Get strings by reference

[d-unsed]

• [x] RString::to_str(&self) -> &str
• [x] RString::to_str_unchecked(&self) -> &str

These methods are similar to to_string() and to_string_unchecked(). The difference is that they return immutable references to underlying Ruby strings.

The new methods do not allocate new Strings on heap, thus they should be faster when a user just needs to get underlying data from Ruby string.

[d-unsed]

Methods

Method	Heap allocation*	Null-bytes check**
to_string	✓	✓
to_string_unchecked	✓	✗
to_str	✗	✓
to_str_unchecked	✗	✗

* heap allocation means that the contents of the original Ruby string are duplicated to a new heap-allocated Rust String

** null-bytes check means that before retrieving an underlying string from a Ruby string, MRI will perform a check for interior null bytes. For example, running some_str.to_string() on some_str = "Hello, \0World" will raise an exception string contains null byte (ArgumentError).

Benchmarks

Some quick benchmarks of these methods. I used MacBook Pro Mid 2014 (2.2 GHz Intel Core i7, 16 GB RAM), Ruby 2.4.0 and Rust 1.14.

(the more iterations per second the better)

• to_str_unchecked() has more or less constant throughput for any string length;
• to_string() has the lowest throughput, because it does both heap allocation and null-bytes check;
• to_str() performs better than to_string_unchecked() on the strings shorter than 512 bytes (possibly, because heap allocation is more expensive than null-bytes checks for shorter strings);
• to_string_unchecked() performs better than to_str() on the strings longer than 512 bytes (possibly, because heap allocation is less expensive than null-bytes checks for longer strings);

[Zapotek]

That is excellent, can't wait to see what difference it makes under real world circumstances.