Latex labels not rendered under very specific conditions

[mtrocadomoreira]

What happened?

I was plotting some Datasets with fairly complicated labels and units, and I noticed a very strange behaviour when using the plot() method.

When both attrs["long_name"] and attrs["units"] contain \mathrm's or \text's, and at least one of them contains a \frac, and if the length of both strings passes a certain threshold, the label on the plot is not rendered as Latex anymore.

This is a sample DataArray plotted with xarray's plot() method:

And this is the same data plotted with matplotlib:

See code below to reproduce these images.

What did you expect to happen?

I expected the Latex label to be rendered correctly, as it is when I apply the label with matplotlib.

Minimal Complete Verifiable Example

import matplotlib.pyplot as plt
import numpy as np
import xarray as xr
xr.show_versions()

data = xr.DataArray(
    np.random.randn(5, 10),
    dims=("y", "x"),
    coords={"x": np.linspace(0, 3, 10), "y": np.linspace(0, 20, 5)},
)

label = r"$\frac{\mathrm{x}}{\mathrm{A}}$"
units = r"$\mathrm{m~hello~very~long}$"

data["x"].attrs["long_name"] = label
data["x"].attrs["units"] = units

data.attrs["long_name"] = label
data.attrs["units"] = units

# with xarray

fig1 = plt.figure()
plt.title("with xarray .plot")
data.plot()

# with matplotlib

fig2 = plt.figure()

im = plt.pcolormesh(
    data["x"].to_numpy(),
    data["y"].to_numpy(),
    data.to_numpy()[:-1, :-1],
)

cb = fig2.colorbar(im)
cb.set_label(label + f"[{units}]")

plt.title("with plt.pcolormesh")

plt.xlabel(label + f"[{units}]")

Steps to reproduce

Use MCVE above to produce two figures and compare the labels in each figure. Try making the labels shorter, or removing \frac or \mathrm from one of them.

MVCE confirmation

• Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
• Complete example — the example is self-contained, including all data and the text of any traceback.
• Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
• New issue — a search of GitHub Issues suggests this is not a duplicate.
• Recent environment — the issue occurs with the latest version of xarray and its dependencies.

Relevant log output

Anything else we need to know?

No response

Environment

Details

INSTALLED VERSIONS

commit: None
python: 3.13.14 | packaged by conda-forge | (main, Jun 12 2026, 09:44:24) [Clang 19.1.7 ]
python-bits: 64
OS: Darwin
OS-release: 25.5.0
machine: arm64
processor: arm
byteorder: little
LC_ALL: en_US.UTF-8
LANG: en_US.UTF-8
LOCALE: ('en_US', 'UTF-8')
libhdf5: 2.0.0
libnetcdf: None

xarray: 2026.7.0
pandas: 3.0.3
numpy: 2.4.6
scipy: 1.18.0
netCDF4: None
pydap: None
h5netcdf: 1.8.1
h5py: 3.16.0
zarr: None
cftime: None
nc_time_axis: None
iris: None
bottleneck: None
dask: 2026.7.1
distributed: None
matplotlib: 3.11.0
cartopy: None
seaborn: 0.13.2
numbagg: 0.9.4
fsspec: 2026.6.0
cupy: None
pint: 0.25.3
sparse: None
flox: 0.11.2
numpy_groupies: 0.11.3
setuptools: 83.0.0
pip: 26.1.2
conda: None
pytest: 9.1.1
mypy: None
IPython: 8.39.0
sphinx: None

[dcherian]

Yes we call a textwrap with a hard break that doesn't know about latex so for very long ones it ends up breaking the rendering.

I don't know of a simple-ish solution to this issue that isn't fragile.