lanework/Kanban/UI/Board/LaneView.swift

import AppKit
import SwiftUI

// MARK: - LaneView

/// One lane: a title bar and a vertically scrolling masonry of cards (03-board-ui.md § Lane).
///
/// ### The title bar (this milestone's subject)
///
/// Leading SF Symbol from `icon` — lenient, an unknown name renders the `square.stack` default
/// (`ItemSymbol`) — then the title or its quiet "Untitled" placeholder, a quiet secondary
/// card-count badge, and a trailing quiet new-card button. **The whole bar is the drag surface**:
/// a click selects the lane — toggling off on a repeat, exactly as empty space does
/// (04-interactions.md § Selection, settled) — and movement begins a **system drag session**
/// carrying the lane (`DragSession`, DRAG-REORDER.md). The click-versus-drag split is the system's
/// own now: `.onTapGesture` and `.onDrag` coexist, so a hesitant click can never start a drag and a
/// drag can never also select. The one thing carved out of the drag region is the new-card button,
/// which sits in an overlay outside it.
///
/// ### The lane's one context menu
///
/// "The lane has one context menu (settled), invoked on the header or on lane empty space alike"
/// (03-board-ui.md § Lane), so both surfaces attach the *same* `laneMenu`. It carries Rename, Style…,
/// the quick-style recents row, the Width stepper and Delete — 11-command-nexus.md ▸ Context menus'
/// Lane row, in its order, complete as of m5.
///
/// ### The card face
///
/// `CardFaceView`, complete as of m5: the search filter narrows what the masonry lays out and what
/// the badge counts, the deferred cut's dim rides the face (`cutTreatment`), and selection changes
/// only the face's styling — never its geometry. **A card has one presentation** (resettled
/// 2026-07-28, reversing the pathfinder's selection-keyed carousel): the masonry never reflows on
/// a click, and viewing an attachment's media is the card window's job, not the face's
/// (03-board-ui.md § Card face).
///
/// ### Equality gate
///
/// The lane is `Equatable` and instantiated through `.equatable()` (`BoardView.laneSlot`), because
/// the strip re-runs for reasons that have nothing to do with any one lane: `BoardView`'s body reads
/// the drag session, so **every drop-proposal change re-evaluates it**, and without a gate that
/// rebuilds every lane on the board — and, through them, every card face — on every cursor move
/// during a drag. See the `==` below for what the gate covers.
struct LaneView: View, Equatable {

    let store: BoardStore
    let lane: Lane

    /// The app-wide quick-style recents (03-board-ui.md § Styling ▸ Controls — "never board data"),
    /// read from the environment rather than threaded down the strip: the list belongs to the app,
    /// not to this board, and every context menu in the window wants it.
    @Environment(AppModel.self) private var appModel

    /// Interior masonry columns — the lane's width units, or the resize session's snapped count
    /// while this lane is being dragged. Passed in rather than read off `lane` so the live drag can
    /// override it (see `BoardView.laneSlot`).
    let columns: Int

    /// This lane's resting slot width — the replica's width, so the image under the cursor is the
    /// lane at its real on-screen size (03-board-ui.md § Motion: "a faithful, full-size replica").
    let slotWidth: CGFloat

    /// The board window's drop machinery: the app-wide session, the geometry registry this lane
    /// registers its card grid into, and the shared retarget every hover and every autoscroll step
    /// goes through (`BoardDropContext`).
    let drops: BoardDropContext

    /// The strip's rubber band: the lane's empty space is one of its three surfaces, and every card
    /// face registers its frame into the same registry (`MarqueeControl`).
    let marquee: MarqueeControl

    /// Opens a card's window — ⌘↩'s second half (04-interactions.md ▸ Grammar, "commits and opens
    /// the card window"). Supplied by the strip, which is supplied by the host: a lane has no
    /// business knowing about `WindowGroup` keys.
    let openCard: (ItemID) -> Void

    /// Reduce Motion, for the card transition below (10-accessibility.md). Read from the environment
    /// and handed to `Motion`, which owns what "reduced" means.
    @Environment(\.accessibilityReduceMotion) private var reduceMotion

    /// Increase Contrast, for the selection stroke and the plate's resting edge below
    /// (10-accessibility.md). Read from the environment and handed to `Accommodations`, which owns
    /// what "increased" does to a stroke.
    @Environment(\.colorSchemeContrast) private var contrast

    /// Reduce Transparency, for the lane plate's wash (`Accommodations.lanePlateWash` — translucent
    /// normally, the standard secondary background under the setting; the trash plate's rule).
    @Environment(\.accessibilityReduceTransparency) private var reduceTransparency

    /// The window's appearance — the input to 10-accessibility.md's runtime-contrast rule, and the
    /// answer for every board whose background is not a hand-written hex (`headerInk`).
    ///
    /// Read in `body` rather than resolved once, which is what makes "recomputed on appearance
    /// change" free: a light/dark flip re-evaluates this view, and the decision below is taken again
    /// against the colours of the appearance the window is now in.
    @Environment(\.colorScheme) private var colorScheme

    /// The board's ruler (03-board-ui.md ▸ Layout — zoom; `BoardZoom`), injected on the strip by
    /// `BoardWindowHost`. An `@Environment` read rather than a value passed down deliberately: this
    /// view is `.equatable()`, and environment values are the one input the gate cannot suppress.
    @Environment(\.boardZoom) private var zoom

    /// The live body metric — every figure this lane lays out on is a multiple of it
    /// (`BoardMetrics`, 10-accessibility.md's full-relative-scaling rule), at the board's zoom.
    private var pointSize: CGFloat { zoom.bodyPointSize }

    /// Spacing between cards, and between the interior columns.
    private var cardSpacing: CGFloat { BoardMetrics.cardSpacing(bodyPointSize: pointSize) }

    /// The lane plate's corner radius — shared by the selection treatment and the accent band, whose
    /// top corners round to exactly this so the band reads as the lane's own edge.
    private var cornerRadius: CGFloat { BoardMetrics.laneCornerRadius(bodyPointSize: pointSize) }

    /// C7 · full-column top edge (03-board-ui.md § Styling ▸ Capabilities).
    private var bandHeight: CGFloat { BoardMetrics.laneAccentBandHeight(bodyPointSize: pointSize) }

    /// The lane's drawn height, for the replica. Measured rather than derived, because a lane is as
    /// tall as the strip gives it.
    @State private var measuredHeight: CGFloat = 0

    /// The title bar's drawn height — **the replica's anchor** (`dragReplica`), measured for the same
    /// reason the lane's height is: the bar is as tall as the text in it, at whatever size the system
    /// is set to.
    @State private var measuredHeaderHeight: CGFloat = 0

    /// This lane's edge-autoscroll driver — one per lane, ticking only while a card session is in
    /// flight (`DragAutoScroller`, DRAG-REORDER.md § Edge autoscroll).
    @State private var autoScroller = DragAutoScroller()

    /// The whole of what this lane is a function of **as far as the strip is concerned**: the lane
    /// value (`Lane` is `Equatable`, its cards included, so an edit anywhere under this lane makes it
    /// a different value), the two layout figures the strip resolves rather than the lane
    /// (`columns` follows the live resize session, `slotWidth` the strip's standard width), and the
    /// window-lived collaborators — the store by identity, the band and the drop machinery by their
    /// own equivalence tests, which exist because `BoardView` rebuilds both structs, closures and
    /// all, on every body pass.
    ///
    /// **What the gate does not suppress is the point.** Everything this body reads through
    /// Observation — `drops.session`'s proposal, members and file target, `store.selection`,
    /// `store.searchFilter`, `store.transient`, `appModel.styleRecents` — invalidates this view
    /// directly, and `.equatable()` has no say in that. So the lane a drag is actually over still
    /// re-runs on every proposal change, and its `shadowRun` animation key still moves with it; what
    /// stops is the *other* lanes re-running because the strip did.
    ///
    /// Deliberately NOT compared: `@State` (per-identity, preserved across updates anyway),
    /// `@Environment` values (SwiftUI invalidates on those itself), and `openCard` — a closure the
    /// host rebuilds every pass, which is a pure hand-off to a `WindowGroup` key and identical in
    /// behaviour whatever closure object carries it.
    nonisolated static func == (lhs: LaneView, rhs: LaneView) -> Bool {
        lhs.lane == rhs.lane
            && lhs.columns == rhs.columns
            && lhs.slotWidth == rhs.slotWidth
            && lhs.store === rhs.store
            && lhs.marquee.isEquivalent(to: rhs.marquee)
            && lhs.drops.isEquivalent(to: rhs.drops)
    }

