The C style memory allocation function free() takes only a pointer to the memory to be deallocated. Theoretically, if additional information could be passed, the memory allocator could be optimized. In particular, if the size of memory requested was passed, that could prevent the memory allocator from needing to store that information. Rust went even further with there memory allocation API and required that the exact size and alignment used when allocating the memory be passed when deallocating. However, they seem to be ignoring issues where the memory size isn't correctly known. They give an example of someone allocating extra memory and constructing a Vec from it. The Vec would then attempt to deallocate the memory using the alignment it would have used had it been the one to allocate it. That may not match the value actually used. Adding an alignment value to every Vec seems like a bad idea. However, there are cases when even the size won't be well known. Imagine consuming a list to construct an array. In that case, one might want to not reallocate the memory of the list, but simply reuse it as the memory for the array. But since the capacity of the list could be bigger than the size of the array, the array type would not know the proper size to deallocate the memory with. For these reasons it seems onerous to require the deallocate operation to require parameters besides the pointer to the memory to be deallocated. Given that Adamant is a high level language and that this was good enough for C it makes sense to go with that and not spend further time on it. The same argument applies to not requiring the previous size or alignment when reallocating.
This repository has been archived by the owner on Nov 18, 2020. It is now read-only.