Lab work 3: Custom Vector, Array & Smart Pointers

Authors (team): Anastasiia Pelekh @drakonchyk

---

Prerequisites

Tool / Library	Minimum version	Why it’s needed
C++ compiler	GCC ≥ 11 / Clang ≥ 14 / MSVC ≥ 19.31	C++20 language features
CMake	3.20	Cross-platform build system
Make / Ninja	—	Backend for CMake (choose one)
GoogleTest	v1.14 (fetched automatically)	Unit-testing framework
(Optional) Valgrind / LLVM ASan/UBSan	latest	Memory-error & UB detection
(Optional) PVS-Studio	7.x	Static analysis

---

Compilation

git clone https://github.com/ucu-cs/lab3-my-vector-__pelekh-1.git
cmake -S . -B build              # configure
cmake --build build -j           # compile library, example & tests

Tip: during development enable sanitizers
cmake -S . -B build -DCMAKE_CXX_FLAGS="-fsanitize=address,undefined -g"

Installation

No global install is required — the project builds a header-only library. If you still prefer make install:

cmake --install build --prefix $HOME/.local

Usage

#include "my_vector.hpp"
#include "my_array.hpp"
#include "my_smart_ptrs.hpp"

int main() {
    my_vector<int> v = {1, 2, 3};
    v.push_back(4);

    my_array<std::string, 3> a{{"one", "two", "three"}};

    my_unique_ptr<int> up(new int(42));
    my_shared_ptr<int> sp1(new int(7));
    my_shared_ptr<int> sp2 = sp1;

    std::cout << v.back()          // → 4
              << " " << a[0]       // → one
              << " " << *up        // → 42
              << " " << sp1.use_count(); // → 2
}

Build & run the example:

cmake --build build --target example
./build/example

Run unit tests:

ctest --test-dir build/tests --output-on-failure

GENERAL BUILD

mkdir build
cd build
cmake .. && make
./example
./tests/tests

---

Important!

• Header-only: include paths are enough — no linking against a compiled library.
• Smart-pointer implementation is minimal and not thread-safe beyond the demonstrated use-case (atomic ref-counts only).
• my_vector provides strong exception-safety for all mutating ops but is ~10 % slower than std::vector in push-back micro-benchmarks due to manual memory management.

---

Results

• Implemented from scratch:
  • my_vector<T> — dynamic array with iterator support
  • my_array<T,N> — fixed-size aggregate array
  • my_unique_ptr, my_shared_ptr, my_weak_ptr
• GoogleTests cover fundamental, complex and nested types.
• Benchmarks show ~1.8× speed-up over naive realloc-every-push strategies; still within 5-15 % of libstdc++ std::vector.

What I learned:

• Low-level manual allocation (operator new[] / placement new).
• Exception-safe move/copy semantics & RAII.
• Intricacies of reference counting and weak ownership.
• Writing allocator-agnostic, STL-compatible iterators.

---

Additional tasks

• pointers.

---