    var body: some View {
        // The strip's gate, observed (`BoardRenderMetrics`) — DEBUG only, and a `let _` because
        // `body` is a `@ViewBuilder` and a bare `Void` call is not a view.
        #if DEBUG
        let _ = BoardRenderMetrics.countLaneBody()
        #endif
        // `spacing: 0` and the padding moved inside: the accent band is **full-width** along the
        // lane's top edge, so it must sit outside the content inset rather than in it.
        VStack(alignment: .leading, spacing: 0) {
            accentBand
            VStack(alignment: .leading, spacing: BoardMetrics.laneStackSpacing(bodyPointSize: pointSize)) {
                header
                cardStack
            }
            .padding(BoardMetrics.lanePlatePadding(bodyPointSize: pointSize))
        }
        .background(selectionBackground)
        // Behind the selection wash, not composed into it: the plate is the lane's resting
        // surface, the wash above it is the selection's own layer, and stacked `.background`s
        // put the later one further back.
        .background(lanePlate)
        .overlay(selectionStroke)
        // The deferred cut's dim (04-interactions.md ▸ Clipboard) — on the whole lane, because a cut
        // lane is cut cards and all.
        .cutTreatment(of: lane.id, in: store)
        .onGeometryChange(for: CGFloat.self) { $0.size.height } action: { measuredHeight = $0 }
        // **This lane's drop target**, on the whole body. It accepts *every* session type and routes
        // internally — card sessions against this lane's masonry zones, lane sessions forwarded to
        // the strip's logic, external Finder file sessions against those same zones — because
        // single-target dispatch has no fall-through (DRAG-REORDER.md).
        .onDrop(of: boardDropTypes, delegate: LaneDropDelegate(context: drops, laneID: lane.id))
        // **The lane is an accessibility container** — "window → lanes (accessibility containers, in
        // lane `order`) → cards (leaf elements, in card `order`)" (10-accessibility.md ▸ The board
        // through VoiceOver). `.contain` rather than `.combine`: the header, the new-card button and
        // every card must stay individually reachable, which is the whole point of a container the
        // VoiceOver cursor enters (`TrashLaneView` states the same rule from the trash's side).
        .accessibilityElement(children: .contain)
        // "⟨title⟩, lane, N cards", where **N is the rendered count and therefore the filter's** —
        // the very collection the visible badge counts, so the spoken count and the drawn one are
        // one number ("the count reads the search filter like the visible badge"). A card the query
        // hid is never built, so it leaves the masonry and the accessibility tree in the same pass,
        // which is 10's "filtered-out cards leave layout and the accessibility tree together" holding
        // by construction rather than by a second rule.
        .accessibilityLabel(AccessibilityPhrases.laneLabel(title: lane.title.value, cards: renderedCards.count))
    }

    // MARK: - Header

    private var header: some View {
        headerContent
            // **The lane title is a heading** — "lane titles are headings, so the headings rotor
            // jumps lane-to-lane; on a one-dimensional board that *is* structural navigation"
            // (10-accessibility.md ▸ Rotor). One flattened element rather than icon + text + badge:
            // the glyph and the count are the container's information, already spoken by its label,
            // and three stops where the design asks for a heading would make the rotor useless.
            //
            // `.contain` while a rename is open, because the flattening would otherwise swallow the
            // text field the user is typing into — the one moment this subtree holds a control
            // rather than chrome.
            .accessibilityElement(children: isRenaming ? .contain : .ignore)
            .accessibilityLabel(AccessibilityPhrases.displayTitle(lane.title.value))
            .accessibilityAddTraits(headerTraits)
            // **VO-Space toggles the lane's selection** — the ⌘-click analogue 10-accessibility.md
            // gives a card, applied to the other selectable thing on the board, and routed through
            // the same `BoardStore.click` funnel the pointer uses so the homogeneity and
            // container rules are `SelectionGrammar`'s single answer rather than a second one.
            // Deliberately **not** the header's own plain-click semantics: "moving the VO cursor
            // never mutates selection … VO-Space on a card toggles its selection (the ⌘-click
            // analogue — a toggle, never plain click's replace)", and a VO-Space that replaced would
            // silently wipe a multi-lane selection the user had just built.
            .accessibilityAction { toggleLaneSelection() }
            // The bar is the drag surface, so it must be hit-testable across its whole width —
            // including the empty stretch between the badge and the button.
            .contentShape(Rectangle())
            // **The header toggles like empty space** (04-interactions.md § Selection, settled): "a
            // click on the already-selected lane's header unselects, one lane-click behavior
            // everywhere, so a full lane keeps a pointer path out of selection". Hence the same
            // `togglesOnRepeat` the empty space passes — the two surfaces differ only in where they
            // are. `.onTapGesture` beside `.onDrag` is the click-versus-drag split: the system holds
            // the drag off until the pointer actually moves, so a click is never a drag.
            .onTapGesture {
                store.click(
                    SelectionTarget(id: lane.id, kind: .lane, container: .board),
                    modifier: .current,
                    togglesOnRepeat: true
                )
            }
            .onDrag(startLaneDrag, preview: { dragReplica })
            // **The header is the file drop's topmost position** (04-interactions.md ▸ Drag and drop,
            // settled 2026-07-28: "a release on the lane header resolves to the topmost position —
            // forgiving beats a dead stripe"). Its frame is registered rather than derived from the
            // grid's because the grid is scroll-view *content*: scrolled down, its top edge climbs
            // past the header and would stop being a boundary at all. The bar itself never moves.
            .onGeometryChange(for: CGRect.self) { $0.frame(in: .global) } action: { frame in
                drops.registry.update(header: frame, for: lane.id)
                // The same measurement answers a second question, so it is read once: the bar is
                // what a lane drag is grabbed by, and its height is where the replica has to hang
                // from the pointer (`replicaHeaderCenterY`).
                measuredHeaderHeight = frame.height
            }
            .onDisappear { drops.registry.removeHeader(lane.id) }
            .overlay(alignment: .trailing) { newCardButton }
            // **10-accessibility.md's ≥ 4.5:1 rule, at the one place on the board where text sits on
            // a colour the user chose** (`BoardTextInk`) — palette name and hand-written hex alike,
            // since the ink they need is the same question and only the *verification* differs
            // (`ContrastMath`).
            //
            // A lane's plate is a *translucent wash* (`lanePlate`), so the title, the icon, the
            // count badge and the rename field still land, effectively, on the board's `background`
            // — the surface the design binds the threshold to ("text does sit on it",
            // `BoardView.boardBackground`); the wash shifts it too little to change the answer. The
            // card faces below are a different matter and deliberately untouched: they carry their
            // own opaque plate (`BoardSurface.cardPlate`), so their titles never see the board
            // colour.
            //
            // **Placed here, not at the end of the chain**, which is the modifier order doing real
            // work: everything above — including the new-card button in the overlay — is text on the
            // board background and takes the computed ink, while the context menu and the Style…
            // popover attached below stay in the window's own appearance, because a menu is system
            // chrome drawn on its own surface and not on this board's colour.
            .boardTextInk(headerInk)
            .contextMenu { laneMenu }
            // **The context menu's plain rows, additionally as custom actions** — "where SwiftUI
            // additionally surfaces menu items as custom accessibility actions, that's free
            // improvement, not a separate design surface" (10-accessibility.md). The menu itself
            // stays the inventory and is reachable the standard way (VO-⇧-M); this is the same four
            // commands one rotor turn closer. Style… is deliberately absent: it opens a popover —
            // its own accessible surface — and the quick-style swatch `Picker` beside it is not an
            // action at all.
            .accessibilityActions { laneActions }
            // The lane's half of the Style… popover. Anchored on the header because that is the
            // lane's own furniture — `styleEditorPresentation` decides whether this lane is the
            // session's presenting anchor at all.
            .popover(isPresented: styleEditorPresentation(store, anchor: lane.id), arrowEdge: .bottom) {
                StyleEditorPopover(store: store, recents: appModel.styleRecents)
            }
    }

