Styling and colors

Per-figure style

Each figure owns a Style (there is no global rcParams). Mutating one figure’s style never affects another.

fig, ax = plotpress.subplots()
fig.style.line_width = 2.5
fig.style.font_family = "Liberation Sans, Arial, sans-serif"

Note

Font choice has a caveat worth reading before you change it – see Fonts and layout below.

Create a variant without mutating the original with Style.copy(**overrides), and pass it to subplots():

dark = plotpress.Style(facecolor="#111", text_color="#eee",
                        axes_facecolor="#111", spine_color="#888")
fig, ax = plotpress.subplots(style=dark)

Style fields

Field	Meaning
`facecolor`	figure background
`dpi`	pixels per inch (figure px = `figsize * dpi`)
`axes_facecolor`	axes background
`spine_color` / `spine_width`	axes frame
`font_family`	CSS font stack for text (see Fonts and layout)
`font_size` / `title_size` / `label_size`	text sizes
`text_color`	all text
`tick_size` / `tick_width` / `tick_label_size`	ticks
`grid_color` / `grid_width` / `grid_alpha`	grid lines
`line_width`	default line width
`marker_size`	default scatter diameter (points)
`color_cycle`	list of colors cycled per axes (default tab10)

Fonts and layout

A figure is laid out before anything draws its glyphs, so plotpress predicts text width from bundled advance-width tables. It bundles Helvetica, Times and Courier – each in regular, bold, italic and bold-italic – plus DejaVu Sans, which covers their metric-compatible clones too: Arial and Liberation Sans are measured as Helvetica, Liberation Serif as Times, Liberation Mono as Courier.

Families outside those groups – Verdana, Tahoma, Arial Black, Arial Narrow – have proprietary metrics matching nothing bundled, so they are measured as Helvetica. They still render, but expect to hand-tune figsize and spacing.

Measuring the fonts you actually have

If you need one of those unmeasurable families to lay out correctly, opt into measuring the real file:

style = plotpress.Style(font_family="Verdana, sans-serif",
                         measure_installed_fonts=True)

Verdana is about 14% wider than the Helvetica it is otherwise measured as, so this widens its margins to what it actually needs. It needs no extra install – Pillow already ships for PNG export and does the measuring.

It is off by default, because it trades away the property the bundled tables exist to provide: with it on, layout depends on which fonts this machine has, so the same script can produce different margins on a colleague’s box or on CI. Turn it on when fidelity on your own machine is worth more than reproducibility across machines. If no candidate file resolves, it silently falls back to the bundled tables rather than failing.

See Only bundled metric families are measured accurately for the full table and the reasoning, and Font metrics: which families plotpress can measure for the measurements.

Colormaps and normalization

28 built-in colormaps – perceptually uniform ("viridis", "plasma", …), sequential ("Blues", "YlOrRd", …), diverging ("coolwarm", "Spectral", …), cyclic ("twilight"), a few classics ("jet", "turbo"), and qualitative/categorical ("tab10", "tab20", "Set1", "Dark2" – banded, not interpolated, for class labels with no natural ordering). See Colormap reference for every one of them as a gradient swatch. Append "_r" to any name to reverse it.

plotpress.available_colormaps()

List the available colormap names.

plotpress.get_cmap(name)

Return a 256x3 uint8 lookup table (or pass an array through).

plotpress.make_cmap(colors, n=256)

Build a continuous colormap by interpolating any list of colors (names, hex, or RGB(A) tuples) – for a custom or brand-specific scale that isn’t one of the 28 built-in names.

plotpress.make_listed_cmap(colors, n=256)

Like make_cmap, but banded rather than interpolated – for categorical data, the same shape as "tab10"/"Set1".

plotpress.register_cmap(name, colors_or_lut, listed=False)

Register a custom colormap (built from a color list, or a ready-made (n, 3) uint8 LUT) under name, so it becomes usable by name – cmap="name", the reversed "name_r", and available_colormaps() – everywhere a built-in colormap is.

plotpress.Normalize(vmin=None, vmax=None)

Linearly map data to [0, 1] for colormapping. Unset limits are inferred from the data on first use. Pass to pcolormesh/imshow/scatter via norm= (or use the vmin/vmax shortcuts).

norm = plotpress.Normalize(0, 1)
ax.pcolormesh(x, y, Z, cmap="plasma", norm=norm)

plotpress.TwoSlopeNorm(vcenter=0.0, vmin=None, vmax=None)

Keep a diverging colormap’s neutral color pinned to vcenter even when vmin/vmax aren’t symmetric around it – a plain Normalize puts the midpoint at (vmin + vmax) / 2, which only lands on a meaningful value (zero, for an anomaly field) by coincidence.

plotpress.BoundaryNorm(boundaries, ncolors=None)

Map data into discrete bins defined by boundaries instead of a gradient – classified rasters, risk tiers, significance thresholds. Figure.colorbar’s ticks land on the boundaries themselves.