Struct GeomLabel 

pub struct GeomLabel {
    pub size: f64,
    pub color: (u8, u8, u8),
    pub fill: (u8, u8, u8),
    pub alpha: f64,
    pub padding: f64,
    pub hjust: f64,
    pub vjust: f64,
    pub fontfamily: String,
    pub check_overlap: bool,
    pub wrap_width: Option<f64>,
    pub line_height: f64,
    pub radius: f64,
}

Label geometry — like text but with a background rectangle.

Fields

size: f64

color: (u8, u8, u8)

fill: (u8, u8, u8)

alpha: f64

padding: f64

hjust: f64

Horizontal justification: 0.0 = left, 0.5 = center (default), 1.0 = right.

vjust: f64

Vertical justification: 0.0 = bottom, 0.5 = middle (default), 1.0 = top.

fontfamily: String

Font family name (informational; actual rendering depends on backend).

check_overlap: bool

When true, skip drawing labels that overlap previously drawn labels.

wrap_width: Option<f64>

Word-wrap width in px (ggtext / str_wrap behaviour), measured with the theme’s text metrics (crate::render::text::text_width). The background box grows to fit however many lines result, plus padding. None (the default) never wraps — unchanged from pre-#176 behaviour.

line_height: f64

Line height as a multiple of size, applied between wrapped lines. Only matters when wrap_width is set and the label wraps to more than one line.

radius: f64

Corner radius in px for the background box (ggplot2 geom_label(label.r), #176). 0.0 (the default) is a plain rectangle, unchanged from pre-#176 behaviour.

Implementations

impl GeomLabel

pub fn with_hjust(self, hjust: f64) -> Self

pub fn with_vjust(self, vjust: f64) -> Self

pub fn with_fontfamily(self, family: &str) -> Self

pub fn with_check_overlap(self, check: bool) -> Self

pub fn with_wrap(self, width_px: f64) -> Self

Word-wrap the label to lines no wider than width_px, growing the background box to fit (#176).

pub fn with_line_height(self, line_height: f64) -> Self

Line height as a multiple of size for wrapped text (default 1.2).

pub fn with_radius(self, radius_px: f64) -> Self

Corner radius in px for the background box (ggplot2 label.r, #176).

Trait Implementations

impl Default for GeomLabel

fn default() -> Self

Returns the “default value” for a type.

impl Geom for GeomLabel

fn draw( &self, data: &DataFrame, coord: &dyn Coord, scales: &ScaleSet, theme: &Theme, backend: &mut dyn DrawBackend, ) -> Result<(), RenderError>

Draw this geometry.

fn required_aes(&self) -> Vec<Aesthetic>

Required aesthetics.

fn default_stat(&self) -> Box<dyn Stat>

Default stat for this geom.

fn default_position(&self) -> Box<dyn Position>

Default position adjustment.

fn default_params(&self) -> GeomParams

Non-mapped visual defaults.

fn name(&self) -> &str

Name for debug/display.

fn apply_element_geom(&mut self, el: &ElementGeom)

Recolour this geom’s built-in default colour/fill from the theme’s ink/paper/accent palette (B2-7’s element_geom), called once per layer before set_series_color — and only when the layer was not configured explicitly (geom_*_with(...); the build pipeline gates the call on the same explicit_style flag set_series_color’s brand-colour application is gated on). That gate, not a field-value comparison, is what keeps an explicit colour untouched: a geom that honours this replaces a field still at its own hard-coded default (black/(50, 50, 50) → ink, white → paper) by comparing its value, which cannot by itself tell an explicitly-set colour that happens to equal that default (e.g. GeomPoint { color: (0, 0, 0), .. } under geom_point_with) apart from one that was simply never touched — only skipping the call entirely for an explicitly-configured layer closes that gap (review PR#141 finding 4). Default no-op; None fields of ElementGeom change nothing, so the default theme (no ink/paper/accent set) stays byte-identical.

fn set_series_color(&mut self, _color: (u8, u8, u8))

Apply a brand/primary color to this geom’s single-series default (its color or fill). The build pipeline calls this only when the layer has no color/fill aesthetic mapped, so an explicit mapping always wins. The default is a no-op; series geoms override it.

fn include_zero_baseline(&self) -> bool

Whether this geom draws from a 0 baseline, so the Y scale should include 0 even when y is explicitly mapped (bars/columns/area/histograms — as in ggplot2). Default false.

fn allows_infinite(&self) -> bool

Whether -Inf/Inf position values are meaningful for this geom (ggplot2: “extend to the panel edge”). Rows with infinite positions are otherwise dropped with a warning before stats run. NaN is always dropped — unless Geom::keeps_na_as_break says otherwise. Default false; geom_rect returns true.

fn keeps_na_as_break(&self) -> bool

Whether a row with a non-finite position is a break rather than a removal (ggplot2’s GeomPath/GeomLine NA handling): the row still counts toward the “removed N rows” warning, but it is not deleted from the built DataFrame, so the geom’s own per-group point list still contains it — as a non-finite pixel position after the scale and coordinate transform, which crate::render::backend::DrawBackend::draw_line splits on into separate polylines (ggplot2’s NA-break semantics) instead of silently joining the finite points across the gap. Default false (the row is dropped, as for every other geom); line/path/step return true.

fn consumes_tooltips(&self) -> bool

Whether this geom calls [set_row_extras] (so it reads Aes::tooltip and Aes::data_attr, and a GGPlot::tooltip template on it has a visible effect). Default false; the build pipeline warns when a tooltip() template is set on a layer whose geom returns false here (review PR#141 finding 6, LOW: a template silently did nothing on e.g. geom_area/geom_boxplot/geom_ribbon). Point, segment, curve, path, line, hex, polygon, sf, rect/tile and col/bar/ histogram return true.

fn setup_data(&self, _data: &mut DataFrame, _warnings: &mut Vec<String>)

Geom-specific data preparation after the stat and position steps but before scale training (ggplot2’s GeomX$setup_data) — e.g. a tile adds its xmin/xmax/ymin/ymax extents so continuous scales train on them. warnings is the layer’s own build-warnings sink (B3-5: added so GeomBoxplot can warn on an unparsable outlier value here, the same channel every other build warning uses — never eprintln!). Default: no-op.

fn inherit_aes(&self) -> bool

Whether the layer inherits the plot-level aesthetic mapping (ggplot2’s inherit.aes). Reference-line geoms return false, as in ggplot2, so a plot-level color/x/y mapping does not leak into a geom_hline. Default true.

fn mapped_only_columns(&self) -> &'static [&'static str]

Canonical column names this geom only reads when the user actually mapped the corresponding aesthetic (D3, B3-5) — never from a raw column that merely happens to share the name (e.g. a SQL alias literally called lower). build_layer drops any of these columns right after resolve_mappings unless the aesthetic was mapped. Default empty; GeomBoxplot returns ["lower", "middle", "upper", "outliers"].

fn validate_layer( &self, _pre_stat_columns: &[String], _stat_name: &str, ) -> Result<(), String>

Stat-dependent layer validation beyond required_aes (B3-5): called once per layer with the resolved (pre-stat, post-D3-drop) column names and the layer’s stat name, before the stat runs. Default no-op. GeomBoxplot requires lower/middle/upper to be mapped under stat="identity", else y.