    /// Which appearance's label vocabulary the header's text is drawn in — the window's own, unless
    /// the **board** paints a background whose composited surface fails ≥ 4.5:1 against it
    /// (10-accessibility.md ▸ Text scaling & visual accommodations; `BoardTextInk`).
    ///
    /// It is not only the exotic case: ten of the twelve palette wells are dark colours, and every
    /// one of them needs the dark appearance's label in a *light* window. A board styled entirely
    /// from the in-app grid reaches this line as often as a hand-written one does.
    ///
    /// It reads the *board*'s background and never this lane's, which is 03-board-ui.md ▸ Styling's
    /// C7 ruling showing up as arithmetic: a lane's colour is an edge band, not a fill, so it is
    /// never behind this text and carries no contrast obligation (`accentBand`).
    private var headerInk: ColorScheme {
        BoardTextInk.scheme(forBoardBackground: store.snapshot.background, appearance: colorScheme)
    }

    /// The lane's colour as C7 — "a lane's color paints a full-width band along its top edge; the
    /// surfaces themselves keep the standard chrome, so colored title text never sits on a colored
    /// fill" (03-board-ui.md § Styling ▸ Capabilities, settled in the pathfinder's treatment
    /// shootout).
    ///
    /// A value that resolves to nothing paints **no band**, and the bytes stay on disk exactly as
    /// written — the card stripe's rule, for its reason: there is no sensible default colour for
    /// "the author meant something we can't read", and a wrong colour is worse than none.
    @ViewBuilder
    private var accentBand: some View {
        if let color = Palette.color(for: lane.background) {
            UnevenRoundedRectangle(topLeadingRadius: cornerRadius, topTrailingRadius: cornerRadius)
                .fill(color)
                .frame(height: bandHeight)
                .frame(maxWidth: .infinity)
                // Decoration only: the header below it owns the lane's click and drag.
                .allowsHitTesting(false)
        }
    }

    // MARK: - The lane's one context menu

    /// Rename, Style…, the quick-style recents row, the Width control, Delete — 11-command-nexus.md ▸
    /// Context menus' Lane row, in its order, complete as of m5.
    @ViewBuilder
    private var laneMenu: some View {
        // Rename: Board ▸ Rename's exact store path (`BoardRenameCommand`) — `beginRename(of:
        // currentTitle:)`, seeded with the lane's live title. The menu-bar item additionally requires
        // this lane to be the *sole* selection; a context menu already names its target by where it
        // was invoked, so — standard macOS practice — it acts on the clicked lane outright.
        Button("Rename") { beginRename() }
            .disabled(!store.acceptsBoardMutations)

        Divider()

        StyleMenuItems(store: store, recents: appModel.styleRecents, target: styleTarget)

        Divider()

        widthControl

        Divider()

        // Delete: File ▸ Delete's exact store path (`store.delete`), on the same widened target set
        // Style… above reads (`targetIDs`, `styleTarget`'s `Set<ItemID>` sibling below) — the
        // successor-selection rule is `delete(_:)`'s own, so this row gets it for free.
        Button("Delete") { deleteTargets() }
            .disabled(!store.acceptsBoardMutations)
    }

    /// The menu's plain rows again, as VoiceOver custom actions (see the `.accessibilityActions`
    /// call site). Every one of them calls the *same* private method its menu row does, so the two
    /// surfaces cannot drift into meaning different things — which is the only way "not a separate
    /// design surface" is checkable rather than merely intended.
    @ViewBuilder
    private var laneActions: some View {
        let units = LaneLayoutMath.displayUnits(of: lane)
        Button("Rename") { beginRename() }
            .disabled(!store.acceptsBoardMutations)
        Button("Increase Width") { store.setLaneWidth(lane.id, units: units + 1) }
            .disabled(!store.acceptsBoardMutations)
        Button("Decrease Width") { store.setLaneWidth(lane.id, units: units - 1) }
            .disabled(!store.acceptsBoardMutations || units <= 1)
        Button("Delete") { deleteTargets() }
            .disabled(!store.acceptsBoardMutations)
    }

    /// Board ▸ Rename's store path, seeded with the lane's live title — one method, two callers
    /// (the context menu row and its accessibility twin).
    private func beginRename() {
        store.transient.beginRename(of: lane.id, currentTitle: lane.title.value)
    }

    /// File ▸ Delete's store path over the context-menu target set — the menu row's body and its
    /// accessibility twin's alike.
    private func deleteTargets() {
        store.delete(targetIDs)
    }

    /// VO-Space's landing: the ⌘-click funnel, on this lane. `togglesOnRepeat` stays false because
    /// only the *plain* branch reads it — the ⌘ branch is already a toggle, which is the point.
    private func toggleLaneSelection() {
        store.click(
            SelectionTarget(id: lane.id, kind: .lane, container: .board),
            modifier: .command
        )
    }

    /// The header element's traits: a heading always, and **selected when the lane is** — "state is
    /// never colour-alone: selection is a ring plus trait" (10-accessibility.md).
    private var headerTraits: AccessibilityTraits {
        isSelected ? [.isHeader, .isSelected] : [.isHeader]
    }

    /// The width stepper — "the header context menu's Width control (stepper, uncapped) is the
    /// precise control … it never touches the window, it **re-divides** the existing width across the
    /// new unit total" (03-board-ui.md § Lane). A +/− pair rather than a slider or a fixed 1×/2×/3×
    /// list, because the control is uncapped in one direction and floored at one unit in the other.
    ///
    /// **Single-lane by nature**, unlike the style entries above it: the design gives the batch to
    /// the ⌥⌘→/⌥⌘← menu items and keeps the stepper on the lane whose menu is open.
    private var widthControl: some View {
        let units = LaneLayoutMath.displayUnits(of: lane)
        return Section("Width — \(units)×") {
            Button("Increase Width") {
                store.setLaneWidth(lane.id, units: units + 1)
            }
            Button("Decrease Width") {
                store.setLaneWidth(lane.id, units: units - 1)
            }
            // A one-unit lane cannot shrink (`width` is ≥ 1), and an item whose only outcome is a
            // no-op reads better disabled than dead — `LaneWidthCommands`' rule, same floor.
            .disabled(units <= 1)
        }
        .disabled(!store.acceptsBoardMutations)
    }

    /// What this lane's menu styles: the whole selection when this lane is part of it, else this lane
    /// alone — standard macOS context-menu targeting (the pathfinder's `styleSelection`, kept).
    /// Right-clicking something outside the selection acts on what was clicked.
    private var styleTarget: StyleTarget {
        guard store.selection.container == .board, store.selection.ids.contains(lane.id) else {
            return .items([lane.id])
        }
        return .items(store.selection.ids)
    }

    /// Delete's target set — the same widening `styleTarget` does, spelled as a plain `Set<ItemID>`
    /// because `store.delete(_:)` takes one directly (the context-menu targeting rule, reused here
    /// on the live side).
    private var targetIDs: Set<ItemID> {
        guard store.selection.container == .board, store.selection.ids.contains(lane.id) else {
            return [lane.id]
        }
        return store.selection.ids
    }

    private var headerContent: some View {
        HStack(alignment: .firstTextBaseline, spacing: BoardMetrics.laneHeaderSpacing(bodyPointSize: pointSize)) {
            Image(systemName: ItemSymbol.name(lane.icon, fallback: ItemSymbol.lane))
                .foregroundStyle(.secondary)
                .imageScale(.medium)
            headerTitle
            countBadge
            Spacer(minLength: 0)
        }
        // Reserves the button's width so a long title truncates before it collides, and keeps the
        // button out of the gestured region. **Font-derived** rather than a fixed 22pt: the button
        // is an `Image` at a relative image scale, so a fixed reserve would be overrun by the glyph
        // itself at a large system text size and 03-board-ui.md's graceful-truncation rule would
        // quietly stop holding (`BoardMetrics.newCardButtonReserve`).
        .padding(.trailing, BoardMetrics.newCardButtonReserve(bodyPointSize: pointSize))
        .padding(.horizontal, BoardMetrics.laneHeaderInset(bodyPointSize: pointSize))
    }

