Thread: How people tackle C limitations?

Some things you can fake up in C, other things you can't do without effectively writing an interpreter for another language in C. The main C limitation versus C++ that can't be overcome is the lack of destructors. In C you have to free memory you allocate explictly. So you can implement a "closure" in C (a functionpointer and a void pointer), but you can't destroy the data the void pointer points to except from the place you set it up. So in practice C "closures" are quite restricted.

You can implement OO with structures containing function pointers. That's a C++ "virtual" function. You can also give structures a field giving the structure size. That allows for future expansion without affected existing code. Microsoft do this a lot in the Windows API.

RAII is easy to do (just have function which allocates the resources and returns a pointer). But the corollary is DIRE (deletion is resorouce expiry), and that you can't do as easily. You have to write a "kill" function and call it explictly.

You generally don't need templates in C because all the data is what C++ calls "plain old data". Functions like qsort() and memcpy() work fine
with C, but not with C++ classes.

There are generics in the latest C standard. However generally treating all your data as double and using ints for counts of thing sin memory or index values works very well.

Originally Posted by thmm

That sounds like a contradiction to me. RAII is based on destructors.

You need to read the next sentence for the context, i.e., Malcolm McLean is trying to address the criticism that the name "RAII" is misleading by giving it a more limited meaning and thus is able to contrast it with "DIRE" as a corollary, whereas the traditional definition combines both. As you have experienced, the attempt to avoid confusion is itself confusing.