text-vs

This is an experiment in benchmarking text editing routines.

Current version is not benchmarking anything.

To start just type:

$ make

It will:

1. download Linux kernel sources and will checkout v4.9 tag,
2. find the top 10 changed files list and save it,
3. get whole diff set for the first file and save it,
4. compile diff2edit program and run it with the diff set file

diff2edit

The program in current version will just convert the diff set to easier to parse format.

The final version of the program will try to produce minimal set of editing commands in simple binary format. The resulting set would be an aproximation of a probable session with a text editor. Or an undo buffer.

Benchmark

What will be benchmarked?

• array of lines (strings) - werf text editing routines
• librope based implementation
• array of lines (array of fragments) - idea similar to ropes, but using arrays instead of trees

How?

Every implementation benchmark will load resulting editing commands set from diff2edit.

The time of the load will be measured. Then iteration through file will be measured.