    /// The title, or the rename editor when this lane is the one being renamed.
    ///
    /// A lane's **only** rename path is Board ▸ Rename (04-interactions.md ▸ Selection: "the menu
    /// item is a lane's only rename path, since Return on a lane creates a card"), so nothing in
    /// this view opens the editor — it only renders one that is already open.
    @ViewBuilder
    private var headerTitle: some View {
        if isRenaming {
            InlineTitleField(
                text: renameDraft,
                prompt: "Lane name",
                onCommit: { store.commitRename() },
                onAbandon: { store.transient.discardRename() },
                onFocusLoss: { store.commitRename() },
                // A lane has no card window; ⌘↩ still commits, which is the half of the rule that
                // applies (04 ▸ Grammar's carve-out is "commits the edit — placeholder or rename —
                // and open[s] the card window", and only a card has one to open).
                onCommitAndOpen: { store.commitRename() }
            )
            .boardFont(.headline)
        } else {
            Text(lane.title.value ?? "Untitled")
                .boardFont(.headline)
                .foregroundStyle(lane.title.value == nil ? .secondary : .primary)
                .lineLimit(1)
                .truncationMode(.tail)
        }
    }

    /// The card-count badge — quiet, secondary (03-board-ui.md § Lane).
    ///
    /// **It counts exactly what the body renders**, because it reads the same `renderedCards` the
    /// masonry iterates. That is deliberate rather than incidental: "The count reads the search
    /// filter like every other surface — during a search it shows the visible count, not the
    /// total", so when m5's search card narrows `renderedCards` to the filter's survivors the badge
    /// follows by construction, with no second rule to keep in step. **A lane the query empties
    /// shows `0` and keeps its slot** (04-interactions.md § Search, settled: "lanes are never
    /// filtered out … the search filters cards, and the board's structure is not a search result") —
    /// which is `BoardView.liveLanes` never consulting the filter at all, made visible here.
    ///
    /// The rename exemption rides along for the same reason every other rule does: the badge counts
    /// what the body renders, and while an inline rename is open its card is one of the things the
    /// body renders (see `renderedCards`).
    private var countBadge: some View {
        Text("\(renderedCards.count)")
            .boardFont(.caption)
            .monospacedDigit()
            .foregroundStyle(.secondary)
            .padding(.horizontal, BoardMetrics.badgeHorizontalPadding(bodyPointSize: pointSize))
            .padding(.vertical, BoardMetrics.badgeVerticalPadding(bodyPointSize: pointSize))
            .background(Capsule().fill(.quaternary))
    }

    /// The new-card button — a **pointer twin** of File ▸ New Card whose click *names its target*:
    /// "the lane header's new-card button overrides [the ⌘N target] rule — the click names its
    /// target lane, selection notwithstanding" (11-command-nexus.md ▸ Pointer grammar, settled), so
    /// it passes this lane and no anchor rather than consulting `NewCardTarget`.
    private var newCardButton: some View {
        Button {
            store.transient.beginPlaceholder(inLane: lane.id)
        } label: {
            Image(systemName: "plus")
                .imageScale(.small)
                .foregroundStyle(.secondary)
                .contentShape(Rectangle())
        }
        .buttonStyle(.plain)
        // "The lane header's new-card button is a labeled child ('New card in ⟨lane⟩')"
        // (10-accessibility.md ▸ The board through VoiceOver) — the header's one child element, which
        // is why it lives in an overlay outside the flattened bar rather than inside it.
        .accessibilityLabel(AccessibilityPhrases.newCardLabel(lane: lane.title.value))
        // Mutating, so the read-only lock disables it like every other write path
        // (02-architecture.md § The lock's scope), and the focused-editor rule closes it while an
        // inline editor is open (04 ▸ Grammar) — the pointer twin of a disabled menu item.
        .disabled(store.isReadOnly || store.isEditingInline)
    }

    // MARK: - The lane drag

    /// Begins the lane's system drag session (DRAG-REORDER.md; 04-interactions.md ▸ Drag and drop).
    ///
    /// **Dragging any member of a multi-selection drags the whole selection**, in board order —
    /// which is the lane level's flatten order. A lane outside the selection drags alone, standard
    /// macOS targeting.
    ///
    /// Refused under the read-only lock and while an inline editor is focused, like every other
    /// mutating gesture (02-architecture.md § The lock's scope; 04 ▸ Grammar's focused-editor rule).
    /// A refusal is an item provider carrying nothing: no session begins, every drop target declines,
    /// and the image snaps back.
    private func startLaneDrag() -> NSItemProvider {
        guard !store.isReadOnly, !store.isEditingInline else { return NSItemProvider() }
        let selection = store.selection
        let ids: Set<ItemID> = selection.container == .board
            && selection.ids.contains(lane.id)
            && selection.ids.count > 1
            ? selection.ids
            : [lane.id]

        let members = store.snapshot.lanes.filter { ids.contains($0.id) }
        guard !members.isEmpty else { return NSItemProvider() }

        let root = store.rootURL
        let payload = DragPayload(
            boardRoot: root,
            kind: .lanes,
            container: .board,
            items: members.map {
                DragPayload.Item(
                    id: $0.id.rawValue,
                    folder: root.appendingPathComponent($0.id.rawValue, isDirectory: true).path,
                    title: $0.title.value
                )
            }
        )
        drops.session.beginLanes(
            members.map(\.id),
            folders: payload.folders,
            // The dragged items' own sizes, frozen at drag start — the one thing that is
            // (03-board-ui.md § Motion).
            units: members.map { LaneLayoutMath.displayUnits(of: $0) },
            source: store
        )
        return payload.itemProvider()
    }

    /// The image travelling under the cursor: **a faithful, full-size replica of the whole lane**,
    /// not the strip of title bar that was grabbed (03-board-ui.md § Motion), fanned with ghosts and
    /// a count badge for a multi-drag.
    ///
    /// A static rendition rather than a live `LaneView`: a drag image is a snapshot, so it carries no
    /// scrolling, no gestures and no geometry observers, and the card list is capped because anything
    /// past the lane's height is clipped anyway.
    ///
    /// **The pointer keeps the point it grabbed** — the Finder-icon promise, and the whole of what
    /// the anchoring padding below buys. SwiftUI centres a preview on the view the drag started
    /// from, which here is the *title bar*: uncompensated, a full-height replica centred on a bar a
    /// line and a half tall hangs half a lane above the cursor, and the cursor lands in the middle of
    /// the image rather than on the bar it grabbed. Padding the replica so its own title bar is the
    /// image's centre undoes exactly that, and lands every other pixel of the replica over the lane
    /// it was lifted from (`DragPreviewAnchor`, which is where the arithmetic and its reasoning live).
    private var dragReplica: some View {
        let count = max(1, draggedLaneCount)
        let anchor = DragPreviewAnchor.padding(length: replicaHeight, anchor: replicaHeaderCenterY)
        return ZStack {
            if count > 2 { replicaFace.offset(x: 12, y: 12).opacity(0.45) }
            if count > 1 { replicaFace.offset(x: 6, y: 6).opacity(0.7) }
            replicaFace
        }
        .overlay(alignment: .topTrailing) { DragCountBadge(count: count) }
        .padding(BoardMetrics.replicaPadding(bodyPointSize: pointSize))
        // Transparent, and only ever on one side — see `DragPreviewAnchor`, whose figure is
        // deliberately invariant under the symmetric margin above, so the two paddings compose in
        // any order.
        .padding(.top, anchor.before)
        .padding(.bottom, anchor.after)
        // **Back to the window's own appearance**, undoing `header`'s runtime-contrast override for
        // this one subtree (`boardTextInk`). The preview is attached inside that modifier and would
        // otherwise inherit it — but the replica is not text on the board background: it draws its
        // own opaque plate (`replicaFace`) and floats over whatever the cursor is above, which during
        // a cross-window drag is another board entirely. The rule's premise is a user-chosen surface
        // behind the glyphs, and here there is none.
        .boardTextInk(colorScheme)
    }

    private var draggedLaneCount: Int {
        let selection = store.selection
        guard selection.container == .board, selection.ids.contains(lane.id) else { return 1 }
        return selection.ids.count
    }

    private var replicaFace: some View {
        VStack(alignment: .leading, spacing: 0) {
            accentBand
            VStack(alignment: .leading, spacing: BoardMetrics.laneStackSpacing(bodyPointSize: pointSize)) {
                headerContent
                VStack(alignment: .leading, spacing: cardSpacing) {
                    ForEach(renderedCards.prefix(12)) { card in
                        HStack(alignment: .firstTextBaseline,
                               spacing: BoardMetrics.cardRowSpacing(bodyPointSize: pointSize)) {
                            Image(systemName: ItemSymbol.name(card.icon, fallback: ItemSymbol.card))
                                .foregroundStyle(.secondary)
                                .imageScale(.medium)
                            Text(card.title.value ?? "Untitled")
                                .boardFont(.body)
                                .lineLimit(2)
                            Spacer(minLength: 0)
                        }
                        .padding(BoardMetrics.cardContentPadding(bodyPointSize: pointSize))
                        .frame(maxWidth: .infinity, alignment: .leading)
                        .background(
                            RoundedRectangle(cornerRadius: BoardMetrics.cardCornerRadius(bodyPointSize: pointSize))
                                .fill(BoardSurface.cardPlate)
                        )
                    }
                    Spacer(minLength: 0)
                }
            }
            .padding(BoardMetrics.lanePlatePadding(bodyPointSize: pointSize))
        }
        .frame(width: replicaWidth, height: replicaHeight, alignment: .topLeading)
        // The lane's own wash over an opaque base — the replica floats over whatever the cursor is
        // above, so unlike the resting lane it cannot let the wash composite against the board; the
        // opaque `.background` behind it stands in for the window's neutral surface.
        .background(lanePlate)
        .background(RoundedRectangle(cornerRadius: cornerRadius).fill(.background))
        .clipShape(RoundedRectangle(cornerRadius: cornerRadius))
        .dragReplicaShadow(zoom: zoom)
    }

