Module chart_window 

ChartWindow<T> — a “last N points per series” streaming projection over a crate::ChartModel.

Wraps a ChartModel<T> and exposes the tail window_size points of every series — the live-scrolling-strip-chart pattern (a sensor feed, a log-rate graph, a stock ticker). Unlike crate::ChartAggregate, ChartWindow copies no point data: it tracks, per series, the source index of the window’s first visible point (starts) and delegates every read straight through to the source. That means a ChartWindow<T> needs no T: Clone bound at all.

Reactivity

The upstream ChartChange stream is translated, not collapsed to a blanket Reset (unlike crate::SortFilterListModel, where an arbitrary sort-key move makes fine-grained translation unsafe — a fixed-size tail window has no such hazard): a tail append into a full window becomes a PointsRemoved + PointsInserted pair (the window slides), a tail append into a still-growing window becomes a plain PointsInserted, and symmetrically a tail removal (trimming the series’ own end — e.g. discarding a bad trailing reading) becomes the mirror-image PointsRemoved + PointsInserted pair: points beyond the new total drop out of the window, and if the window slid backward to stay full, the newly-uncovered prefix is revealed as an insertion. Anything that isn’t a clean tail append/removal (a mid-series insert or removal) falls back to a per-series rebuild reported as SeriesDataReplaced.

use teksilo_data::{ChartModel, ChartWindow};
let model: ChartModel<i32> = ChartModel::new();
let s = model.add_series("sensor");
for i in 0..100 {
    model.push_point(s, i, i as f32);
}
let window = ChartWindow::new(model.clone(), 10);
assert_eq!(window.point_count(s), 10); // last 10 points only

Structs

ChartWindow

A “last N points per series” streaming projection over a ChartModel<T>.