In Defense of C++ | slacker news

When it comes to programming, I generally decide my thoughts based on pain-in-my-ass levels. If I constantly have to fiddle with something to get it working, if it's fragile, if it frequently becomes a pain point - then it's not great.

And out of all the tools and architecture I work with, C++ has been some of the least problematic. The STL is well-formed and easy to work with, creating user-defined types is easy, it's fast, and generally it has few issues when deploying. If there's something I need, there's a very high chance a C or C++ library exists to do what I need. Even crossing multiple major compiler versions doesn't seem to break anything, with rare exceptions.

The biggest problem I have with C++ is how easy it is to get very long compile times, and how hard it feels like it is to analyze and fix that on a 'macro' (whole project) level. I waste ungodly amounts of time compiling. I swear I'm going to be on deaths door and see GCC running as my life flashes by.

Some others that have been not-so-nice:

* Python - Slow enough to be a bottleneck semi-frequently, hard to debug especially in a cross-language environment, frequently has library/deployment/initialization problems, and I find it generally hard to read because of the lack of types, significant whitespace, and that I can't easily jump with an IDE to see who owns what data. Also pip is demon spawn. I never want to see another Wheel error until the day I die.

* VSC's IntelliSense - My god IntelliSense is picky. Having to manually specify every goddamn macro, one at a time in two different locations just to get it to stop breaking down is a nightmare. I wish it were more tolerant of having incomplete information, instead of just shutting down completely.

* Fortran - It could just be me, but IDEs struggle with it. If you have any global data it may as well not exist as far as the IDE is concerned, which makes dealing with such projects very hard.

* CMake - I'm amazed it works at all. It looks great for simple toy projects and has the power to handle larger projects, but it seems to quickly become an ungodly mess of strange comments and rules that aren't spelled out - and you have no way of stepping into it and seeing what it's doing. I try to touch it as infrequently as possible. It feels like C macros, in a bad way.

If you are willing to give up incremental compilation, concatenating all C++ files into a single file and compiling that on a single core will often outperform a multi-core compilation. The reason is that the compiler spends most of its time parsing headers and when you concentrate everything into a single file (use the C preprocessor for this), it only needs to parse headers once.

Merely parsing C++ code requires a higher time complexity than parsing C code (linear time parsers cannot be used for C++), which is likely where part of the long compile times originate. I believe the parsing complexity is related to templates (and the headers are full of them), but there might be other parts that also contribute to it. Having to deal with far more abstractions is likely another part.

That said, I have been incrementally rewriting a C++ code base at a health care startup into a subset of C with the goal of replacing the C++ compiler with a C compiler. The closer the codebase comes to being C, the faster it builds.