    /// The replica's size — **the lane's own**, floored for a lane that has not measured itself yet
    /// (`BoardMetrics`). Named rather than inlined in the frame because the anchoring below has to
    /// ask the same question the drawing does, and two derivations of one figure would be two
    /// answers.
    private var replicaWidth: CGFloat {
        max(slotWidth, BoardMetrics.laneReplicaMinimumWidth(bodyPointSize: pointSize))
    }

    private var replicaHeight: CGFloat {
        max(measuredHeight, BoardMetrics.laneReplicaMinimumHeight(bodyPointSize: pointSize))
    }

    /// Where the title bar sits inside the replica — its **centre**, measured down from the replica's
    /// top edge, which is the point the pointer must hold (`dragReplica`).
    ///
    /// The replica stacks exactly what the lane stacks, in the same order and off the same figures,
    /// so this is the lane's own layout arithmetic rather than a second description of it: the accent
    /// band when the lane's colour resolves to one (`accentBand` draws nothing when it does not, and
    /// contributes no height either), the plate's inset, and half the measured bar.
    private var replicaHeaderCenterY: CGFloat {
        let band = Palette.color(for: lane.background) == nil ? 0 : bandHeight
        return band + BoardMetrics.lanePlatePadding(bodyPointSize: pointSize) + measuredHeaderHeight / 2
    }

    // MARK: - Body

    /// The card stack. Its empty space is a click target in its own right (04 ▸ Selection): one
    /// click selects the lane or, when it is already the selection, clears it; a double click
    /// creates a card at the bottom with its title editor focused.
    ///
    /// **"Selection scrolls into view"** (04-interactions.md ▸ Grammar): the reader watches the
    /// navigation head — the cursor the arrows move, not the whole selection — and scrolls only when
    /// the head names a card *this* lane renders, so exactly one lane responds to any one press.
    /// Deliberately unwrapped by `withAnimation`: 03-board-ui.md § Motion has selection follow
    /// "whatever transaction is active rather than easing on its own".
    private var cardStack: some View {
        ScrollViewReader { proxy in
            scrollableCards
                .onChange(of: store.transient.selectionHead) { _, head in
                    guard let head, let card = renderedCards.first(where: { $0.id == head }) else { return }
                    proxy.scrollTo(LaneSlot.identity(of: card.id))
                }
        }
    }

    private var scrollableCards: some View {
        // Evaluated ONCE per body, deliberately: the per-element `.accessibilitySortPriority`
        // below reads `slots.count`, and reading the computed property from inside the `ForEach`
        // closure re-runs the whole chain (`slots` → `renderedCards` → the O(n) card filter and
        // the session's hidden-member resolution) once per element — O(n²) per lane body, which
        // a drag pickup's synchronous whole-board layout multiplied into a visible stall.
        let slots = self.slots
        return ScrollView(.vertical) {
            // Cards stay standard width whatever the lane spans: at a slot width of
            // `units × standard + (units - 1) × gap`, `MasonryLayout` divides back into exactly
            // `units` columns of `standard` (03-board-ui.md § Layout — full visibility).
            MasonryLayout(columns: columns, spacing: cardSpacing) {
                ForEach(Array(slots.enumerated()), id: \.element.id) { index, slot in
                    Group {
                        switch slot {
                        case let .card(card):
                            CardFaceView(
                                store: store,
                                card: card,
                                role: .board(openCard: openCard),
                                marquee: marquee,
                                drops: drops
                            )
                            // **The value gate** (`CardFaceView.==`) — the lane's own, one level
                            // down: this body re-runs on every proposal change while a drag is over
                            // *this* lane, and the faces it draws are almost never what changed.
                            .equatable()
                        case let .placeholder(phase):
                            NewCardStubView(store: store, phase: phase, openCard: openCard)
                        case let .shadow(_, height):
                            // One of the drag's N contiguous shadows, at the dragged card's frozen
                            // height — the run's real footprint, so the drop lands exactly here.
                            DragShadow(cornerRadius: BoardMetrics.cardCornerRadius(bodyPointSize: pointSize))
                                .frame(height: height)
                        }
                    }
                    // "Appear/disappear is scale + fade (cards scale from ~0.8 …)"
                    // (03-board-ui.md § Motion), which is how a create, a delete, a restore and
                    // (m5) a search filter's leavers all reach the masonry. The placeholder wears it
                    // too: it is the card, one round trip early. Whether any of it *performs* is
                    // decided upstream — at the reload for the real cards (`Motion.reloadAnimates`),
                    // at the gesture for the placeholder, which touches no disk.
                    //
                    // **The create handoff deliberately never reaches it.** A committed placeholder
                    // is already keyed by the arriving card's identity (`LaneSlot`), so when the echo
                    // reload swaps the pseudo-card for the real one the `ForEach` element is neither
                    // inserted nor removed — only its content changes — and an appear/disappear
                    // transition has nothing to run on. That is the whole of "the handoff must read
                    // as one arrival" (02-architecture.md ▸ TransientBoardState ▸ overlays), and it
                    // holds identically under Reduce Motion: a transition that does not fire has no
                    // variant to choose between.
                    .transition(Motion.cardTransition(reduced: reduceMotion))
                    // **VoiceOver reads the masonry by `order`, not by drawn position** —
                    // 10-accessibility.md ▸ Logical order, not masonry position (decided).
                    //
                    // The divergence narrowed when the masonry went column-major — walking down
                    // one column now *is* consecutive `order` — but it is still real, and it is
                    // why an explicit priority is needed at all: a geometry-sorted accessibility
                    // tree (which is what a container does without this) sweeps in reading order,
                    // left-to-right then down, which over a column-major grid interleaves the
                    // columns: 1, 4, 7, 2, 5, 8 …, an order that exists nowhere in the model, on
                    // disk, or in the keyboard grammar. Priority descends with the slot index, so
                    // the highest reads first and the list is exactly `slots` — the same sequence
                    // the masonry is handed and the same one `SelectionGrammar` flattens.
                    //
                    // The drag shadows are inert here: `DragShadow` hides itself from the tree, and
                    // a slot that is not an element consumes no priority.
                    .accessibilitySortPriority(Double(slots.count - index))
                    // The scroll target. `ForEach` already carries this identity, but `scrollTo`
                    // resolves against an explicit `.id`, and it goes outermost so the transition
                    // above stays inside the identified view rather than around it.
                    .id(slot.id)
                }
            }
            .frame(maxWidth: .infinity, maxHeight: .infinity, alignment: .topLeading)
            // The drag's reflow-to-make-room inside the lane, keyed on **this lane's shadow run**
            // and nothing broader (03-board-ui.md § Motion). `MasonryLayout` is a `Layout` over one
            // `ForEach` precisely so the round-robin reshuffle animates as positional slides rather
            // than as remove/insert blinks (DRAG-REORDER.md § The card masonry).
            .animation(Motion.dragReflow(reduced: reduceMotion), value: shadowRun)
            // Where this lane's card grid is drawn, in the window's global space — the analytic
            // resting grid the drop model replays `MasonryPlacement` over. Registered rather than
            // re-derived, so the zones and the drawn grid cannot disagree.
            .onGeometryChange(for: CGRect.self) { $0.frame(in: .global) } action: { frame in
                drops.registry.update(
                    LaneDropRegistry.Grid(frame: frame, columns: max(1, columns), spacing: cardSpacing),
                    for: lane.id
                )
            }
            .onDisappear { drops.registry.removeGrid(lane.id) }
            // The edge-autoscroll anchor, **inside** the scroll view's content so
            // `enclosingScrollView` resolves (`DragAutoScrollAnchor`). Every scroll step re-resolves
            // the proposal through the same shared retarget the drop delegate uses, because the
            // cursor is stationary while the content moves under it.
            .background {
                DragAutoScrollAnchor(scroller: autoScroller) {
                    drops.retargetCards(inLane: lane.id)
                }
            }
            .contentShape(Rectangle())
            // Order matters: the two-tap recogniser must be attached first so a double click is not
            // consumed as two singles.
            .onTapGesture(count: 2) {
                guard !store.isReadOnly, !store.isEditingInline else { return }
                store.transient.beginPlaceholder(inLane: lane.id)
            }
            // "Single click selects the lane (click again to unselect)" — the toggle the header
            // shares (04-interactions.md § Selection), and the modifier grammar on top of it.
            .onTapGesture {
                store.click(
                    SelectionTarget(id: lane.id, kind: .lane, container: .board),
                    modifier: .current,
                    togglesOnRepeat: true
                )
            }
            // The rubber band's first surface — "click-drag rubber-bands across lanes". Simultaneous
            // so the taps above stay instant; the band's own begin guard is what keeps a drag that
            // started on a card face out of it (`MarqueeControl`).
            .simultaneousGesture(marquee.gesture(in: .board))
            // The same menu the header carries — "one menu, invoked on the header or lane empty
            // space alike" (03-board-ui.md § Lane, settled).
            .contextMenu { laneMenu }
        }
        // The autoscroll driver, **structurally terminated**: a `.task(id:)` keyed on whether a card
        // session is in flight at all, so it is cancelled the moment the session ends — and
        // `DragSession`'s watchdog guarantees that flag clears however the drag finished
        // (DRAG-REORDER.md § Edge autoscroll). Within a session, a pointer outside this lane's
        // engagement rect simply scrolls nothing.
        .task(id: drops.session.isDraggingCards) {
            guard drops.session.isDraggingCards else { return }
            // The engagement rect's reaches are distances to this lane's own furniture, so they are
            // measured on the board's ruler (03-board-ui.md ▸ Layout — zoom). Read once at the head
            // of the drag, which is once per level: zoom is inert while a session is in flight.
            autoScroller.bodyPointSize = pointSize
            await autoScroller.run()
        }
    }

