Beagle SCM

LITE-023 DONE: active line/token in the lite pager + keyboard nav

gritzko's ruling 2026-08-14: the lite pager gets a notion of an ACTIVE line and an ACTIVE token, rendered as a colour wash (bg lightness shift for the line, a stronger second wash for the token — composes over the diff/why washes, never a hue that drowns them). Keyboard then navigates the cursor, not just the scroll: Enter follows the ACTIVE token's U target when there is one, falling back to the active row's entry (today Enter only fires on the TOP row, _followRow(v.scroll)). A mouse click sets the cursor before following. lite/ pager ONLY for now — be's pager and C bro follow later, on their own tickets.

Input

Context

lite/view/pager.js (LITE-002) has scroll-only state: view = {hunks, path, rows, scroll, cols, wrap}; Enter/click follow via _followRow/_uriAt (hidden U toks), _screenToByte maps a click to a byte. lite/view/bro.js paints rows with attr composition (aOr, the LITE-010 diff wash slots) — the wash machinery the active line/token rendering rides on.

• Current keys (SHORTCUTS/_keyScroll): q, j/k scroll, space/b page, g/G, m mouse, : bar, Enter, -/BS back, R/r, w wrap, h help.
• A row knows its hunk + byte range ({hunk, off, end}); tokens per row come from the hunk's tok32 stream, same walk _screenToByte does.
• The cursor is render-side state; hunks/toks stay untouched — no new tok tags.

Goals

Cursor state + wash rendering + keys, in lite/view/pager.js (+ a paint hook in lite/view/bro.js if the row painter needs an "extra attr" arg).

• view.cur = {row, tok}: active display row, active token index within it (tok optional — a line with no tokens is still an active line).
• Active line: a subtle bg LIGHTNESS wash OR'd over every cell of the row at paint time; active token: a stronger wash on its span. Both compose with existing washes (diff in/rm, why blame) — pick shifts, not hues.
• Token hops land only on FOLLOWABLE tokens — a visible token with a hidden U follower, or an F nav-target tok (ticket codes / file refs, the LITE-015 click path) — skipping plain text; a line with none has no active token.
• Enter follows the active token exactly as a click on it would (U target, F door hand-off), else the active row (the _followRow fallback); click = set cursor to the hit cell, then follow as today.
• Cursor moves scroll the viewport when it walks off-screen; scroll alone never moves the cursor (the two are independent, vim-window style).
• Status line may show the active target's name/path (the _uriAt string) instead of the top-row #L.
• Help hunk (h) regenerates from the ONE SHORTCUTS table, per BRO-007.

Constraints

• lite/ ONLY: lite/view/pager.js, lite/view/bro.js, lite/test/. No jsrc/, no jab/, no C. No VC ops.
• Verify through the real UI path: pty probe (LITE-004 style) showing the wash moving under the keys and Enter following the active token — not unit probes alone.

WIP

Design decisions

• KEY SCHEME RULED (gritzko 2026-08-14): scheme B — vim keys drive the cursor, asdw scrolls + back/fwd. j/k = active line down/up, h/l = active token prev/next; w/s = scroll up/down; a/d = back/forward through the view stack (back stays on -/BS too; FORWARD is NEW — a fwd stack popView currently drops, in scope). Displaced keys: help h → ?, wrap w → W.
• Wash values: reuse the LITE-010 wash-slot shape — one WASH_CUR_LINE + one WASH_CUR_TOK attr const in bro.js, OR'd in the row paint when the pager passes the cursor down; theme decides light/dark-safe shifts.

TODOs

• gritzko: pick scheme A / B / other, and whether fwd-stack is in scope
• cursor state + clamp (resize, refresh, popView keep it sane)
• wash rendering: line + token, composing over diff/why washes
• followable-token walk (next/prev U-backed tok, cross-row)
• Enter/click rewire to the active token, _followRow fallback
• scheme B keys: j/k/h/l cursor, w/s scroll, a/d back/fwd (+ fwd stack), h→?, w→W
• pty test: keys move the wash, Enter follows the active token
• gritzko: 8 checks in test/pager/pty.js contradict scheme B — amend or veto (list in Outcome)

Blockers and bummers

• aWash paints an UNWASHED cell grey 253 (near-white) — presumes a light default bg; a dark-terminal shift likely wants a theme probe (follow-up material, not amended here).

Outcome

Worker wt work/LITE-023, base ///lite/#2dc61618, 4 files staged (jab put), NO commit. Orchestrator re-ran the tests in the wt: green as reported.

• view/bro.js +21: WASH_CUR_LINE/WASH_CUR_TOK + aWash()/washBg() — a bg LIGHTNESS step, not aOr (OR over bg VALUES would garble the LITE-010 diff washes into one grey); in/rm stay apart under the cursor.
• view/pager.js +200/-23: per-view cur = {row, tok} (tok=-1 none), wash paint, followable walk (U follower or F tok), scheme-B keys, NEW fwd stack (a/d; a push clears it), help→?, wrap→W, cursor-drags-viewport clamp.
• test/cursor/{pty.js,run.sh} new: PASS cursor 43 checks under jab (real openpty; frame-byte asserts: j/k wash moves w/o scroll, h/l cross-row hops + stronger wash, Enter follows the active token incl. a hidden-U chooser row, a/d back/fwd, s/w scroll w/o cursor move, ? help from SHORTCUTS, W wrap, click sets cursor). RED verified against the landed pager first.
• Full lite suite: failure counts IDENTICAL to the pre-change baseline (blob 12/13, cat 16/35, click, first, index, list, logspine, merge, tree — all pre-existing env fails); PASS: chat commit commitnav cursor diff now.
• NOT amended, gritzko's call — test/pager/pty.js, 8 checks now red under scheme B: frame0-bar-top (h:→?:), key-j-scrolled frame1-shifted frame1-bar-pct (j = cursor now), key-w-wraps wrap-adds-rows wrap-shows-tail (w→W), enter-on-banner-notes (Enter fires on the active line, not the top row — the ticket's core change).

Suggested commit: LITE-023: active line + token in the lite pager, scheme B keys