[panarch]

Summary

• edtui's built-in word motions use is_ascii_alphanumeric(), causing CJK characters to each be treated as a separate word boundary. This replaces them with custom Unicode-aware implementations using char::is_alphanumeric().
• Fixes vim-incompatible behaviors: de deleting trailing whitespace, db including cursor char, w not skipping leading whitespace on line crossings, b going to last char instead of word start on line crossings.
• Removes insert-mode Alt+F/Alt+B word motions and their keymap help entries.

Test plan

• cargo clippy --all-targets -- -D warnings passes
• cargo test --workspace — 71 tests pass (19 new: 8 CJK + 11 word motion)
• Manual testing with Korean text for w/e/b/diw/ciw

Summary by CodeRabbit

• Improvements

  • Enhanced Unicode/CJK-aware word navigation and selection for more accurate cursor movement across languages and lines.

• Changes

  • Removed Alt+F and Alt+B word-navigation shortcuts in insert mode and updated the displayed keymap accordingly.

• Tests

  • Added extensive automated tests covering word-motion, cross-line behavior, and CJK scenarios.

[coderabbitai]

📝 Walkthrough

Walkthrough

Refactors editor word-motion into Unicode-aware helper functions, removes Alt+F/Alt+B insert-mode bindings and keymap entries, updates normal/visual mode transitions to call the new helpers, and adds extensive word-motion tests (including CJK) plus small Tester helpers for assertions.

Changes

Cohort / File(s)	Summary
Word Motion Infrastructure tui/src/transitions/notebook/textarea.rs	Added Unicode-aware word-motion subsystem: char_class, skip_ws/skip_ws_rev, max_col_for_mode, update_visual_selection, forward/backward motions (move_word_forward, move_word_forward_to_end, move_word_backward), and select_inner_word. Exposed as pub(super) for local use.
Mode Handler Refactors tui/src/transitions/notebook/editing_normal_mode.rs, tui/src/transitions/notebook/editing_visual_mode.rs	Replaced direct MoveWord* action executions with calls to the new helper functions; adjusted delete/selection flows (e.g., DeleteInsideWord, DeleteWordBack) to use selection helpers. Removed references to removed MoveWord* enum variants.
Insert-Mode Keybindings & Keymap Display tui/src/context/notebook.rs, tui/src/views/dialog/editor_keymap.rs	Removed Alt+F / Alt+B handling and the corresponding keymap display entries for Alt-word navigation in insert mode; deleted related imports/usages.
Tests & Test Utilities tui/tests/editor_word_motion.rs, tui/tests/editor_cjk.rs, tui/tests/tester.rs	Added comprehensive word-motion tests (including CJK and cross-line scenarios) and two Tester helpers: cursor() and editor_text() for inspecting editor state in tests.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related PRs

• Replace tui-textarea with edtui for notebook editor #286 — overlaps: replaces/removes MoveWord* action variants and introduces/uses move_word_forward / move_word_backward / move_word_forward_to_end helpers.
• Introduce Tester and simplify TUI integration tests #197 — related: changes to test harness and Tester API (test helpers for inspecting editor state).

Suggested labels

codex

Poem

🐰 I hop through words in Unicode light,
Skipping spaces, left and right,
Alt-keys trimmed, helpers sing,
Tests verify each CJK spring,
Little rabbit cheers the night ✨

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately and specifically summarizes the main change: implementing Unicode-aware word motion with specific vim commands (w/e/b/diw/ciw/de/db).
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch unicode-word-motion

---

No actionable comments were generated in the recent review. 🎉

🧹 Recent nitpick comments

tui/src/transitions/notebook/textarea.rs (1)

121-132: skip_ws_rev relies on implicit synchronization between the iterator position and idx.col.

The rev().skip(…) pattern works correctly but is fragile: the correctness depends on skip being exactly line.len() - (idx.col + 1) so the first yielded character corresponds to idx.col, and each subsequent character lines up with the decrement. Consider a simpler formulation that makes the index tracking explicit:

♻️ Optional: explicit index loop

 fn skip_ws_rev(lines: &Lines, idx: &mut Index2) {
     if let Some(line) = lines.get(RowIndex::new(idx.row)) {
-        let skip = line.len().saturating_sub(idx.col + 1);
-        for ch in line.iter().rev().skip(skip) {
-            if !ch.is_whitespace() {
+        while idx.col > 0 {
+            if !line[idx.col].is_whitespace() {
                 break;
             }
             idx.col = idx.col.saturating_sub(1);
         }
     }
 }

Note: this simplification omits the check for the initial position (idx.col) itself — you'd need to decide whether the current position should also be tested. The existing code does test it, which matches the caller's expectation (cursor is already at col - 1 when skip_ws_rev is called). Either way, making the index explicit avoids a subtle coupling.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[gluesql-coverage-bot]

Glues Coverage Report

• Commit: a9221f6e36b59ef9d82d72c54f02d1c6c0aac29c
• Timestamp: 2026-02-08T162924Z
• Report: View report

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@tui/src/transitions/notebook/textarea.rs`:
- Around line 235-290: The recursive upward moves in move_word_backward_once
(calls to itself from the col==0 branch and the whitespace-at-col-0 branch after
skip_ws_rev) can cause deep recursion on many blank lines; refactor this
function to an iterative loop: use a while loop that moves idx/ editor.cursor up
by one row when at col 0 or when whitespace at col 0 instead of recursing,
reusing skip_ws_rev and the same char_class/backward-scan logic (Index2,
editor.lines, editor.cursor) until you find the word start or reach the buffer
start; ensure each iteration updates idx/cursor and continues rather than
calling move_word_backward_once.

🧹 Nitpick comments (2)

tui/tests/editor_cjk.rs (1)

8-23: setup() helper is duplicated from editor_word_motion.rs.

This helper is identical to the one in editor_word_motion.rs (lines 8–21). Consider extracting it into tester.rs as a shared method (e.g., Tester::setup_editor_with_text) to avoid the duplication. Not blocking since test-file-level helpers are common.

tui/src/transitions/notebook/textarea.rs (1)

122-132: skip_ws_rev iterates beyond the logical stop point when idx.col saturates at 0.

When all characters from idx.col down to 0 are whitespace, idx.col saturates at 0 but the iterator continues visiting remaining characters (all of which are whitespace), calling saturating_sub(1) as a no-op each time. The result is still correct (caller handles the "still on whitespace" case at lines 259-266), but the extra iterations are wasted work.

An early break when idx.col == 0 would short-circuit cleanly:

Proposed fix

         for ch in line.iter().rev().skip(skip) {
             if !ch.is_whitespace() {
                 break;
             }
+            if idx.col == 0 {
+                break;
+            }
             idx.col = idx.col.saturating_sub(1);
         }