    /// Where a card drag would land **in this lane**, or `nil` when the proposal is elsewhere.
    private var cardProposal: Int? {
        drops.session.laneProposal(onBoardRooted: store.rootKey, laneID: lane.id)
    }

    /// The shadow run this lane opens, or `nil` when no proposal names it — the masonry's one
    /// make-room mechanism, and the reflow's narrow animation key.
    ///
    /// Two sessions feed it and they are mutually exclusive by construction (a file session never
    /// arms `DragSession`, so `isActive` is false for exactly as long as one is in flight):
    ///
    /// - **a card drag**, at the dragged cards' frozen heights — the run's real footprint, so the
    ///   drop lands exactly where the shadows are;
    /// - **a Finder file drag**, at the nominal height, one shadow per file — the cards being
    ///   proposed do not exist yet, so there is no measured height to be faithful to.
    private var shadowRun: ShadowRun? {
        if let position = cardProposal {
            return ShadowRun(position: position, heights: drops.session.cardHeights)
        }
        if let proposal = drops.session.fileLaneProposal(onBoardRooted: store.rootKey, laneID: lane.id) {
            return ShadowRun(
                position: proposal.index,
                heights: Array(repeating: drops.registry.nominalCardHeight, count: proposal.count)
            )
        }
        return nil
    }

    /// What the masonry lays out: the rendered cards, the drag's N contiguous shadows at the
    /// proposal, and the new-card placeholder when this lane is the one being created into.
    ///
    /// The placeholder is inserted **at the position the card will actually take** — after its anchor
    /// for ⌘N's "immediately after it", at the bottom otherwise — by asking the very function the
    /// commit uses to compute the rank (`BoardStore.insertionIndex`). One answer, so the pseudo-card
    /// cannot appear anywhere but where the real card lands. Both insertions are computed against
    /// `renderedCards`, and the placeholder's is shifted past a shadow run that opened in front of
    /// it, so neither displaces the other.
    ///
    /// **The phase rides along**, because it is what keys the slot: a committed placeholder wears the
    /// arriving card's identity so the handoff is one arrival rather than two (`LaneSlot`). The
    /// position math above is untouched by that — the key changes, the index does not — so the
    /// masonry cannot flinch at the moment of commit.
    private var slots: [LaneSlot] {
        var result = renderedCards.map(LaneSlot.card)

        let run = shadowRun
        let shadowPosition = run.map { min(max(0, $0.position), result.count) }
        var placeholder: (position: Int, phase: NewCardPlaceholder.Phase)?
        if let pending = store.transient.newCardPlaceholder, pending.laneID == lane.id {
            let position = BoardStore.insertionIndex(after: pending.anchorCardID, among: renderedCards)
                ?? result.count
            placeholder = (position, pending.phase)
        }

        let heights = run?.heights ?? []
        if let shadowPosition {
            let shadows = heights.enumerated().map { LaneSlot.shadow(index: $0.offset, height: $0.element) }
            result.insert(contentsOf: shadows, at: shadowPosition)
        }
        if var placeholder {
            if let shadowPosition, placeholder.position >= shadowPosition {
                placeholder.position += heights.count
            }
            result.insert(.placeholder(placeholder.phase), at: min(placeholder.position, result.count))
        }
        return result
    }

    /// **A deleted card renders nowhere, and needs no rule to**: deletion is a *move* into `.trash/`
    /// (03-board-ui.md § Trash, resettled 2026-07-28), so a deleted card has physically left
    /// `lane.cards` and the trash column renders it instead. The tombstone era's ancestor walk and
    /// effective-liveness predicate are retired with the flag they read.
    ///
    /// **A card being dragged *as a move* renders nowhere, for as long as the session lasts.** It is
    /// lifted out of the resting layout at pickup and stays out until release. A card being dragged
    /// as a **copy** stays exactly where it is and dims (`CardFaceView`), because that is what the
    /// copy will leave behind — so flipping ⌥ or ⌘ mid-drag reflows this lane once, which is the
    /// feedback the modifier is asking for (DRAG-REORDER.md § Resting-layout zones, ruled
    /// 2026-08-01). Either way the drop's index is counted in the layout on screen, which is why
    /// `copyCards` and `moveCards` take their index in different spaces.
    ///
    /// **A card the live search filter hides renders nowhere either** (04-interactions.md § Search):
    /// "cards whose title *and* body both miss the query animate out". This is the one collection
    /// that narrowing, which is what makes the filter "the single source of truth for what's on the
    /// board" true of this lane's every surface at once — the masonry, the count badge (see
    /// `countBadge`), the drop zones' resting layout, the marquee registration and the Finder
    /// file-drop targets all read this list or the registry it populates, so none of them needs a
    /// rule of its own.
    ///
    /// **With one exception, and it is the open inline rename** (04-interactions.md § Search,
    /// settled): "an open inline rename survives the filter hiding its card — the editor is a surface
    /// the filter doesn't reach; it stays open and focused, commits by UUID wherever the card lives,
    /// Escape abandons". The editor is drawn *inside* its card's slot (`CardFaceView.isRenaming`), so
    /// on this board a surface the filter doesn't reach means precisely a slot the filter doesn't
    /// take away: filtering the card out would unmount the field mid-keystroke and silently drop what
    /// the user had typed, which is the dirty-buffer courtesy read backwards. The exemption lasts
    /// exactly as long as the editor — commit or Escape retires it, the predicate applies again in
    /// the same pass, and a card that no longer matches animates out then (which is also the whole of
    /// "rename deliberately gets no carve-out": the *query* still stands throughout).
    ///
    /// It cannot arrive by typing, because focusing the search field is focus loss and commits the
    /// rename first; the case it serves is a foreign edit that stops the card matching while the user
    /// is renaming it.
    private var renderedCards: [Card] {
        Self.rendered(
            lane.cards,
            hiddenByDrag: drops.session.hiddenMembers(onBoardRooted: store.rootKey),
            filter: store.searchFilter,
            renaming: store.transient.renameEditor?.targetID
        )
    }

