A friend, asking me what I was up to these days, was surprised that I'd use C++ to start a new project. We've written a lot of C++ together, going back to the end of 80s. So this made me realise it wasn't as obvious as I might have thought, hence this post.

vs CSharp

His main thing was "why don't you use something like C# and be rid of managing memory ?". I don't feel I'm having any problems managing memory in C++. For ages, constructors, destructors and scopes have been my main tool to ensure resources (not just memory) get tracked properly. Also std::unique_ptr and std::shared_ptr are here to keep the more involved situations under control. As a matter of fact, the idea that I know precisely and for sure when resources get disposed of is one of my main reasons for liking C++.

vs Javascript

This may sound as an odd point to bring up, here's why: if you want to write code that will run on pretty much any platform, target web browsers. So, of course, I considered it: there is no reason I couldn't replicate most of what I had written over ten years ago, on today's machines, in a browser. I actually did experiment. I ran into two main problems: UI/DOM and sandboxing. Running inside the browser means a lot of your UI is decided by the browser, you have to relinguish control over things like the right-mouse button (workarounds are possible but tricky). The sandboxing, though, is the dealbreaker: you cannot access the machine from a web page, this is a problem when your goal is to index the user's content. The sandboxing is an absolute necessity in a browser, so JS is just out of the question. (... sigh of relief)

vs Java, Rust, Go, Swift or otherwise

For Javascript, there is the argument of the ubiquity of the browser. For other languages, so far, there isn't an argument or feature that'd make me want to go further than experimenting.

Innovative UI

I strongly believe that applications would benefit from building UI's like we do in videogames. I am going to dive again in a pan&zoom UI, a very dynamic and reactive experience. The easiest way to do this is to go down to the graphics API and handle resources at that lower level. The Windows version runs on Vulkan, the OSX version on Metal ; I have built a version on the Raspberry Pi with Vulkan, so Linux should be covered too. That's a great advantage of C++ : it runs on pretty much anything.

Libraries have C interfaces

There are a few libraries that make a project like Girolamo possible, handling basic needs. They are all written in C++ or C, they're all trivial to integrate in a C project:

• libuv: the lib that makes node.js asynchronous
• sqlite: easiest way to store large amounts of structured data inside one file
• libcurl: access (that works) to internet resources
• opencv: everything computer vision
• llama.cpp: I don't need to introduce this, do I ?
• ...

I am going to talk more about all the tools I use in later posts similar to this one, but the point is clear: C++ has all the toys !

Conclusion

I have been away from C++ for a long time because it wasn't a good option to build for the new platforms of the time: iOS and Android. I have tried many things and, I must admit, I missed C++. The best reason I can give for choosing C++ as the tool of choice for Girolamo is this: I enjoy it.