Abstract:

This video is an interview with Alex Dkovski, a Director of Software Engineering, about his upcoming talk on the practical value of C++ 20 features. Dkovski explains that his talk intentionally focuses on the smaller, often overlooked features of the standard, rather than just the "big four" (ranges, concepts, modules, coroutines), to encourage wider adoption by companies. He discusses his personal experience, highlighting concepts as the most impactful feature for generic programming and the spaceship operator for its ability to drastically reduce boilerplate code. Conversely, he notes that modules remain impractical due to a lack of mature compiler and build-system support. The conversation also explores future standards, with Dkovski expressing great enthusiasm for reflections in C++ 26 for code generation and skepticism about the complexity and potential performance impact of contracts. Additional topics include his use of the Conan package manager, his perspective on AI as a helpful but still limited development assistant, and a concluding thought on prioritizing code safety over raw performance.

Unlocking the Power of C++20 and Beyond: An Interview with Alex Dkovski

• 00:01:43 Talk Focus - Smaller C++20 Features: Alex's talk is designed to highlight the smaller, yet powerful and practical features of C++ 20 that developers often miss. His goal is to demonstrate their value and encourage companies to migrate to the now-stable C++ 20 standard, which he argues makes developers' lives significantly easier.
• 00:05:36 The Challenge with Modules: Alex confirms that modules, one of the "big four" features, are still not practically usable in a production environment. This is primarily due to lagging support in build systems and compilers, making them the one major feature he has been unable to implement.
• 00:07:20 Advocating for std::format: He is a strong proponent of using std::format over the external {fmt} library. His key reason is the importance of relying on the standard library whenever possible to reduce external dependencies.
• 00:13:45 The Impact of the Spaceship Operator (<=>): Alex praises the spaceship operator as a great feature for simplifying code. He notes that by using it, he was able to remove hundreds of lines of boilerplate comparison code from his projects.
• 00:14:51 Most Valuable C++20 Feature - Concepts: For Alex, concepts are the most useful feature of C++ 20. He describes them as a "jewel in the crown" for anyone engaged in heavy template metaprogramming or generic programming, as they dramatically improve code clarity and compiler errors.
• 00:10:51 Excitement for C++26 Reflections: Looking ahead, Alex is extremely enthusiastic about the upcoming reflections feature in C++ 26. He has experimented with it using special compiler builds and sees its massive potential for automating code generation and eliminating boilerplate, especially for tasks like creating Python bindings.
• 00:22:05 Skepticism Towards Contracts: While excited about reflections, Alex is wary of contracts. He fears they will add complexity, negatively impact runtime performance, and create fragmentation in the ecosystem if their use is inconsistent across different libraries.
• 00:16:33 Managing Dependencies with Conan: Alex recommends using Conan (specifically Conan 2) as a package manager for C++ projects. He finds it easy to use and effective for creating and managing internal and external dependencies.
• 00:31:14 Using AI as a Development Assistant: Alex uses AI tools like ChatGPT and Claude as assistants for tasks like writing scripts, generating Conan packages, and finding simple bugs. However, he cautions that AI is still not reliable for complex C++ code and that developers must have the expertise to verify its output.
• 00:38:52 Prioritizing Safety Over Performance: When faced with the classic trade-off, Alex's stance is firm: safety should always come first. He warns that relying on undefined behavior for performance is a dangerous practice, as code can break unexpectedly with compiler updates. His motto is that it is better to be correct than "fast and wrong."