Question about working other plotting packages like seaborn.

[gepcel]

Many other packages provide numerous supplements for Matplotlib. Occasionally, I need a feature that Ultraplot lacks. For instance, I miss useful statistical methods and finely tuned color themes of seaborn. I learned about ax.external(), so I gave it a try. Let's look at an example:

# %% import seaborn and prepare data
import seaborn as sns
x = np.random.randn(1000)+3
df = pd.DataFrame(x, columns=["Test Values"])

# %% plot with only seaborn and move legend
# legend was moved successfully
ax = sns.histplot(df, kde=True, legend=True)
sns.move_legend(ax, "upper left")

# %% plot with ultraplot and seaborn
# legend cannot be moved.
fig, ax = pplt.subplot()
with ax.external():
 sns.histplot(df, ax=ax, kde=True, legend=True)
 sns.move_legend(ax, "upper right")

I understand that working with other packages isn't really Ultraplot's agenda, and cooperation isn't always straightforward. I was also surprised to find that I can't control the legend position in Seaborn's histplot with some kind of loc argument—perhaps I missed something. Are there other ways to integrate Seaborn with Ultraplot that I'm unaware of?

Maybe we could consider adding some useful statistical fitting methods directly into Ultraplot?

[cvanelteren]

One thing you could do in this particular example is use ax.legend instead

fig, ax = uplt.subplot()
with ax.external():
 sns.histplot(df, ax=ax, kde=True, legend=False)

# sns.move_legend(ax, "upper right")
ax.legend(loc = "ll")

While I generally cannot guarantee compatibility across all packages, I am sympathetic to just making it work. I think for this in particular, we can make it work by adding this method to our legend.