wraplotly

wraplotly is a small wrapper around plotly built to:

• Make it easier to create subplots by using wrappers containg both the plotly express and the graph objects (sometimes ever figure factory).
• Automaticly resample scatter plots that contain too much information using the plotly-resampler library.
• Add some data analysis related functionalities to plotly.

A quick demonstration of how wraplotly handles subfigures for instance:

import wraplotly as wp
from sklearn import datasets
from sklearn.cluster import AgglomerativeClustering

n_samples = 500
points, _ = datasets.make_circles(n_samples=n_samples, factor=0.5, noise=0.05)
x, y = points[:, 0], points[:, 1]

c_single = AgglomerativeClustering(linkage="single").fit(points).labels_
c_average = AgglomerativeClustering(linkage="average").fit(points).labels_
c_complete = AgglomerativeClustering(linkage="complete").fit(points).labels_
c_ward = AgglomerativeClustering(linkage="ward").fit(points).labels_

grid = wp.grid([[0, 1, 2, 3]], subplot_titles=["Single", "Average", "Complete", "Ward"])

grid(wp.scatter(x, y, color=c_single))
grid(wp.scatter(x, y, color=c_average))
grid(wp.scatter(x, y, color=c_complete))
grid(wp.scatter(x, y, color=c_ward))

fig = grid.fig
fig = fig.update_layout(title="Different linkage impacting the agglomerative clustering of a circle")
fig.show()

There is also a built-in function to stack figures horizontaly:

hstack = wp.hstack(
    wp.scatter(x, y, color=c_single),
    wp.scatter(x, y, color=c_average),
    wp.scatter(x, y, color=c_complete),
    wp.scatter(x, y, color=c_ward),
    subplot_titles=["Single", "Average", "Complete", "Ward"],
)

fig = hstack.fig
fig = fig.update_layout(title="Different linkage impacting the agglomerative clustering of a circle")
fig.show()

Installation

pip3 install git+https://github.com/goncalogiga/wraplotly

Other arragements

Aside from the grid and the hstack methods, it is also possible to use vertical stacking (vstack) and combination of plots (combine)

Vertical stacking

oil = pd.read_csv("/kaggle/input/store-sales-time-series-forecasting/oil.csv")
transactions = pd.read_csv("/kaggle/input/store-sales-time-series-forecasting/transactions.csv")

oil['date'] = pd.to_datetime(oil['date'], format = "%Y-%m-%d")
transactions['date'] = pd.to_datetime(transactions['date'], format = "%Y-%m-%d")

stack = wp.vstack(
    wp.line(transactions, "date", "transactions", name="transactions"),
    wp.line(oil, x="date", y="dcoilwtico", name="oil price"),
    subplot_titles=["Number of transactions per day", "Daily oil price"]
)

fig = stack.fig
fig = fig.update_layout(height=750)
fig.show()

Combining

oil = pd.read_csv("/kaggle/input/store-sales-time-series-forecasting/oil.csv")
transactions = pd.read_csv("/kaggle/input/store-sales-time-series-forecasting/transactions.csv")

oil['date'] = pd.to_datetime(oil['date'], format = "%Y-%m-%d")
transactions['date'] = pd.to_datetime(transactions['date'], format = "%Y-%m-%d")

transactions["transactions"] = 1 + transactions["transactions"] / transactions["transactions"].max() * 100

wp.combine(
    wp.line(transactions, "date", "transactions", name="transactions"),
    wp.line(oil, x="date", y="dcoilwtico", name="oil price"),
    subplot_titles=["Number of transactions per day & Daily oil prices"]
)

Full list of wraplotly wrapper functions

Definition	Supports arragements (grid, hstack ect.)	Wraps outside arragement	Wraps inside arragement
line	Yes	px.line	go.Scatter
scatter	Yes	px.scatter	go.Scatter
bar	Yes	px.bar	go.Bar
box	Yes	px.box	go.Box
histogram	Yes	px.histogram	go.Histogram
imshow	Yes	px.imshow	go.Image
density_heatmap	Yes	px.density_heatmap	go.Histogram2d
heatmap	No	px.imshow
distplot	No	ff.create_distplot
pairplot	No	ff.create_scatterplotmatrix
colored_line	No	px.scatter

About plotly-resampler

wraplotly uses the plotly-resampler library, which solves memory issues when displaying large plots, by downsampling (aggregating) the data respective to the view and then plotting the aggregated points. This is done automaticly when the data passed to the wraplotly objects is too large. Here is an example:

x = np.arange(1_000_000)
noisy_sin = (3 + np.sin(x / 200) + np.random.randn(len(x)) / 10) * x / 1_000

wp.line(noisy_sin)

/home/wraplotly/base.py:85: UserWarning:

Data was too large (~1000000) and had to be downsampled using plotly-resampler.