optimise: stores char data as bytes instead of their code representation

[Samy-33]

No description provided.

[jeaye]

Two stylistic points.

1. We always use /* */ comments in jank. We also use correct grammar and punctuation, since comments are documentation. So we'll want a period at the end of this.
2. This for loop is more idiomatically written like this:

for(size_t i{}; i < 2; ++i)

In particular, a few things change:

1. We use size_t because we're counting up from 0
2. We use direct initialization, as we do in modern C++
3. We zero-base our indices/loop counters
4. We name the counter i, even if it's ignored
5. We use ++i instead of i++; while there's no practical difference here, the postfix increment has serious performance implications in other scenarios and we want to only use it when we actually want postfix; here, we actually want prefix increment

[Samy-33]

Could you point me to a resource for why i++ would have perf implications?

[jeaye]

The ++ operator in C++ is overloadable. We do this regularly for custom iterator types. For more complex iterator types, copying the iterator may be very slow. i++ requires copying i so it can be returned and then incrementing the actual i. So, imagine an iterator which holds a vector and an index in it. By doing i++, you need to copy the iterator, which means deep copying the entire vector. If you use ++i, the iterator is updated in place.

As I said, in this situation, the compiler will treat i++ to mean ++i because clang knows that people like to write i++ when they mean ++i. But C++ is all about building good habits so that you don't get bitten. By using ++i as a default, and only i++ when you actually want a copy of the value before it was incremented, you won't be surprised down the road.

[jeaye]

Since you asked for a reference:

• https://stackoverflow.com/questions/5223950/stl-iterators-prefix-increment-faster
• https://stackoverflow.com/questions/41765835/why-use-the-prefix-increment-form-for-iterators

[jeaye]

Nice work, Saket!