Showcase: What You Can Build¶

Well-known chart designs from the matplotlib, seaborn, Altair, and D3 traditions — built in Ferrum's Rust engine. Every image below was produced entirely in Python using Ferrum's grammar-of-graphics API; no matplotlib, no browser, no external renderer. These designs lean on Ferrum features like categorical color ranges, temporal axis auto-inference, size and shape legends, stroke routing on line marks, sort specs, annotation date coordinates, X2/Y2 floating geometry, polar coordinates, continuous-color schemes, hue-split density, per-group error extents, and two-way row-by-column faceting.

Where the Gallery catalogs each individual mark and helper, the Showcase assembles those primitives into complete, recognizable chart designs. Each card shows the concise API call alongside the rendered output.

Four-encoding charts¶

Encode x, y, size, and color simultaneously for maximum information density.

Gapminder Bubble

fm.Chart(df).mark_point().encode(x="gdp_per_capita", y="life_expectancy", size=fm.Size("population"), color="region")

Four simultaneous encodings — position, size, and color — in one readable chart. The size legend uses Ferrum's new multi-legend stacking; both legends render side by side.

Comparison and annotation¶

Charts built for the "story in the data" — where one series, one event, or one gap is the point.

Highlight Line

ctx = fm.Chart(ctx_df).mark_line(stroke="#cccccc", opacity=0.55).encode(x="x", y="y", detail="series") + fm.Chart(hi_df).mark_line(stroke="#e4572e")

Nine gray context series recede; one accent series commands attention. detail= groups the context lines without adding a color legend.
Dumbbell Chart

rule = fm.Chart(df).mark_rule(stroke="#cccccc").encode(x="score_before", x2=fm.X2("score_after"), y="category") + fm.Chart(df_pts).mark_point().encode(x="score", color="period")

Before-and-after per category. Connecting rules use x + x2 encoding; a categorical color range pins "Before" and "After" to consistent hues.
Time Series with Events

fm.annotate_rect(x1="2023-06-01", x2="2023-09-01", y1=70, y2=125, fill="#fbbf24") + fm.Chart(df).mark_line().encode(x=fm.X("date", axis=fm.Axis(label_format="%b"))) + fm.annotate_vline(x="2023-09-01")

Temporal columns auto-infer their scale. Annotation coordinates accept ISO date strings. The yellow band is annotate_rect; the rule is annotate_vline.

Ranked and sorted¶

Sorted layouts for ranked comparisons — sorted along the axis, not alphabetically.

Cleveland Dot Plot

stem = fm.Chart(df).mark_rule(stroke="#d1d5db").encode(x=fm.X("score_zero", axis=fm.Axis(title="score")), x2=fm.X2("score"), y=fm.Y("skill", scale=ordinal)) + fm.Chart(df).mark_point(fill="#3b82f6")

Lollipop composition: a mark_rule stem anchored to zero plus a mark_point dot. Category order is explicit via OrdinalScale(domain=sorted_domain).

Small multiples¶

Faceted panels with shared scales and consistent encodings across categories.

Faceted Small Multiples

fm.Chart(df).mark_point(opacity=0.65).encode(x="petal_length", y="sepal_width", color=fm.Color("species", scale=fm.OrdinalScale(domain=species_order, range=colors))).facet(col="species")

Three scatter panels sharing a common axis range. An explicit OrdinalScale(domain=..., range=...) keeps each species consistently colored across facet panels.
Two-Way Facet Trellis

fm.Chart(df).mark_point().encode(x="x:Q", y="y:Q").facet(row="row", col="col")

A 2 × 2 grid of scatter panels driven by simultaneous row= and col= facet keys — two-way faceting reveals interaction effects across both dimensions.

Polar and radial¶

Circular layouts where angle encodes a categorical dimension.

Coxcomb / Wind-Rose

fm.Chart(df).mark_bar().encode(x="month:N", y="rainfall:Q", color="season:N").coord(fm.CoordPolar(theta="x"))

Equal angular slices whose radius encodes magnitude. CoordPolar(theta="x") maps the nominal x channel to arc angle; coloring by 4-category season avoids palette wrapping across 12 months while adding a meaningful grouping.

Financial and categorical geometry¶

Floating geometry via X2/Y2 encodings plus variable-column-width mosaic layouts.

Candlestick (OHLC)

body = fm.Chart(df).mark_bar().encode(x="day:O", y=fm.Y("open:Q", axis=fm.Axis(title="Price")), y2="close:Q", color="dir:N") + wick = fm.Chart(df).mark_segment(stroke="#333333", stroke_width=1.5).encode(x="day:O", x2="day:O", y=fm.Y("low:Q", axis=fm.Axis(title="Price")), y2="high:Q") → (body + wick)

Body uses mark_bar with y/y2 for the open-close range; wick uses mark_segment for the high-low tail. The wick layer renders on top so tails are visible at both ends.
Marimekko / Mosaic

fm.Chart(df).mark_rect().encode(x="x0:Q", x2="x1:Q", y="y0:Q", y2="y1:Q", color="region:N")

Variable-width stacked columns: precompute x0/x1 per product (column widths proportional to revenue share) and y0/y1 per region (stacked proportions). mark_rect with four positional encodings places each tile precisely.
Diverging Likert

fm.Chart(df).mark_bar().encode(x="x0:Q", x2="x1:Q", y="statement:N", color=fm.Color("level:N", scale={"scheme": "redblue"}))

Bars float from a center axis by precomputing x0/x1 that straddle zero. The redblue scheme maps agreement levels from negative (red) to positive (blue); x/x2 encoding places each segment without any stacking transform.

Continuous color and density¶

Sequential and diverging continuous-color scales, plus multi-group density layouts.

Continuous-Color Heatmap

fm.Chart(df).mark_rect().encode(x="x:O", y="y:O", color=fm.Color("z:Q", scale={"scheme": "viridis"}))

A quantitative color field automatically resolves to a continuous sequential scale. The viridis scheme maps the full [-1, 1] domain from dark-purple to bright-yellow; a gradient legend replaces the usual categorical swatches.

Distribution comparisons¶

Violin and error-band marks that split or stratify by a hue variable.

Split Violin by Hue

fm.Chart(df).mark_violin().encode(x="day:N", y="tip:Q", color="smoker:N")

When color= maps to a two-level nominal variable, each violin is automatically mirrored — left half for one group, right half for the other. The inner box-plot whiskers remain for both halves.
Per-Hue Error Band

fm.Chart(df).mark_errorband().encode(x="month:Q", y="y:Q", color="model:N")

mark_errorband computes a confidence interval per (x, color) group. Two models with different noise levels produce visibly different band widths — Model A narrow, Model B wide — at every x position.