Encapsulated Figure Types

Writing the full layers stack by hand is flexible but verbose. For the two most common physics figures, set type: on the figure and pass a handful of keys — Jarvis-PLOT expands it into the equivalent layers (data transform, render, contours) automatically.

A typed figure is pure syntactic sugar: after expansion it is an ordinary figure, so you can mix typed and hand-written figures in the same file, override the auto-built frame/style, and stack your own layers on top with extra_layers.

Available types

type	Figure	Page
posterior_2d	2D posterior probability-density map (Bayesian)	posterior_2d
profile_2d	2D profile-likelihood map (frequentist)	profile_2d

Shared structure

Every typed figure follows the same shape. The type-specific keys are documented on each page above.

- name: my_figure
  type: posterior_2d                       # or profile_2d
  data: my_samples                         # DataSet name, or a list → concatenated
  x: {expr: xx, lim: [0, 5], label: "$x$"}
  y: {expr: yy, lim: [0, 5], label: "$y$"}
  # ... type-specific keys ...
  colorbar: { ... }                        # colorbar label / cmap / vmin / vmax
  style_card: [a4paper_2x1, rectcmap]      # optional: override the auto-picked card
  extra_layers: [ ... ]                    # optional: standard layers drawn on top

Common keys

Key	Required	Default	Purpose
type	yes	—	posterior_2d or profile_2d
data	yes	—	DataSet name, or a list of names (concatenated)
x, y	yes	—	Axis coordinate dicts (see below)
name	no	the type name	Output file name
colorbar	no	see below	Colorbar styling
style_card	no	[a4paper_2x1, rectcmap]	Override the auto-selected layout card
extra_layers	no	—	Standard layers appended on top
frame	no	auto	Deep-merged over the auto-built frame

Coordinate dicts

In a typed figure the coordinate dicts accept a label (used to set the axis label) in addition to the usual expr / lim / scale:

x:
  expr: "np.log10(mass)"   # value (required)
  lim: [0.1, 10]           # axis range
  scale: log               # linear | log
  label: "$m$ [GeV]"       # axis label (defaults to the expr text)

Colorbar block

Both types accept the same colorbar block, which fills the colorbar axis (axc):

Key	Default	Purpose
label	density / the z label	Colorbar label
cmap	jarvis_rainbow2_r	Colormap
scale	linear	linear or log
vmin	0 (posterior) / auto (profile)	Lower color limit
vmax	auto	Upper color limit

When to use the long form instead

Reach for hand-written layers when you need something the typed figures don't expose: more than the built-in layers, custom per-layer transforms, multi-panel cards, ternary axes, or any non-standard method. The two paths produce identical figure objects, so there is no penalty for switching — you can even copy what a typed figure expands to (run with --debug) and edit it by hand.

---

See also: Plot Methods · Transforms · Figures and Layers