Chore: simplify and improve performance for autofix

[mysticatea]

What is the purpose of this pull request? (put an "X" next to item)

[X] Other, please explain:

What changes did you make? (Give an overview)

This PR refactors the way of autofix.

The current way is:

1. Makes the array of characters.
   • This creates many string instances.
2. Applies fixes to the array with Array#splice in reverse.
   • Array#splice is a O(n) operation, then it's repeated as many times as the number of fixes.
3. Joins the array.

The new way of this PR is:

1. Makes the fixed string by concatenation (+=). It concatenates "the part followed by a fix" and "the text of the fix" by turns.

As a result, the performance of autofix improved from O(n×m) to O(n+m). (n is the number of characters, m is the number of fixes)
I measured it by this script then I saw the new way gets 25% faster than the current way.

Is there anything you'd like reviewers to focus on?

• This PR changes the order of fixing. This is reverse of the current order. So I needed to modify tests for overlapped fixes. However, we are doing multipass fixing, so I think it's no problem.

[mention-bot]

@mysticatea, thanks for your PR! By analyzing the history of the files in this pull request, we identified @not-an-aardvark, @platinumazure and @nzakas to be potential reviewers.

[eslintbot]

LGTM

[not-an-aardvark]

Can start and end be defined in a smaller scope? (They don't seem to be used outside of the for-of loop.)

[not-an-aardvark]

I think this will fail to detect a conflict between these two fixes, because there is a special check for lastPos > 0:

({ range: [0, 0], text: "foo" });
({ range: [0, 0], text: "bar" });

lastPos will be 0 after applying the first fix, so lastPos > 0 will be false and the fix will be applied.

Maybe it would be better to set the initial value of lastPos to -Infinity instead.

[mysticatea]

Good catch!

[eslintbot]

LGTM

[mysticatea]

I modified perf-fix2.js to avoid spawn. This is more accurate than perf-fix.js to measure performance of autofix part.

Then, it's about ×2 faster.

[nzakas]

LGTM, just want another set of eyes to review.