py-probelab

Shared ProbeLab utilities. Install as py-probelab, import as probelab.

uv add py-probelab
# or: pip install py-probelab

Today this is brand colors and matplotlib styling, so every chart we publish looks like it came from the same place. More common utilities will land here over time.

Charts

import matplotlib.pyplot as plt
import numpy as np
import probelab.plot as pplot

pplot.use()

fig, ax = plt.subplots()
ax.plot(np.arange(100), np.random.randn(100).cumsum(), label="dht")
ax.set_title("Lookup latency")
ax.set_xlabel("time")
ax.legend()

pplot.watermark(fig, loc="lower right")
fig.savefig("latency.png")

use() applies the theme globally for the rest of the process. Inside library code, or when you only want it for one figure, use the context manager instead so the caller's session is left alone:

with pplot.style():
    fig, ax = plt.subplots()
    ...

Importing probelab.plot registers the theme with matplotlib itself, so you can skip the helpers entirely and compose it with any other style:

import probelab.plot  # noqa: F401  - registers the style

plt.style.use("probelab-dark")
plt.style.use(["probelab-dark", "seaborn-v0_8-poster"])

Watermark

pplot.watermark(fig)                      # centered, faint
pplot.watermark(ax, loc="upper right")    # per-panel
pplot.watermark(fig, alpha=0.2, zoom=0.1) # more visible, smaller

Pass a Figure to place the mark in figure coordinates (usually what you want for a multi-panel chart) or an Axes to place it within one panel. Positions are center, upper left, upper right, lower left, lower right.

Colors

probelab.colors has no third-party imports, so it is safe to use from dashboards, reports and terminal output as well as from plotting code:

from probelab import colors

colors.PALETTE           # the 9 categorical series colors, in cycle order
colors.categorical(4)    # first 4 of them
colors.PRIMARY           # "#FFD7C5" - titles, axis labels
colors.BASE_100          # page/figure background

Fonts

The package vendors eight static DM Sans weights and registers them with matplotlib on first use, so charts render identically on a laptop and in CI without anyone installing fonts system-wide.

pplot.fonts_available()   # False means charts fell back to a system sans
pplot.use(fonts=False)    # prefer a system-wide DM Sans instead

Development

uv sync
uv run pytest
uv build

Changing the theme

src/probelab/plot/styles/probelab-dark.mplstyle is the source of truth for the theme; src/probelab/colors.py mirrors its color tokens for non-matplotlib consumers. A test asserts the two agree, so change them together.

Two things to know about editing .mplstyle files:

• Hex colors are written without a leading #, because # starts a comment there. figure.facecolor: 2e3438, not #2e3438 - the latter silently parses as an empty value and raises on load.
• Font weights must match a weight the vendored family actually declares, or matplotlib logs a findfont warning on every render. DM Sans ExtraLight declares 250, not 200.

Licensing

The code is MIT. The bundled DM Sans family is redistributed under the SIL Open Font License 1.1 (src/probelab/plot/assets/fonts/OFL.txt).