    /// `renderedCards` as a pure function of its four inputs — see there for every rule it applies.
    /// Split out only so the rules can be pinned without a view (`SearchFilterTests`); the lane's
    /// masonry, its count badge and its drop zones all read the property, which reads this.
    nonisolated static func rendered(
        _ cards: [Card],
        hiddenByDrag hidden: Set<ItemID>,
        filter: SearchFilter,
        renaming: ItemID?
    ) -> [Card] {
        cards.filter { card in
            guard !hidden.contains(card.id) else { return false }
            return filter.matches(card) || card.id == renaming
        }
    }

    // MARK: - Selection

    private var isSelected: Bool {
        store.selection.container == .board && store.selection.ids.contains(lane.id)
    }

    /// The lane's resting plate — the standard chrome's middle step (03-board-ui.md § Styling ▸
    /// Capabilities: window, washed lane, opaque card), the same quaternary wash the trash column
    /// wears (`Accommodations.lanePlateWash`). Translucent deliberately: a board-chosen colour
    /// shows through, so the header ink computed against the board's background (`headerInk`)
    /// keeps its premise on a painted board.
    private var lanePlate: some View {
        RoundedRectangle(cornerRadius: cornerRadius)
            .fill(Accommodations.lanePlateWash(reduceTransparency: reduceTransparency).style)
    }

    /// The selection treatment: a subtle whole-lane accent wash and stroke. Deliberately quiet —
    /// 03-board-ui.md gives lane *colour* to the top-edge accent band, so selection must not read as
    /// a fill that would compete with it once that lands.
    private var selectionBackground: some View {
        RoundedRectangle(cornerRadius: cornerRadius)
            .fill(isSelected ? AnyShapeStyle(Color.accentColor.opacity(0.08)) : AnyShapeStyle(.clear))
    }

    /// The selection ring — and, under Increase Contrast, the plate's resting edge as well
    /// (10-accessibility.md: "Increase Contrast strengthens borders and the selection indicator";
    /// `Accommodations`, and `CardFaceView.plateStroke` for the same three-way branch on a card).
    ///
    /// A lane is otherwise bounded by its quiet wash and the gap between it and its neighbour,
    /// which is the distinction the setting most needs to restore here.
    private var selectionStroke: some View {
        RoundedRectangle(cornerRadius: cornerRadius)
            .strokeBorder(plateStroke, lineWidth: plateStrokeWidth)
    }

    private var plateStroke: AnyShapeStyle {
        if isSelected {
            AnyShapeStyle(Color.accentColor)
        } else if Accommodations.drawsRestingBorder(contrast: contrast) {
            AnyShapeStyle(.separator)
        } else {
            AnyShapeStyle(.clear)
        }
    }

    private var plateStrokeWidth: CGFloat {
        Accommodations.borderWidth(isSelected ? 1.5 : 1, contrast: contrast)
    }

    // MARK: - Rename plumbing

    private var isRenaming: Bool {
        store.transient.renameEditor?.targetID == lane.id
    }

    /// The draft, as a binding onto transient state rather than as `@State`: the editor's text lives
    /// in `TransientBoardState` because a reload has rules about it (the vanish discard), and a
    /// second copy in the view would be the one the commit did not read.
    private var renameDraft: Binding<String> {
        Binding(
            get: { store.transient.renameEditor?.draftTitle ?? "" },
            set: { store.transient.updateRenameDraft($0) }
        )
    }
}

// MARK: - Lane slots

/// The run of shadows a lane opens for whichever session is proposing into it — where it starts and
/// what each shadow is worth in height.
///
/// `Equatable` because it is the reflow's animation key: within a session the heights never change,
/// so the value moves exactly when the proposal does.
private struct ShadowRun: Equatable {
    var position: Int
    var heights: [CGFloat]
}

/// What a lane's masonry lays out — its cards, plus at most one pseudo-card.
///
/// The placeholder is not a `Card` and never will be: it has no disk presence and no UUID until its
/// title commits (02-architecture.md § Layering, the one named exception to the one-way flow).
/// Modelling it as a sibling case rather than as a fake `Card` is what keeps that true — nothing can
/// accidentally hand it to code expecting an item that exists.
///
/// ### The create handoff, which is entirely a question of `id`
///
/// 02-architecture.md ▸ TransientBoardState ▸ overlays (settled, 2026-07-28) requires the handoff to
/// **read as one arrival**: "the placeholder renders at the arriving card's exact geometry/chrome".
/// The mechanism is this enum's identity function and nothing else — no `matchedGeometryEffect`, no
/// second transaction, no suppression flag.
///
/// A placeholder carries its phase, and the phase decides its key:
///
/// - **`.editing`** keys to the constant `"placeholder"`. There is only ever one placeholder in one
///   lane at a time, and it must hold its identity — and therefore its keyboard focus — for as long
///   as the user types.
/// - **`.awaitingArrival(id)`** keys to `identity(of: id)`: **the very key the arriving card will
///   use**. The Writer's create has run, so the real card's UUID exists a full round trip before its
///   `Card` does, and adopting it early is what makes the handoff a *content swap inside one
///   `ForEach` element* rather than a removal and an insertion at coincident slots. The element
///   persists across the echo reload, so `Motion.cardTransition` — attached per slot in the masonry
///   — never fires on it: one arrival, one geometry, no double scale-and-fade.
///
/// The key therefore changes exactly once, at commit, and that change is deliberately outside every
/// animated transaction the board runs (`BoardStore.commitPlaceholder` is a plain synchronous call
/// from the editor's Return; `land`'s `withAnimation` comes a round trip later), so it costs no
/// motion either.
///
/// Two slots can never collide on the arriving key: `BoardStore.land` assigns the snapshot and
/// re-grounds the transient state — the discard-on-arrival among it — inside one transaction, so no
/// render pass ever sees both the real card and the placeholder standing in for it.
enum LaneSlot: Identifiable {
    case card(Card)
    /// The new-card placeholder, carrying the phase that decides both what it draws and what it is
    /// keyed by. Passed down rather than re-read in the stub so the key and the face cannot disagree
    /// about which half of the handoff this is.
    case placeholder(NewCardPlaceholder.Phase)
    /// One of a drag's N contiguous shadows, at the dragged card's frozen height.
    case shadow(index: Int, height: CGFloat)

    var id: String {
        switch self {
        case let .card(card): Self.identity(of: card.id)
        case let .placeholder(phase): Self.identity(ofPlaceholderIn: phase)
        // Constant per position in the run, so the shadows animate as slides when the proposal moves
        // rather than blinking out and back in.
        case let .shadow(index, _): "shadow:\(index)"
        }
    }

    /// A card slot's id, spelled once so the scroll-into-view call and the slot itself cannot
    /// disagree about what `scrollTo` is looking for.
    static func identity(of card: ItemID) -> String { "card:\(card.rawValue)" }

    /// The key an open editor holds while it has no identity of its own to hold.
    static let editingPlaceholderIdentity = "placeholder"

    /// **The handoff, as one pure function.** See the type's doc comment: a committed placeholder
    /// answers with the arriving card's key, which is what makes the swap continuous.
    static func identity(ofPlaceholderIn phase: NewCardPlaceholder.Phase) -> String {
        switch phase {
        case .editing: editingPlaceholderIdentity
        case let .awaitingArrival(id): identity(of: id)
        }
    }
}

