Overlapping labels in relplot with seaborn 0.12

[sorenwacker]

import seaborn as sns

sns.set_context("paper")
sns.set_style("white")

data = (sns.load_dataset('iris').set_index('species')*1e7).reset_index()
g = sns.relplot(data=data, x='sepal_length', y='sepal_width', col='species', 
                col_wrap=2, height=2.5)
g.set_titles(row_template="{row_name}", col_template="SOMEWHATLONG-{col_name}")
for axes in g.axes.flat:
    axes.ticklabel_format(axis='both', style='scientific', scilimits=(0, 0))

import seaborn as sns

sns.set_context("paper")
sns.set_style("white")

data = (sns.load_dataset('iris').set_index('species')*1e7).reset_index()
g = sns.relplot(data=data, x='sepal_length', y='sepal_width', col='species', 
                col_wrap=2, height=2.5, facet_kws=dict(sharex=False, sharey=False))
g.set_titles(row_template="{row_name}", col_template="SOMEWHATLONG-{col_name}")
for axes in g.axes.flat:
    axes.ticklabel_format(axis='both', style='scientific', scilimits=(0, 0))

seaborn 11.2:

seaborn 12.1:

y

[sorenwacker]

Is there a way to turn the xlabels off again or to increase the spacing between the xaxis labels and the figure titles?

[mwaskom]

Please turn this into a reproducible example, thanks.

[sorenwacker]

Here is an reproducible example.

[sorenwacker]

The columtemplate has a \n at the end to prevent the longer title to overlap with the scaling indicator (for the lack of a better name).

data = (sns.load_dataset('iris').set_index('species')*1e7).reset_index()
g = sns.relplot(data=data, x='sepal_length', y='sepal_width', col='species', 
                col_wrap=2, height=2.5, facet_kws=dict(sharex=False, sharey=False))
g.set_titles(row_template="{row_name}", col_template="SOMEWHATLONG-{col_name}")
for axes in g.axes.flat:
    axes.ticklabel_format(axis='both', style='scientific', scilimits=(0, 0))

Also seaborn 0.12.1.

[mwaskom]

Thanks, see #3180

[sorenwacker]

Is there a way to remove the x-axis labels for plots which are not on the bottom? In my example, only the upper left plot? So that it mimics the 11.2 behaviour. As a workaround.

[sorenwacker]

I think, I will just use 11.2 until 12.2 is out.
Thanks for your great work!

[mwaskom]

Yes, you can easily clean up the figure using the matplotlib api