Hacker News Reader

Comment by pmiller2 | Hacker News Reader

I will always upvote something like this, but I am wondering: why write it in C++? I didn't really read the whole thing, but I skimmed a bit, and didn't see a lot of advantage taken of the higher-level facilities of C++. It seems like you've picked a route that gives you all the disadvantages of using C, while simultaneously destroying the possibility of it ever being self-hosting.

Still, a fun project, and you have my upvote. :)

View on HN

Not everyone wants to write the most idiomatic C++, so writing variably C-ish code in C++ isn't really all that uncommon nowadays. If you're a C programmer but you notice C++ has a convenient feature (or features) that you might find useful in a pinch, hell, maybe it's worth it to bite the bullet (although that's not to say that using C++ in lieu of C can have its drawbacks, though)

I would completely agree with you if this weren't a C compiler project. In this case, the advantage of having a self-hosting compiler is huge! Among other things, you can use the compiler source itself to test the compiler, which is pretty cool.

I have worked on projects that were essentially "C with classes," and it was fine. I'm not saying every C++ project needs to go all out using every feature from C++ 11, but I didn't even see much beyond use of the string class here. It seems not worth giving up self-hosting to me.

> the advantage of having a self-hosting compiler is huge!

I agree that it's kind of neat to compile your compiler with itself, but it's not clear to me that this constitutes a significant productivity boost. There are countless other C programs that have already been written and can be used as test cases.

So I'm left wondering what the huge advantage that self-hosting provides is. I am interested so, if you have more to say on the topic, I will certainly read it.

If we're talking about toy compilers and learning, then productivity isn't really in consideration. Having a self-hosting compiler for a non-trivial language means you've achieved the "bootstrap" stage, which is a significant milestone and really gives you a practical understanding of what's involved in developing a programming language. A compiler that is powerful enough to compile itself is on the way to exiting the "toy"/"theoretical" realm that most people who study compilers don't ever go beyond.

You could just start to replace C++ things by plain C until it is self-hosting.

while simultaneously destroying the possibility of it ever being self-hosting.

...at least until it begins turning into a C++ compiler ;-)

That makes me wonder, given that there seem to be quite a lot of "toy" C/C-subset compilers being written these days, just how much harder C++ would be. For a long time it was commonly thought that even the simplest C compiler would be extremely complex and difficult, but that notion appears to have been somewhat defeated, so the next logical step in that direction is C++ or maybe C++-to-C (like Cfront) compilers. The closest I'm aware of would be the C++-subset used in TempleOS, which is self-hosting.

Horribly. C++ has so many features, and interactions between them, that you really don't want to be dragged into that as a toy project. I'd wager that the C++ language spec is longer than the source of a functional C compiler (both without the standard libraries).

Maybe the interactions aren't actually the issue because the features are quite orthogonal and independent of each other, and the complexity is really the natural result of the interactions between them. If I wanted to write a C++ compiler I'd probably start by adding features in this order:

    - Member functions in structures; constructors; destructors
    - Single inheritance
    - Virtual functions
    - Exceptions
    - Multiple inheritance
    - Virtual inheritance
    - Templates

There's definitely more I haven't listed, but the first few don't seem too difficult...

I was wondering the same thing... ?? Oh, and you get an upvote from me too.

Having a self hosting compiler gives a very handy tool to assess the gains of a speed optimisation: if the speed improvement of the compiler compensate the more complex compilation algorithm, then it is justified.

Reply on news.ycombinator.com