Legend and labels are not customizable enough

[Tsai002]

The current chart rendering has very limited customization support for both legends and labels, which makes the visualization difficult to use in real-world scenarios.

Legend issues:

• There is no way to hide or disable it.

Label issues:

• Labels cannot be formatted or customized. No support for custom text formatting, precision, abbreviations, prefixes/suffixes, etc.

I tried using Charton to recreate Economist-style charts, but the current customization capabilities are too limited for production-quality results.

use charton::{alt::color, core::guide::LegendPosition, prelude::*};
use polars::prelude::*;

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let mut df = CsvReadOptions::default()
        .try_into_reader_with_file_path(Some(
            "private-investment-in-artificial-intelligence-by-focus-area.csv".into(),
        ))
        .unwrap()
        .finish()
        .unwrap();

    df = df
        .lazy()
        .with_columns([col("Year").cast(DataType::String)])
        .collect()
        .unwrap();

    let mask = df.column("Entity")?.str()?.not_equal("Total");
    let df = df.filter(&mask)?;

    let entity = df.column("Entity")?.str()?;
    let highlight_colors = [
        "Data management, processing, cloud",
        "Medical and healthcare",
        "Retail",
        "Natural language, customer support",
        "Facial recognition",
    ];
    let mask: BooleanChunked = entity
        .into_iter()
        .map(|opt_v| opt_v.map(|v| highlight_colors.contains(&v)))
        .collect();

    println!("{:?}", df.shape());

    let highlight = df.filter(&mask)?;
    println!("{:?}", highlight.shape());
    let not_highlight = df.filter(&!mask)?;
    println!("{:?}", not_highlight.shape());
    let ds = load_polars_df!(not_highlight)?;
    let ds_highlight = load_polars_df!(highlight)?;

    println!("dataset: {:?}", ds_highlight.head(10));
    println!("dataset: {:?}", ds.head(10));

    let cp = vec![
        "#17648d".into(),
        "#51bec7".into(),
        "#008c8f".into(),
        "#d6ab63".into(),
        "#843844".into(),
    ];

    let highlight_chart = Chart::build(ds_highlight)?
        .mark_line()?
        .configure_line(|l| l.with_stroke_width(1.25))
        .encode((
            alt::x("Year"),
            alt::y("total_private_investment_by_focus_area_inflation_adjusted"),
            color("Entity"),
        ))?
        .configure_theme(|t| {
            t.with_palette(ColorPalette::Custom(cp))
                .with_legend_position(LegendPosition::Top)
        })
        .with_title("Annual global private investment in artificial intelligence, by focus area")
        // .with_y_ticks([
        //     -1e9, 0.0, 1e9, 2e9, 3e9, 4e9, 5e9, 6e9, 7e9, 8e9, 9e9, 1e10, 11e9, 12e10,
        // ])
        .with_y_label("");

    highlight_chart.save("highlight.svg")?;

    let base_chart = Chart::build(ds)?
        .mark_line()?
        .configure_line(|l| l.with_stroke_width(0.75).with_color("#d4dddd"))
        .encode((
            alt::x("Year"),
            alt::y("total_private_investment_by_focus_area_inflation_adjusted"),
            alt::color("Entity"),
        ))?
        .configure_theme(|t| {
            t.with_palette(ColorPalette::Custom(vec!["#d4dddd".into()]))
                .with_legend_position(LegendPosition::Top)
        })
        .with_title("Annual global private investment in artificial intelligence, by focus area")
        .with_y_label("");

    base_chart.save("base.svg")?;
    highlight_chart.and(base_chart).save("economist.svg")?;

    Ok(())
}

[wangjiawen2013]

Thanks for the detailed feedback. I completely agree that Charton’s current legend and label customization support is still quite limited for production-quality or publication-style charts. The project is still in a very early stage, and right now it mainly has the core rendering architecture in place, so many higher-level customization features are still missing.

At the moment, most of my time is going into the new wgpu backend and distributed rendering infrastructure, since those are foundational systems. Charton is also mostly a spare-time project outside of my full-time job, so progress will probably stay gradual for a while. Still, feedback like this is very valuable and helps prioritize future improvements.

[Tsai002]

Honestly, building a visualization library from scratch — especially with a modern Rust + wgpu architecture and distributed rendering ambitions — is an enormous undertaking. Even getting the core rendering pipeline stable is already a huge achievement. Many people underestimate how much engineering depth is hidden behind “just drawing charts.”

Really looking forward to seeing where Charton goes next — happy to be witnessing the growth of something that could become truly great.