Clarifications about Input semantics #83
Description
Hi! Thank you for your work on a Rust yaml parsing crate :)
Splitting this off from #81 because I have a couple of issues that aren't to do with error handling but more questions about the semantics. For context I am trying to wrap a Reader that reads bytes which should be utf8.
The Input trait's documentation says that implementations of skip_n should "Consume the next count character[s]."
Is it also the case that I can assume that lookahead is called prior to skip* calls just like for peek? Or should I handle the case that fewer than N characters are left in the buffer and read more?
Also, regarding lookahead, would be at all possible to start allowing implementers to buffer more than count chars? I ask because right now it is hard to implement a performant Reader implementation since I have to do multiple reads to fill a buffer with count chars, but if I could buffer more than count, I could opportunistically read 4*N bytes and load as many chars as I get from that, all in a single read and buffer the incomplete bytes as well.