref by default I think well written maintainable C++ should maybe have a dozen heap objects tops. C++ is a better language if you generally forget that pointers and bare arrays exist.
Just again - I think you’re right and the fact that your list is only three things long (and arguably two of them would be misuses) is a pretty clear sign that this is an incredibly niche feature.
I don’t think this operator is a real feature, tbh 😅
The fact that he claims it’s in C++ 29, while we are in 2024 is a good hint.
Or maybe he is a time traveler. Quick, go ask the next lottery numbers!
Mostly because at the lowest level of computing (machine code and CPU instructions), pointers are the only method (that I know of) of any kind of indirection.
At the lowest level, there are 2 types of references:
Every higher level language feature for memory management (references, objects, safe pointers, garbage collection, etc) is just an abstraction over raw pointers
Pointers themselves are really just abstractions over raw integers, whose sole purpose is to index into RAM
With that in mind, pointers to pointers are a natural consequence of any kind of nested object hierarchy (linked lists, trees, objects with references to other objects, etc)
The only other kind of indirection would be self-modifying machine code (like a Wheeler Jump). But the computing world at large has nixed that idea for a multitude of reasons
That’s not a pointer to another pointer, but a pointer to a data structure that happens to contain another pointer.
The distinction is meaningless in the land of Opcode’s and memory addresses
For example, a struct is just an imaginary “overlay” on top of a contiguous section of memory
Say you have a struct
struct Thing { int a; int b; Thing* child; } Thing foo {}You could easily get a reference to
foo->child->bby doing pointer arithmetic*((*((*foo) + size(int)*2)) +size(int))(I’ve not used C much so I’ve probably got the syntax wrong)
Magic.
Blasphemy.
Yes, you can do crazy shit if you try hard enough, but every reasonable programmer would access
foo->child->balsfoo->child->band not via that crazy LISPy expression.By question was: Why would you have a pointer to a memory address that itself only holds a pointer somewhere else?
So far the only reasonable explanation is from @[email protected]:
I’m more talking about theory than practical.
I’ve not developed anything in C/C++, so I don’t know practical uses for a double pointer, aside from multidimensional arrays, or arrays of pointers
My point was that, conceptually, pointers to pointers is how most complex data structures work. Even if the C representation of said code doesn’t have a
int**somewhereBecause stuff can own other stuff and be owned at the same time. Also, arcane jackarsery.
Edit: if you want to give a function a pointer that it may change this may occur in a constructive way. I.e. replace an owned object.
Yeah… But it’s usually a good practice to put a struct somewhere between your 30 levels of ownership.
Exceptions exist, but they are not very common. Also, in C++, operators overloading may help you if you keep needing to write code like this.
In C++ you should never have owning raw pointers. Unless you have a good reason™.
Raw pointers are great, but not for ownership.
I just use
unique_ptr99% of the timeAnd you should.
It even works for classes whose constructors your implementation cannot see, if you aren’t a bitch about it.
Pointers just point to memory addresses, and because pointers are stored in memory addresses, it just kind of naturally falls out that way.
Pointer to pointer is fairly common stuff once you get serious with C++ but I’ve no idea what this abomination is supposed to even be, haha!
Because int main(int argc, char** argv)
Because pointers are real.
That’s the best thing I’ve read in a while, and I’m only on page 2!
James Mickens does good work.
That reads differently after 2024.