Bump chart
A bump chart follows the rank of each series across discrete time points or conditions. Lines connect consecutive ranks and the best rank sits at the top, so crossings are the story: a line crossing another is one series overtaking another.
WITH d AS (SELECT * FROM read_csv_auto('{{DFK_BASE_URL}}data/bump.tsv'))
SELECT kuva_render(to_json({
'title': 'Rank over time',
'series': [{
'type': 'bump',
'series': (SELECT list({'name': series, 'ranks': rs} ORDER BY series)
FROM (SELECT series, list(rank ORDER BY time) AS rs FROM d GROUP BY series)),
'x_labels': (SELECT list(CAST(time AS VARCHAR) ORDER BY time)
FROM (SELECT DISTINCT time FROM d))
}]
})) AS chart;
Each series carries a ranks list with one entry per time point, and x_labels names the points. The two
must be the same length — they are paired by position.
Auto-ranking from raw values
Instead of pre-computed ranks, a series can carry values and let the chart rank them at each time point.
That is the honest mode when the input really is a score rather than a standing.
WITH d AS (SELECT * FROM read_csv_auto('{{DFK_BASE_URL}}data/bump.tsv'))
SELECT kuva_render(to_json({
'title': 'Ranked from values',
'series': [{
'type': 'bump',
'series': [
{'name': 'Alpha', 'values': [95, 80, 88, 92]},
{'name': 'Beta', 'values': [80, 95, 72, 86]},
{'name': 'Gamma', 'values': [70, 85, 95, 78]}
],
'x_labels': ['Q1', 'Q2', 'Q3', 'Q4'],
'tie_break': 'average'
}]
})) AS chart;
By default a higher value ranks first. rank_ascending: true flips it, for the measures where less is
better — race times, error counts, latency.
SELECT kuva_render(to_json({
'title': 'Lower is better',
'series': [{
'type': 'bump',
'series': [
{'name': 'Alpha', 'values': [12.4, 11.8, 11.9, 11.2]},
{'name': 'Beta', 'values': [13.1, 12.6, 11.5, 10.9]},
{'name': 'Gamma', 'values': [11.9, 12.2, 12.0, 11.6]}
],
'x_labels': ['Q1', 'Q2', 'Q3', 'Q4'],
'rank_ascending': true
}]
})) AS chart;
Highlight mode
highlight names one series to emphasise: it gets a heavier stroke and bolder endpoint labels while
everything else drops to 20 % opacity. It is the standard way to make a bump chart about one subject
without deleting the context.
WITH d AS (SELECT * FROM read_csv_auto('{{DFK_BASE_URL}}data/bump.tsv'))
SELECT kuva_render(to_json({
'title': 'Highlighting one series',
'series': [{
'type': 'bump',
'series': (SELECT list({'name': series, 'ranks': rs} ORDER BY series)
FROM (SELECT series, list(rank ORDER BY time) AS rs FROM d GROUP BY series)),
'x_labels': (SELECT list(CAST(time AS VARCHAR) ORDER BY time)
FROM (SELECT DISTINCT time FROM d)),
'highlight': 'Beta'
}]
})) AS chart;
Missing time points
A null in ranks means the series was absent at that point, and the line breaks rather than interpolating
through it — which is the difference between "not measured" and "ranked last".
SELECT kuva_render(to_json({
'title': 'With gaps',
'series': [{
'type': 'bump',
'series': [
{'name': 'Alpha', 'ranks': [1, NULL, 2, 1]},
{'name': 'Beta', 'ranks': [2, 1, 1, 3]},
{'name': 'Gamma', 'ranks': [3, 2, NULL, 2]}
],
'x_labels': ['A', 'B', 'C', 'D']
}]
})) AS chart;
Tie-breaking
When auto-ranking finds equal values, tie_break decides the ranks:
tie_break | Ranks awarded |
|---|---|
"average" | The average of the positions they occupy (e.g. 2.5, 2.5) (default) |
"min" | All of them get the best rank |
"max" | All of them get the worst rank |
"stable" | They keep their input order |
SELECT kuva_render(to_json({
'title': 'Ties, broken by minimum',
'series': [{
'type': 'bump',
'series': [
{'name': 'Alpha', 'values': [90, 90, 80]},
{'name': 'Beta', 'values': [90, 70, 95]},
{'name': 'Gamma', 'values': [60, 90, 85]}
],
'x_labels': ['Q1', 'Q2', 'Q3'],
'tie_break': 'min'
}]
})) AS chart;
Average ranks are fractional, which is why ranks accepts decimals — 2.5 is a legitimate pre-ranked
value meaning "tied for second".
Curve style and labels
| Field | Default | What it sets |
|---|---|---|
curve_style | "sigmoid" | "sigmoid" curves through the point; "straight" joins them with a line |
show_rank_labels | false | Write the rank number inside each dot |
show_series_labels | true | Series names at the left and right edges |
dot_radius | 6 | Dot radius in pixels |
stroke_width | 2.5 | Line width in pixels |
legend | true | Show the legend |
SELECT kuva_render(to_json({
'title': 'Straight lines, rank labels',
'series': [{
'type': 'bump',
'series': [
{'name': 'Alpha', 'ranks': [1, 3, 2, 1]},
{'name': 'Beta', 'ranks': [2, 1, 1, 3]},
{'name': 'Gamma', 'ranks': [3, 2, 3, 2]}
],
'x_labels': ['2021', '2022', '2023', '2024'],
'curve_style': 'straight',
'show_rank_labels': true,
'dot_radius': 8,
'stroke_width': 2
}]
})) AS chart;
Fields
| Field | Type | What it sets |
|---|---|---|
series | series[] | Required. Each entry is {name, ranks?, values?, color?}. |
ranks | (number | null)[] | Pre-computed ranks; null breaks the line. |
values | (number | null)[] | Raw values, ranked automatically; null breaks the line. |
x_labels | string[] | One label per time point. |
curve_style | string | "sigmoid" (default) or "straight". |
rank_ascending | boolean | Lower values rank first (default off). |
tie_break | string | "average" (default) · "min" · "max" · "stable". |
highlight | string | Emphasise one series by name; the rest are muted. |
show_rank_labels / show_series_labels | boolean | Rank numbers in the dots, series names at the edges. |
dot_radius / stroke_width | number | Dot and line sizes. |
legend | boolean | Show the legend (default on). |
Notes
seriesmust not be empty, and every entry needs anameplus eitherranksorvalues.- Do not mix the two within one chart:
valuesare ranked within the group that gave values, so a series withranksand a series withvaluesare on different scales. - Use
nullfor an absent time point, not0—0is a rank, and it would sort above rank 1. x_labelsmust cover every time point; a shorter list leaves the extra points unlabelled.- The best rank is at the top, and
rank_ascendingonly changes which raw values get the low numbers — it does not flip the axis.
See also
- kuva — Bump chart — the plotting library's own reference for this chart.
- Slope — the two-time-point version.
- Parallel coordinates — ranks across more than a time axis.