ENHANCED: Read more characters at once, significantly speeding up library(pio)

[triska]

Enjoy!

[ghost]

Why read one byte at a time? There is take. And for the name string may be matter than str (less confusion with the type str).

[triska]

Yes, I noticed take, however I found the description unclear in case of errors ("may succeed": what is that supposed to mean, that it may also fail even if the error is resolved?).

In fact I chose str also to avoid confusion with the type String!

[ghost]

There is the type str used as &str or Box<str> but for string there isn't such type but it's close to String.

The current implementation is like take, on the first error it halts but that doesn't mean EOF has been reached.

[triska]

Thank you @notoria for these suggestions, I have addressed these issues in additional commits.

Please review, and merge if applicable! Thank you again.

[ghost]

I don't have the numbers but an iterator could be faster (and nicer?). Something like this, it could be done for both binary and text mode. With push there could be reallocation.

[triska]

I suggest to leave micro-optimizations to contributors who want to measure the performance of such changes, and ideally address this in Rust itself where possible.

[triska]

I have rebased the changes against the current master branch!

One additional note: The next and final big step to improve efficiency is to map files without 0-bytes directly into memory using mmap (see #251 ). However, for now, reading in larger chunks as implemented in these commits already improves performance significantly, and also directly uses the compact internal representation for strings.