BENCHMARKS.md

Benchmarks

The following benchmarks measure crop's performance against Jumprope and Ropey across 4 different use cases:

• creating a rope from a string or using a builder;
• slicing the rope and converting the slice back to an owned Rope;
• iterating over the rope's chunks, bytes, chars, and lines;
• editing the rope by continuously inserting/deleting/replacing text at random offsets;

using 4 different input texts to track the change in performance as the size of the underlying buffer increases:

• tiny.txt: 669 bytes, 9 lines;
• small.txt: 1.5 KB, 20 lines;
• medium.txt: 219 KB, 3086 lines;
• large.txt: 1.5 MB, 21596 lines.

It wasn't possible to add Jumprope to the builder and *_slice benchmarks because it doesn't provide similar APIs, and it wasn't included in the iter_* benchmarks because even though some iterators are implemented, they are not exported publicly.

The xi_rope project was not included in the benchmarks as it's been discontinued together with the rest of the Xi editor (and it was about an order of magnitude slower that all the other ropes on most use cases).

The code used to run the benchmarks can be found here.

All the benchmarks were run on a 2018 MacBook Pro with a (not so) mighty 2.2 GHz 6-Core Intel Core i7.

Creation

cargo bench from_str	cargo bench builder

Slices

cargo bench byte_slice	cargo bench line_slice

cargo bench from_slice

Iterators

cargo bench iter_chunks	cargo bench iter_bytes

cargo bench iter_chars	cargo bench iter_lines

Edits

cargo bench insert_char	cargo bench insert_sentence

cargo bench insert_large	cargo bench delete_char

cargo bench delete_sentence	cargo bench delete_large

cargo bench replace_char	cargo bench replace_sentence

cargo bench replace_large