Rebuild an existing report to receive the current component defaults. Previously generated HTML keeps its old appearance until you rebuild and deploy the complete output. Explicit additions below need report code changes; upgrading never enables those options automatically.
These are genuine native captures from the v0.9.5 tag and the current implementation at the time this guide was captured, using identical synthetic values. They are fixed screenshots, not live charts. Wide captures use a 900px viewport in Trellum Dark; phone captures use 390px in Trellum Light. The screenshots keep their captured theme.
Return to the component showcase for live examples, or the demo gallery.
python -m trellum.run reports/<slug> --no-serve
python -m trellum validate reports/<slug>Source values and metric definitions remain yours. Existing line observations and null gaps keep their meaning. New totals and forecasts are never inferred from column names. This guide covers a representative set of visible changes.
Compact striped rows and right-aligned numbers arrive on rebuild. Matching and CSV counts explain the population. Search and sorting now run before the display limit: max_rows limits visible rows, while search and CSV cover all matching supplied rows. Static CSV keeps raw numeric precision; numeric sorting can differ from rounded text order.
The explicit example declares currency units, a target of 14,000 and an additive total. Targets and totals do not appear by default. Totals cover matching delivered rows before the display limit. The code uses a channels DataFrame with Channel and numeric Revenue columns; import the named classes from trellum.components.
The rebuilt default uses a compact ring beside a wrapping HTML legend with all category values and shares. The center stays empty until center_total=True is authored. Totals require additive quantities and complete finite nonnegative values. Shares include legend-hidden categories; hiding an arc changes visibility, not filters or CSV.
On phones the legend stacks under the ring, so the complete component can grow taller. Existing signed/incomplete arc calculations remain unsupported for part-to-whole interpretation.
The default plot is now 220px at every viewport width. In the 900px comparison the old plot is 260px; at 390px it was 180px. Automatic dots adapt to point spacing without dropping observations. Straight lines remain the default.
Existing explicit LineChart.colors lists now apply to unsplit dataset-bound value/ratio series, where they previously were ignored. Omitted or empty lists and missing entries keep theme fallback; active category splits keep category mapping precedence. This correction is not illustrated in the frozen screenshots.
The explicit example chooses a 160px plot, a daily average header and a curve. Its daily DataFrame has ISO day and numeric revenue columns. Curves interpolate and can overshoot; they are not averages or forecasts. The eight underlying daily values stay identical.
On rebuild, ordinary LineCharts mark each dated event with a small marker in the plot and name it beside the marker where the whole name fits. Crowded names collapse to the markers. Tap a marker for its exact name and original dates, or open All events. The current 220px plot, measured values and event dates stay the same.
The comparison uses two identical supplied events. The old phone labels overlap; the current markers keep both events visible. Use event_labels="full" for in-plot labels or event_labels="grouped" for the earlier numbered-marker presentation. Other chart families keep their existing labels.
Upgrading does not add fitted lines automatically. Enable trend_line on an eligible daily LineChart or numeric ScatterChart to show a descriptive linear fit. The same observations and 220px plot appear below with the option omitted and enabled. The dash-dot overlay stops at the observed range; it does not create a forecast.
In a filtered report the fit recalculates after aggregation. Readers can hide it or open details for observation counts, slope and R². Daily ratio fits and bubble points are unweighted. Only complete snapshot sources are supported; unavailable or oversized populations show a reason instead of a sampled fit.
LineChart(daily, "day", "revenue", dataset_id="daily",
trend_line={"method": "linear", "series": "each"})Automatic grids retain chart titles and each chart family’s height. Grid columns is now a maximum even when min_width is supplied; explicitly set columns=4 if an old automatic grid relied on fitting four tiles. SplitPane stacks according to its own panel width.
Ordinary tabs become wrapping underlined navigation. Sidebar mode is an author choice and becomes a searchable selector in narrow panels. Existing control state and URL keys are preserved.
Grid(children, columns=2, min_width=320)After rebuilding your own reports, check desktop and phone widths, long legends, automatic grids and custom CSS. Keep source/query bounds for large table populations; max_rows is not a payload-size limit.