Styling Matplotlib figures¶

Arcadia has a set of visual communication standards for figures. The arcadia_pycolor package provides utilities for automatically styling figures in Python to meet these standards.

Note: While equivalent styling utilities exist for Plotly figures, this user guide was written specifically for working with Matplotlib figures. There is currently an open issue for writing an equivalent user guide (or updating this one) for Plotly users.

In [1]:

import pathlib

import matplotlib as mpl
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
from sklearn import datasets

import arcadia_pycolor as apc

import pathlib import matplotlib as mpl import matplotlib.pyplot as plt import pandas as pd import seaborn as sns from sklearn import datasets import arcadia_pycolor as apc

Load datasets¶

First, load some datasets to use for example plots.

In [2]:

iris = datasets.load_iris()
iris_data = pd.DataFrame(iris.data)  # type: ignore
iris_data.columns = [i.replace(" (cm)", "") for i in iris.feature_names]  # type: ignore
iris_data["species"] = iris.target_names[iris.target]  # type: ignore

x_data = "sepal length"
y_data = "sepal width"
c_data = "species"

glue_dataset = sns.load_dataset("glue").pivot(index="Model", columns="Task", values="Score")

iris = datasets.load_iris() iris_data = pd.DataFrame(iris.data) # type: ignore iris_data.columns = [i.replace(" (cm)", "") for i in iris.feature_names] # type: ignore iris_data["species"] = iris.target_names[iris.target] # type: ignore x_data = "sepal length" y_data = "sepal width" c_data = "species" glue_dataset = sns.load_dataset("glue").pivot(index="Model", columns="Task", values="Score")

Example plots without Arcadia's styles¶

Using plotting tools like seaborn, you can create plots with their default styles. These styles don't conform to Arcadia's style standards.

In [3]:

sns.scatterplot(x=x_data, y=y_data, hue="species", data=iris_data)
plt.title("An example scatterplot")
plt.show()

sns.violinplot(x=c_data, y=y_data, data=iris_data, hue=c_data)
plt.title("An example violin plot")
plt.show()

sns.heatmap(glue_dataset)
plt.title("An example heatmap")
plt.show()

sns.scatterplot(x=x_data, y=y_data, hue="species", data=iris_data) plt.title("An example scatterplot") plt.show() sns.violinplot(x=c_data, y=y_data, data=iris_data, hue=c_data) plt.title("An example violin plot") plt.show() sns.heatmap(glue_dataset) plt.title("An example heatmap") plt.show()

Applying Arcadia's styles¶

Using arcadia_pycolor's apc.mpl.setup() function, you can update matplotlib's default style settings to conform to Arcadia's standards.

For example, the above plots don't use the Suisse family of fonts, which are the official fonts specified by Arcadia's style guide.

If you have Suisse family fonts installed, apc.mpl.setup() will use them. If you don't have them installed, apc.mpl.setup() will still work for other style settings, but the fonts will be the default ones.

In [4]:

apc.mpl.setup()

colors = {
    "setosa": apc.aegean,
    "versicolor": apc.amber,
    "virginica": apc.seaweed,
}

fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_square"), layout="constrained")
sns.scatterplot(
    x=x_data,
    y=y_data,
    hue=c_data,
    data=iris_data,
    palette=colors,
    ax=ax,
)
plt.title("An example scatterplot")
plt.show()

fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_square"), layout="constrained")
sns.violinplot(
    x=c_data,
    y=y_data,
    data=iris_data,
    hue=c_data,
    palette=colors,
    ax=ax,
)
plt.title("An example violin plot")
plt.show()

fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_wide"), layout="constrained")
sns.heatmap(glue_dataset, ax=ax, cmap=apc.gradients.reds.to_mpl_cmap())
plt.title("An example heatmap")
plt.show()

apc.mpl.setup() colors = { "setosa": apc.aegean, "versicolor": apc.amber, "virginica": apc.seaweed, } fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_square"), layout="constrained") sns.scatterplot( x=x_data, y=y_data, hue=c_data, data=iris_data, palette=colors, ax=ax, ) plt.title("An example scatterplot") plt.show() fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_square"), layout="constrained") sns.violinplot( x=c_data, y=y_data, data=iris_data, hue=c_data, palette=colors, ax=ax, ) plt.title("An example violin plot") plt.show() fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_wide"), layout="constrained") sns.heatmap(glue_dataset, ax=ax, cmap=apc.gradients.reds.to_mpl_cmap()) plt.title("An example heatmap") plt.show()

