Charts Overview

Vizb folds the same x/y/z dimension model into every chart type — you describe your data once with a group pattern (e.g. -p x,y,z for CSV or -p x/y/z for benchmarks), and each selected renderer interprets it in parallel. One pipeline, six different shapes. Use --charts (or -c) to control which renderers are bundled into the output file.

Chart Types

Bar Chart

Compare values across categories. Supports linear and log scale, grouped 2D bars, and WebGL bar3D for z-axis data.

Line Chart

Track trends across X-axis values. One line per Y-series in 2D; switches to WebGL line3D with a z-axis.

Scatter Chart

Plot individual points in 2D or 3D. Grouped categories or auto-value from all-numeric columns.

Pie Chart

Show proportional distribution. Scales from one pie (1D) to two side-by-side (2D) to three pies (3D).

Radar Chart

Multi-dimensional comparison on a spoke diagram. Spokes change meaning across 1D/2D/3D — great for profiles and benchmarks.

Heatmap

X × Y colored grid — works with or without a z-axis. With z, cell value sums z-series and cell color blends the z palette.

Dimensions: 1D, 2D, and 3D

Every vizb chart is driven by up to three chart dimensions: X-axis (x), Y-axis (y), and Z-axis (z). You declare them with the -p flag:

Pattern	Dimensionality	What you get
-p x	1D — X only	X is the only axis
-p x,y (CSV) or -p x/y (bench)	2D — X + Y	Y subdivides each X point into groups or series
-p x,y,z (CSV) or -p x/y/z (bench)	3D — X + Y + Z	Z adds a third dimension; bar/line go WebGL, pie/radar add a third slice set, heatmap blends Z into cell color

Pie and radar have no cartesian axes, but the same -p pattern governs how many series or spokes they render:

Chart	1D (-p x)	2D (-p x,y / -p x/y)	3D (-p x,y,z / -p x/y/z)
Bar	Single bar series over X	Grouped bars, one series per Y value; or auto-value (all-numeric)	WebGL bar3D — Z stacked per (X, Y) cell; or auto-value continuous 3D
Line	Single line over X	One line per Y value across X; or auto-value (all-numeric)	WebGL line3D — sparse points per Z; or auto-value continuous 3D
Scatter	One point per X	One series per Y value across X; or auto-value (all-numeric)	WebGL scatter3D — points per Z; or auto-value continuous 3D
Pie	One pie (X = slices)	Two pies side-by-side (By X / By Y)	Three pies (By X / By Y / By Z)
Radar	One polygon, X = spokes, value = stat total	One polygon per X, Y = spokes, legend = X	One polygon per Z, X = data points, Y = spokes
Heatmap	n/a — needs at least X + Y	Matrix X × Y, single-series gradient	Same matrix; cell = Σ z, color blends z palette

Note

Heatmap requires at least a 2D pattern (-p x,y for CSV or -p x/y for benchmarks). Passing -p x with -c heatmap will produce an empty chart panel.

4D: Named Chart Groups

The n axis is the 4th dimension. Unlike x/y/z — which add visual structure inside a chart — n splits data into named groups, each rendered as its own chart panel. Every unique name value becomes a separate chart.

Pattern	Result
-p n/x	One chart per name, each a 1D view
-p n/x/y	One chart per name, each a 2D view
-p n/x/y/z	One chart per name, each a 3D view — maximum dimensionality

For CSV, -g maps columns to pattern slots positionally:

# 4D: name=algo, x=input_size, y=variant, z=run_type
vizb data.csv -g algo,input_size,variant,run_type -p n,x,y,z -c bar -o out.html

Note

The n axis is optional. Without it, all rows share a single unnamed chart. With it, each unique name value gets its own panel — useful for comparing independent algorithms, datasets, or experiments side by side.

Settings

Most settings apply to every chart type. The exceptions are scale (log), which only affects cartesian charts, and 3d-rotate (3D auto-spin), which only applies when bar, line, or scatter renders in WebGL mode.

Setting	bar	line	scatter	pie	radar	heatmap
sort / --sort	✓	✓	✓	✓	✓	✓
labels / --show-labels	✓	✓	✓	✓	✓	✓
swap (axis reorder)	✓	✓	✓	✓	✓	✓
scale log	✓	✓	✓	✗	✗	✗
3d-rotate (3D auto-rotate)	✓ 3D only	✓ 3D only	✓ 3D only	✗	✗	✗

To override a setting for a single chart type without affecting the others, use the --chart <type>:key=val syntax:

# Turn on labels for bar only
vizb data.csv -g category --chart bar:labels=true -o output.html

# Enable sort on radar, keep others unsorted
vizb data.csv -g category,metric -p x,y -c bar,radar --chart radar:sort=asc -o output.html

See Commands: root for the full flag reference.

Selecting Chart Types

The --charts flag (short: -c) controls which renderers are compiled into the output HTML. The default is:

bar,line,pie

Add scatter, heatmap, or radar explicitly when you need them: -c bar,line,pie,scatter,heatmap,radar.

Each chart renderer ships as a separate compressed chunk; only the ones you select are embedded. This keeps the output small when you only need one or two shapes:

# Radar and bar only
vizb data.csv -g category,metric -p x,y -c bar,radar -o output.html

# Single chart — smallest possible output
vizb data.csv -g category -c bar -o output.html

Tip

For sharing or deploying reports, pass --charts to include only what your audience needs. The 3D engine (~380 kB encoded) is pulled in automatically when bar, line, or scatter is selected and your data needs a 3D scene — and omitted otherwise. See Output File Size for the size breakdown.

Next Steps

SettingsCLI flags and UI settings — sort, labels, scale, swap, and more.

Axis SwappingRearrange x/y/z dimensions interactively in the UI.

StatisticsPer-series descriptive stats and a Pearson/Spearman correlation matrix.

GroupAggregate rows before charting with --group.