Tryst::RangeSlider

A dual-thumb range slider. Built on OwnerDrawnWidget, same App-layer pattern as ValueSlider/Switch/SegmentedControl: no ui.<type>, no bind: (see CUSTOM_WIDGETS.md for why a stateful, animated widget doesn't fit the WidgetType/AppContract seam).

slider = Tryst::RangeSlider.new(app, min: 0.0, max: 100.0, low: 20.0, high: 70.0)
slider.pack
slider.on_action { |(low, high)| puts "now #{low}..#{high}" }

One canvas, two thumbs, one focus

OwnerDrawnWidget tracks exactly one hover/focused/pressed state per canvas - there's no per-region concept, and layering a second real Tk widget over a thumb to give it its own focus/hit-target doesn't work either: a plain Tk widget is OPAQUE (paints its own solid -background), and there's no flat color that could stand in for an antialiased circle sitting on a track that's sometimes fill-colored and sometimes not. So this widget has ONE real Tab stop (the canvas), same as every sibling, plus @active_thumb tracking which of the two DRAWN thumbs currently owns keyboard input and the focus ring. Clicking either thumb (nearest-hit-test in #on_press, not a second widget) makes it active; Tab, while the canvas already has focus, cycles active thumb low->high without leaving the widget, then lets a second Tab through to real Tab traversal once already on high (signal.break! is what makes that non-trapping - see #initialize).

Value flow

#low=/#high=/#set_range set values programmatically and do NOT fire #on_action - only a user-driven change (drag, click-to-position, keyboard) does, the same split every other stateful widget in this codebase draws. The two thumbs can never cross: each clamps against the other's CURRENT position plus min_gap (defaults to step), so dragging one past the other simply stops it min_gap away rather than swapping which is "low" and which is "high".