Questions about implementation and experience?

[Ygg01]

Let me preface this as this project is awesome and I quite enjoyed reading through code. I'm reading cause I'm intrested in SIMD-ifying my YAML parsing crate.

But I got few questions about the implementation details, or link me to some alternate discussion board.

Here are the questions:

1. Why are structural indexes:
   

   simd-json/src/lib.rs

   Lines 779 to 781 in b249c70

   	pub(crate) unsafe fn find_structural_bits(
   	input: &[u8],
   	structural_indexes: &mut Vec<u32>,

   
   Vec<u32> and not Vec<u64>?
   1. This means largest JSON parsed (assuming UTF-8 encoding) is 4GB, right?
   2. Is it for performance reasons or something else I might have missed?

2. Does using macros like this one:
   

   simd-json/src/stage2.rs

   Lines 128 to 144 in b249c70

   	macro_rules! s2try {
   	($e:expr) => {
   	match $e {
   	::std::result::Result::Ok(val) => val,
   	::std::result::Result::Err(err) => {
   	// We need to ensure that rust doesn't
   	// try to free strings that we never
   	// allocated
   	unsafe {
   	res.set_len(r_i);
   	};
   	return ::std::result::Result::Err(err);
   	}
   	}
   	};
   	}

   Offer benefits over using an #[inline] on function?

3. In Readme.md you mentioned something about a We work around some performance bottlenecks imposed by safe rust Does this means the indexing penalty you pay for safe array access?