feat: add g_ motion (move to last non-blank character)

[legalcodes]

Manual Review

I tested using this motion in the replit link -- it worked.

Adds the g_ motion, which moves the cursor to the last non-whitespace character on the line.

Behavior

• g_ moves to the last non-blank character of the current line
• 2g_ moves to the last non-blank of the next line (count - 1 lines down)
• Works with operators: dg_ deletes to the last non-blank, preserving trailing whitespace
• Inclusive motion, matching vim's native behavior

This is the counterpart to ^ (first non-blank), which is already implemented. Currently g_ is unmapped and does nothing.

Implementation

• New helper findLastNonWhiteSpaceCharacter() mirrors the existing findFirstNonWhiteSpaceCharacter()
• New motion moveToLastNonWhiteSpaceCharacter registered with inclusive: true
• Tests cover basic movement, no-trailing-whitespace lines, count support, and operator interaction

Test plan

• Added automated tests in vim_test.js
• Manual verification via web demo

Open web-demo @ d6eeff49181ef1b2a41ff8f40aba9e6ee427cc96

[Copilot]

Pull request overview

Adds Vim’s g_ motion to move the cursor to the last non-blank character on a line (inclusive), including count support and operator compatibility, aligning behavior with existing whitespace-aware motions.

Changes:

• Register g_ as an inclusive motion mapped to moveToLastNonWhiteSpaceCharacter.
• Implement moveToLastNonWhiteSpaceCharacter and findLastNonWhiteSpaceCharacter() helper.
• Add automated tests covering basic movement, count usage, and dg_ behavior.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
src/vim.js	Adds the g_ keybinding, implements the motion, and introduces a helper to find the last non-whitespace character.
test/vim_test.js	Adds a new test case for g_ movement, count behavior, and delete-operator interaction.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

moveToLastNonWhiteSpaceCharacter computes line = head.line + motionArgs.repeat - 1 and immediately calls cm.getLine(line) without clipping the line number to the document bounds. With a large count (e.g. 100g_ near EOF) this can produce an out-of-range line and return a Pos with an invalid line, which then flows into operator selections (dg_, etc.) without being clipped and can break selection logic. Clamp the target line to [cm.firstLine(), cm.lastLine()] (or use cm.clipPos(new Pos(line, 0)).line) before calling getLine and before returning the Pos.

Suggested change

	var lineText = cm.getLine(line);
	var lastNonWS = findLastNonWhiteSpaceCharacter(lineText);
	return new Pos(line, Math.max(0, lastNonWS));
	var clippedLine = cm.clipPos(new Pos(line, 0)).line;
	var lineText = cm.getLine(clippedLine);
	var lastNonWS = findLastNonWhiteSpaceCharacter(lineText);
	return new Pos(clippedLine, Math.max(0, lastNonWS));

[Copilot]

The new g_ tests cover a small count (2g_) but don’t cover boundary behavior when the count would move past the end of the document. Since g_ is countable, adding a test like 100g_ on a short buffer (and asserting it lands on the last line’s last non-blank without throwing) would help lock in correct clipping behavior.

[masad-frost]

question: do we need the conditional linewise case for counted operators here? In Vim, d2g_ from the first nonblank of a\nlast \nthird deletes the first two complete lines, while this implementation leaves the trailing spaces from the second line. Starting inside the first line remains characterwise, so this seems to need a conditional rather than linewise: true on the mapping.

~ written by Zerg 👾 (valiant-disruptor-aba4)

[masad-frost]

question: can this return the start of the final code point? With x😀 from column 0, this returns the low-surrogate index. Cursor clipping then lands on the first trailing space, so vg_ includes that space; Vim stops on the emoji.

~ written by Zerg 👾 (valiant-disruptor-aba4)