attributions severely wrapped upon savefig

[Matthew-J-Payne]

Hey contextily, thanks for all your work for this library - really great.

Apologies for the gigantic .pngs!

So when I visualise a plot using contextily.add_basemap(), when viewing the plot using jupyter inline, I get a nice unwrapped attribution (showing you via windows snipping tool, hence low dpi), which I would like to be reflected when the .png is saved:

But, when saving I get this:

My script, below:

import geopandas as gpd
import contextily as ctx

rivers_shapefile = gpd.read_file('datasets/rivers.shp')
rivers_shapefile['class'] = 'rivers'

fig, ax = plot.subplots(figsize = (8, 8))
rivers_shapefile.plot(column = 'class',
                         ax = ax,
                         cmap = 'bwr',
                         linewidth = 0.6,
                         zorder = 4,
                         alpha = 0)
ctx.add_basemap(ax = ax,
                source = ctx.providers.CartoDB.Voyager,
                zorder = 1,
                zoom = 8,
                attribution = "© OpenStreetMap contributors 2023. Distributed under the Open Data Commons Open Database License (ODbL) v1.0",
                attribution_size = 6,
                crs = 32649)

ax.xaxis.set_major_locator(plot.MaxNLocator(4))
ax.yaxis.set_major_locator(plot.MaxNLocator(4))

custom_xticklabels_4 = ["0", "114.1°E", "114.6°E", "115.1°E", "115.6°E"]
custom_yticklabels_4 = ["0", "2.7°N", "3.3°N" , "3.8°N", "4.3°N"]
ax.set_xticklabels(custom_xticklabels_4)
ax.set_yticklabels(custom_yticklabels_4)

fig.savefig(fname = "outputs/for_pr.png", dpi = 1200, facecolor = "white", bbox_inches = "tight")
plot.show() # plots for debugging, defaults to low res plot

The plot is for a paper in review, so I can't publicly share the data, but happy to send across.
Contextily version is 1.3.0 and geopandas 0.12.2.

Any pointers in the right direction is appreciated, thank you!

All the best,
Matt

[darribas]

Perhaps related to #95 ?

The actual method is at:

contextily/contextily/plotting.py

Lines 259 to 298 in d821ac7

	def add_attribution(ax, text, font_size=ATTRIBUTION_SIZE, **kwargs):
	"""
	Utility to add attribution text.
	Parameters
	----------
	ax : AxesSubplot
	Matplotlib axes object on which to add the attribution text.
	text : str
	Text to be added at the bottom of the axis.
	font_size : int
	[Optional. Defaults to 8] Font size in which to render
	the attribution text.
	**kwargs : Additional keywords to pass to the matplotlib `text` method.
	Returns
	-------
	matplotlib.text.Text
	Matplotlib Text object added to the plot.
	"""
	# Add draw() as it resizes the axis and allows the wrapping to work as
	# expected. See https://github.com/darribas/contextily/issues/95 for some
	# details on the issue
	draw()
	text_artist = ax.text(
	0.005,
	0.005,
	text,
	transform=ax.transAxes,
	size=font_size,
	path_effects=[patheffects.withStroke(linewidth=2, foreground="w")],
	wrap=True,
	**kwargs,
	)
	# hack to have the text wrapped in the ax extent, for some explanation see
	# https://stackoverflow.com/questions/48079364/wrapping-text-not-working-in-matplotlib
	wrap_width = ax.get_window_extent().width * 0.99
	text_artist._get_wrap_line_width = lambda: wrap_width
	return text_artist

[harningle]

Seems wrapping here does not work with bbox_inches="tight". In contextily, we always wrap the attribution (line 307):

contextily/contextily/plotting.py

Lines 300 to 309 in 00ca031

	text_artist = ax.text(
	0.005,
	0.005,
	text,
	transform=ax.transAxes,
	size=font_size,
	path_effects=[patheffects.withStroke(linewidth=2, foreground="w")],
	wrap=True,
	**kwargs,
	)

However, wrap=True "does not work together with savefig(..., bbox_inches='tight')", as said in Matplotlib docs. A hacky solution would be disabling wrapping, if we are sure the text can comfortably sit in a single line:

ctx.add_basemap(ax=ax, ...)
txt = ax.texts[-1]
txt.set_wrap(False)