// MARK: - The new-card placeholder

/// The card being created, drawn as a pseudo-card in the masonry flow at standard card width —
/// 02-architecture.md § Layering's one named exception to the one-way flow, finally rendered.
///
/// Two faces, one per phase:
///
/// - **`.editing`** — a focused text field. Return commits, Escape abandons, and **click-away
///   discards**: the placeholder's rule, "the deliberate exception because nothing exists on disk
///   yet" (04-interactions.md ▸ Grammar).
/// - **`.awaitingArrival`** — the committed title as plain text, deliberately *not* an editor. The
///   Writer's create has run and the overlay is only covering the gap until the watcher round-trips
///   the real card; leaving a live field there would invite edits that have nowhere to go, and its
///   focus loss would fire the discard rule against a card that is already on its way.
///
/// ### The two faces are two *different* faces, on purpose
///
/// The editing face is an editor — the accent-stroked well the user is typing into. The awaiting
/// face is **a card**: the same plate, inset, stripe gutter, icon, font and title position a default
/// new card gets (`CardFaceView`, via the `BoardMetrics` both read). That is the second half of
/// 02-architecture.md ▸ TransientBoardState ▸ overlays' one-arrival rule — the first half is
/// `LaneSlot` keying a committed placeholder by the arriving card's identity, which makes the echo
/// reload a content swap inside one persistent element; drawing that content identically is what
/// makes the swap invisible rather than merely un-animated.
///
/// A newly created card is always default-styled — `BoardWriter.createCard` writes `schema`, `title`
/// and `order` and nothing else — so "the arriving card's chrome" is exactly: the level-default
/// symbol, the standard secondary tint, no accent stripe, and no selection stroke (the commit
/// re-selects the *lane*).
private struct NewCardStubView: View {

    let store: BoardStore

    /// How far along the birth is — handed down from the slot rather than re-read from the store, so
    /// the face this view draws and the identity the slot is keyed by are the same answer.
    let phase: NewCardPlaceholder.Phase

    let openCard: (ItemID) -> Void

    /// Increase Contrast, for the editor well's stroke below (10-accessibility.md; `Accommodations`).
    @Environment(\.colorSchemeContrast) private var contrast

    /// The board's ruler (`BoardZoom`) — the same environment the real face reads, so the placeholder
    /// and the card that replaces it are drawn at one zoom.
    @Environment(\.boardZoom) private var zoom

    /// The live body metric — the same one the real face reads, which is what makes "the numbers are
    /// the same numbers rather than equal ones" survive the move to font-derived metrics.
    private var pointSize: CGFloat { zoom.bodyPointSize }

    var body: some View {
        switch phase {
        case .editing: editor
        case .awaitingArrival: arrivingFace
        }
    }

    /// The editor well — an accent-stroked plate around the field, which is what a thing being typed
    /// into should look like and deliberately not what a card looks like.
    private var editor: some View {
        InlineTitleField(
            text: draft,
            prompt: "Card title",
            onCommit: { commit() },
            onAbandon: { store.transient.discardPlaceholder() },
            onFocusLoss: { store.transient.discardPlaceholder() },
            onCommitAndOpen: {
                // The one board command that stays enabled mid-edit: commit, then open
                // (04 ▸ Grammar's carve-out). A commit that discarded — empty title, a
                // vanished lane, a failed create — hands back no id and opens nothing.
                if let id = commit() { openCard(id) }
            }
        )
        .boardFont(.body)
        .frame(maxWidth: .infinity, alignment: .leading)
        .padding(BoardMetrics.cardContentPadding(bodyPointSize: pointSize))
        .background(
            RoundedRectangle(cornerRadius: BoardMetrics.cardCornerRadius(bodyPointSize: pointSize))
                .fill(BoardSurface.cardPlate)
        )
        .overlay(
            RoundedRectangle(cornerRadius: BoardMetrics.cardCornerRadius(bodyPointSize: pointSize))
                // Increase Contrast takes the well to full strength and a point heavier — a 60%-alpha
                // outline is exactly the kind of border the setting exists to rescue
                // (`Accommodations`).
                .strokeBorder(
                    Color.accentColor.opacity(Accommodations.accentOpacity(0.6, contrast: contrast)),
                    lineWidth: Accommodations.borderWidth(1.5, contrast: contrast)
                )
        )
    }

    /// **The arriving card, drawn a round trip early.** Every line below has a counterpart in
    /// `CardFaceView.body`/`titleRow`, and the numbers are the same numbers rather than equal ones
    /// (`BoardMetrics`) — when the echo reload swaps this view for the real face inside the one
    /// slot they share, nothing about the plate, the icon, the font or the title's position changes.
    ///
    /// No stripe overlay and no selection stroke: both would be `.clear` for a default, unselected
    /// new card, and a shape that paints nothing is better left unwritten than written and disabled.
    private var arrivingFace: some View {
        HStack(alignment: .firstTextBaseline, spacing: BoardMetrics.cardRowSpacing(bodyPointSize: pointSize)) {
            Image(systemName: ItemSymbol.card)
                .foregroundStyle(.secondary)
                .imageScale(.medium)
            Text(store.transient.newCardPlaceholder?.draftTitle ?? "")
                .boardFont(.body)
                .lineLimit(4)
                .frame(maxWidth: .infinity, alignment: .leading)
        }
        .frame(maxWidth: .infinity, alignment: .leading)
        .padding(BoardMetrics.cardContentPadding(bodyPointSize: pointSize))
        .padding(.leading, BoardMetrics.cardStripeWidth(bodyPointSize: pointSize))
        .background(
            RoundedRectangle(cornerRadius: BoardMetrics.cardCornerRadius(bodyPointSize: pointSize))
                .fill(BoardSurface.cardPlate)
        )
    }

    /// Commits, then **re-selects the lane** — "Return commits and re-selects the lane (next Return
    /// = next card)" (04-interactions.md ▸ Grammar). The lane rather than the new card is what makes
    /// a run of Return-type-Return file a stack of cards without the user's hands leaving the
    /// keyboard.
    ///
    /// The lane is read before the commit, because every discard path clears the overlay that holds
    /// it — and re-checked after, because one of those paths is *the lane vanished*, and selecting
    /// something that renders nowhere would break the one-container invariant until the
    /// next reload swept it away.
    @discardableResult
    private func commit() -> ItemID? {
        let lane = store.transient.newCardPlaceholder?.laneID
        let id = store.commitPlaceholder()
        if let lane, store.snapshot.lanes.contains(where: { $0.id == lane }) {
            store.select([lane], in: .board)
        }
        return id
    }

    private var draft: Binding<String> {
        Binding(
            get: { store.transient.newCardPlaceholder?.draftTitle ?? "" },
            set: { store.transient.updateDraft($0) }
        )
    }
}

// MARK: - The inline title field

/// The one text field all three inline editors wear — the new-card placeholder, a card rename, and
/// a lane rename — so the grammar around it is written once (04-interactions.md ▸ Grammar).
///
/// The four exits, and who differs on them:
///
/// | Exit | Placeholder | Rename |
/// |---|---|---|
/// | Return | commits | commits |
/// | Escape | discards | abandons |
/// | Click-away | **discards** | **commits** |
/// | ⌘↩ | commits + opens | commits + opens |
///
/// Only the click-away row differs, which is why it is a caller-supplied closure rather than a
/// branch in here: this view knows *that* focus left, never what that should mean.
///
/// **Every handler must be idempotent**, because the exits overlap by construction: Return commits
/// and then the field disappears, which also fires the focus-loss handler an instant later. The
/// store's `commitRename`/`commitPlaceholder` and the transient state's `discard…` all no-op against
/// an editor that is already closed, so the overlap costs nothing.
struct InlineTitleField: View {

    @Binding var text: String
    let prompt: String
    let onCommit: () -> Void
    let onAbandon: () -> Void
    let onFocusLoss: () -> Void
    let onCommitAndOpen: () -> Void

    @FocusState private var isFocused: Bool

    var body: some View {
        TextField(prompt, text: $text)
            .textFieldStyle(.plain)
            .lineLimit(1)
            .focused($isFocused)
            // The editor is born focused: every entry point to it is a deliberate "edit this now"
            // (Return, ⌘N, the header button, a double click, Board ▸ Rename), and one that landed
            // unfocused would need a second click to do anything.
            .onAppear { isFocused = true }
            .onSubmit(onCommit)
            // ⌘↩ before the field sees the Return: the one board command enabled mid-edit
            // (04 ▸ Grammar). Anything without the modifier is passed straight through, so plain
            // Return still reaches `onSubmit`.
            .onKeyPress(keys: [.return], phases: .down) { press in
                guard press.modifiers.contains(.command) else { return .ignored }
                onCommitAndOpen()
                return .handled
            }
            // Escape reaches a focused text field as AppKit's cancel operation on some paths and as
            // a plain key press on others; both are wired to the same idempotent abandon rather than
            // guessing which one this control will get.
            .onKeyPress(.escape) {
                onAbandon()
                return .handled
            }
            .onExitCommand(perform: onAbandon)
            .onChange(of: isFocused) { _, focused in
                guard !focused else { return }
                onFocusLoss()
            }
    }
}