curved-text 0.2.0

Draw text along an arbitrary curve in matplotlib, with arc-length positioning and a perpendicular offset.

curved-text

Draw text that follows an arbitrary curve in matplotlib.

Label curves along their own paths instead of in a legend, so the eye never leaves the data to decode a colour key.

Each character is placed in display coordinates and rotated to the chord across its own advance, which follows the curve's local tangent while staying smooth even when the curve is coarsely sampled. The layout is recomputed on every draw, so the label keeps following the curve through layout, resizing, and interactive panning or zooming. Placement is controlled by three independent parameters:

• pos -- where the label is anchored along the curve, as a fraction of arc length (0.0 = first point, 1.0 = last).
• anchor -- which part of the label lands at pos: "start", "center", or "end".
• offset -- a perpendicular shift off the curve, in typographic points, along the normal of the label's chord (positive is above a left-to-right curve).

A label that overruns either end of the curve is not clipped: the curve is extended along its end tangent and the overrunning glyphs sit on that straight extension.

If you know LaTeX, this is matplotlib's analogue of TikZ's text along path decoration (from decorations.text): pos/anchor play the role of text align and the indents, offset plays the role of raise, and overrunning text rides the tangent extension instead of being truncated at the path's end.

Install

pip install curved-text

Or, from a clone, an editable install:

pip install -e .

Usage

import numpy as np
import matplotlib.pyplot as plt
from curved_text import curved_text

x = np.linspace(0, 2 * np.pi, 400)
y = np.sin(x)

fig, ax = plt.subplots()
ax.plot(x, y)
curved_text(ax, x, y, "text that follows the curve",
            pos=0.5, anchor="center", offset=6.0, color="C3")
plt.show()

More worked examples -- the three placement controls, overrun behaviour, and integration with seaborn and pandas -- are in examples/.

The object-oriented form is also available:

from curved_text import CurvedText

CurvedText(x, y, "along the curve", ax, pos=0.2, anchor="start", offset=4.0)

Note the axes argument position: the CurvedText class takes it after x, y, text (matching matplotlib.text.Text), while the curved_text function takes it first (matching matplotlib's axes-first helper functions).

Any extra keyword arguments (color, fontsize, alpha, fontfamily, ...) are passed through to each character's matplotlib.text.Text.

Mathtext

A $...$ run in the label is laid out by matplotlib's mathtext engine and bent through the same arc-length frame as plain text. Every glyph outline and rule box is mapped through the curve, so radicals, fractions, and sized delimiters stay connected and follow it at any curvature. Plain and math runs mix freely in one string:

curved_text(ax, x, y, r"flux $\propto \sqrt{D_{\mathrm{eff}}}\,(L/L_0)^2$",
            pos=0.5, anchor="center", offset=8.0)

Pass parse_math=False to treat dollar signs literally. Tall expressions compress vertically on the inside of tight bends, so choose the label size relative to the curvature. text.usetex is not supported.

Works with seaborn, pandas, and other matplotlib-backed libraries

curved_text needs a matplotlib.axes.Axes, so it works with any library that draws on matplotlib. seaborn's axes-level functions return an Axes, its figure-level functions expose one through .axes, and pandas DataFrame.plot returns an Axes as well. Pass that axes in directly:

import seaborn as sns

ax = sns.lineplot(data=df, x="x", y="y")
curved_text(ax, df["x"], df["y"], "along the curve",
            pos=0.5, anchor="center", offset=6.0)

Notes

• The curve (x, y) should be ordered along the curve (monotonic in arc length) and have at least two points.
• Arc length and the offset are computed in display space, so spacing and the offset are correct at any DPI and figure size.
• The curve should be smooth relative to the glyph size. Each glyph (and each mathtext run) follows the local tangent, so when the curve reverses direction within a glyph's width -- as raw noisy data does -- the label collides with itself. Label a smoothed or fitted trend line rather than the raw samples.

License

MIT