Tucan - a plotting library on top of VegaLite

[pnezis]

Tucan is an Elixir plotting library built on top of VegaLite designed to simplify the creation of interactive and visually stunning plots. With Tucan you can effortlessly generate a wide range of plots, from simple bar charts to complex composite plots, all while enjoying the power and flexibility of a clean composable functional API. Version 0.1.1 was just released

You can find the docs here.

Sample usage:

scatter =
  Tucan.scatter(:iris, "petal_width", "petal_length", tooltip: true)
  |> Tucan.color_by("species")
  |> Tucan.shape_by("species")

lines = Tucan.lineplot(:stocks, "date", "price", color_by: "symbol", x: [type: :temporal])

area =
  Tucan.area(:stocks, "date", "price", color_by: "symbol", mode: :normalize, x: [type: :temporal])

density = Tucan.density(:penguins, "Body Mass (g)", color_by: "Species", fill_opacity: 0.2)

strip =
  Tucan.stripplot(:tips, "total_bill", group: "day", style: :jitter)
  |> Tucan.color_by("sex")

boxplot = Tucan.boxplot(:penguins, "Body Mass (g)", color_by: "Species")
histogram = Tucan.histogram(:cars, "Horsepower", color_by: "Origin", fill_opacity: 0.5)

pie = Tucan.pie(:barley, "yield", "site", aggregate: :sum, tooltip: true)

donut = Tucan.donut(:barley, "yield", "site", aggregate: :sum, tooltip: true)

heatmap =
  Tucan.density_heatmap(:penguins, "Beak Length (mm)", "Beak Depth (mm)", fill_opacity: 1.0)

Tucan.vconcat([
  Tucan.hconcat([scatter, lines, area]),
  Tucan.hconcat([density, Tucan.vconcat([strip, boxplot]), histogram]),
  Tucan.hconcat([pie, donut, heatmap])
])

will generate the following:

Credits go to @jonatanklosko for the awesome VegaLite bindings.

[billylanchantin]

Hey this is looks great!

I’ve been playing with VegaLite recently. I have a Matplotlib background and I think because of that I found the new API a little confusing. I’ll give this a try!

[pnezis]

Hey @billylanchantin that was the main reason i implemented it. I have also a seaborn/matplotlib background and the learning curve of VegaLite was steep.

[jonatanklosko]

@pnezis this is amazing and charts in the docs are absolute fire!!

[josevalim]

Amazing!

[mat-hek]

Very cool! Any plans for live/dynamically updated charts? It’s the part of VegaLite that made me struggle the most.

[pnezis]

@jonatanklosko thanks!

The charts are auto-generated from plain markdown code blocks. A custom markdown processor is used that compiles the block, formats it and renders both the original code block and the vega lite spec.

The underlying package originated from a discussion with @josevalim in the ex_doc repo some time ago.

[pnezis]

Hi! dynamically updated charts as well as a high level API for interactive charts is in the plans for future releases.

[josevalim]

@pnezis would you also support heatmaps? My understanding is that they are similar to your current density heatmap, except there is no binning on X and no aggregate on Y. (X may also need to be nominal).

[pnezis]

@josevalim correct heatmaps are very similar to denisty heatmaps, but for categorical variables on both axes (ordinal or nominal in vegalite terminology) instead of numerical.

I guess I missed it, will definitely add it.

[pnezis]

@josevalim heatmap/5 added Will release a new version later this week

[outlog]

amazing work!

[pnezis]

New version released today: v0.2.0, with many new features:

• heatmap, punchcard and jointplot plots
• vruler and hruler for adding vertical/horizontal rulers on any plot
• Helper modules for configuring the Legend and the Scale
• Docs improvements, added a livebook guide

You can find the full CHANGELOG here

Any feedback welcome

[josevalim]

The new (and old) plots are great!

[pnezis]

v0.3.0 is out. It includes many new features and enhancements:

• Nx support, you can pass directly tensors as series to all plots
• imshow allows you to plot pseudo-color images
• errorbar and errorband plots added
• lollipop plot added as an alternative to bar chart
• Tucan.Geometry module provides an API for plotting common geometric shapes and polylines

You can find the full CHANGELOG here

Happy new year!

[wkgumaru]

I am trying to change the color of the bar graph that is beside the default blue. I am looking at the :color encoding but I am unsure how that works. Can someone give me an example code of how that works? I wanna use hexcodes.

[pnezis]

This is currently not supported, I will add it the new release, for now you can do:

data = [
  %{"a" => "A", "b" => 28},
  %{"a" => "B", "b" => 55},
  ...
]

plot = Tucan.bar(data, "a", "b", color: [datum: "red"])

%VegaLite{
  plot
  | spec: update_in(plot.spec, ["mark"], fn opts -> Map.merge(opts, %{"color" => "#fa1243"}) end)
}

PS. Fix pushed, you can now use fill_color: your_hex_here in the bar chart. You need to add it as git dependency for now.

[wkgumaru]

So it would be tucan.bar(data,“a”,“b”,fill_color: hex_value)??

[pnezis]

So it would be tucan.bar(data,“a”,“b”,fill_color: hex_value)??

Correct

[wkgumaru]

Can you show what I add as a git dependency?