Automatic registration of Arcadia's colors and gradients in Matplotlib¶

After apc.mpl.setup() is called, all of Arcadia's named colors and colormaps are automatically registered in Matplotlib. This allows you to use them in your plots without having to directly pass the colors or gradients.

These registered colors and colormaps are prefixed with "apc:". For example, "apc:sky" is a registered color, and "apc:blues" is a registered colormap. You can reverse gradients by appending "_r" to the colormap name, e.g. "apc:blues_r".

In [5]:

registered_colors = {
    "setosa": "apc:sky",
    "versicolor": "apc:vital",
    "virginica": "apc:lapis",
}

fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_square"), layout="constrained")
sns.scatterplot(
    x=x_data,
    y=y_data,
    hue=c_data,
    data=iris_data,
    palette=registered_colors,
    ax=ax,
)
plt.title("An example scatterplot")
plt.show()

fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_wide"), layout="constrained")
sns.heatmap(glue_dataset, ax=ax, cmap="apc:blues")
plt.title("An example heatmap")
plt.show()

fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_wide"), layout="constrained")
sns.heatmap(glue_dataset, ax=ax, cmap="apc:blues_r")
plt.title("An example heatmap with reversed gradient")
plt.show()

registered_colors = { "setosa": "apc:sky", "versicolor": "apc:vital", "virginica": "apc:lapis", } fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_square"), layout="constrained") sns.scatterplot( x=x_data, y=y_data, hue=c_data, data=iris_data, palette=registered_colors, ax=ax, ) plt.title("An example scatterplot") plt.show() fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_wide"), layout="constrained") sns.heatmap(glue_dataset, ax=ax, cmap="apc:blues") plt.title("An example heatmap") plt.show() fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_wide"), layout="constrained") sns.heatmap(glue_dataset, ax=ax, cmap="apc:blues_r") plt.title("An example heatmap with reversed gradient") plt.show()

Generating publication-ready plots¶

You can use apc.mpl.style_plot to further style plots to conform to Arcadia's standards. You can call this function after creating a plot to apply the following changes:

• Capitalize legend and axis titles.
• Add a horizontal line between legend title and entries.
• Converts axis labels to monospaced font (when monospaced_axes=True).
• Capitalize text, removes tickmarks, and adjusts padding (when categorical_axes=True).
• Convert colorbar tick labels to monospaced font (when colorbar_exists=True).

In [6]:

fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_square"), layout="constrained")
sns.scatterplot(
    x=x_data,
    y=y_data,
    hue="species",
    data=iris_data,
    palette=colors,
    ax=ax,
)
apc.mpl.style_plot(monospaced_axes="all")
plt.title("An example scatterplot")
plt.show()

fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_square"), layout="constrained")
sns.violinplot(
    x=c_data,
    y=y_data,
    data=iris_data,
    hue=c_data,
    palette=colors,
    ax=ax,
)
apc.mpl.style_plot(categorical_axes="x", monospaced_axes="y")
plt.title("An example violinplot")
plt.show()

fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_wide"), layout="constrained")
sns.heatmap(glue_dataset, ax=ax, cmap="apc:reds")
apc.mpl.style_plot(colorbar_exists=True)
apc.mpl.set_xaxis_categorical()
apc.mpl.set_yaxis_categorical()
plt.title("An example heatmap")
plt.show()

fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_square"), layout="constrained") sns.scatterplot( x=x_data, y=y_data, hue="species", data=iris_data, palette=colors, ax=ax, ) apc.mpl.style_plot(monospaced_axes="all") plt.title("An example scatterplot") plt.show() fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_square"), layout="constrained") sns.violinplot( x=c_data, y=y_data, data=iris_data, hue=c_data, palette=colors, ax=ax, ) apc.mpl.style_plot(categorical_axes="x", monospaced_axes="y") plt.title("An example violinplot") plt.show() fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions("float_wide"), layout="constrained") sns.heatmap(glue_dataset, ax=ax, cmap="apc:reds") apc.mpl.style_plot(colorbar_exists=True) apc.mpl.set_xaxis_categorical() apc.mpl.set_yaxis_categorical() plt.title("An example heatmap") plt.show()

Using default figure sizes¶

The apc.mpl.get_figure_dimensions() function allows you to choose from a fixed set of plot sizes. These are compatible with our Adobe Creative Cloud templates.

When generated with layout="constrained" and saved using apc.mpl.save_figure(), the resulting PDF images should drag and drop neatly into place with the Illustrator templates.

There are three size categories, based on the Creative Cloud templates:

• "full" (1000px wide)
• "float" (700px wide)
• "half" (500px wide)

Each size category has a few options for aspect ratio:

• "wide": wider on the horizontal axis, for "full" and "float" sizes
• "square": equal width and height, for all sizes

In [7]:

for plot_size in ("full_wide", "full_square", "float_wide", "float_square", "half_square"):
    fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions(plot_size), layout="constrained")
    sns.scatterplot(x=x_data, y=y_data, hue="species", data=iris_data, ax=ax)
    apc.mpl.style_plot(monospaced_axes="all")
    plt.title(f'"{plot_size}" figure size')
    plt.show()

for plot_size in ("full_wide", "full_square", "float_wide", "float_square", "half_square"): fig, ax = plt.subplots(figsize=apc.mpl.get_figure_dimensions(plot_size), layout="constrained") sns.scatterplot(x=x_data, y=y_data, hue="species", data=iris_data, ax=ax) apc.mpl.style_plot(monospaced_axes="all") plt.title(f'"{plot_size}" figure size') plt.show()

Simulating colored backgrounds¶

The default Creative Cloud templates use an apc.parchment background color. You can simulate this background color in your plots by modifying matplotlib's rcParams within a with statement. When saved using apc.mpl.save_figure(), PDF files default to using a transparent background.

In [8]:

with mpl.rc_context({"figure.facecolor": apc.parchment}):
    sns.scatterplot(x=x_data, y=y_data, hue="species", data=iris_data)
    apc.mpl.style_plot(monospaced_axes="all")
    plt.show()

with mpl.rc_context({"figure.facecolor": apc.parchment}): sns.scatterplot(x=x_data, y=y_data, hue="species", data=iris_data) apc.mpl.style_plot(monospaced_axes="all") plt.show()

Saving figures¶

The arcadia_pycolor package provides a convenience function apc.mpl.save_figure to save figures with the correct resolutions (DPI) for web and print publication.

In [9]:

output_dirpath = pathlib.Path("output")
output_dirpath.mkdir(exist_ok=True)

sns.scatterplot(x=x_data, y=y_data, hue="species", data=iris_data)
apc.mpl.style_plot(monospaced_axes="all")

# Save a figure with the resolution for websites.
apc.mpl.save_figure(
    str(output_dirpath / "example-figure-web.pdf"), size="full_square", context="web"
)

# Save a figure with the resolution for print.
apc.mpl.save_figure(
    str(output_dirpath / "example-figure-print.pdf"), size="full_square", context="print"
)

output_dirpath = pathlib.Path("output") output_dirpath.mkdir(exist_ok=True) sns.scatterplot(x=x_data, y=y_data, hue="species", data=iris_data) apc.mpl.style_plot(monospaced_axes="all") # Save a figure with the resolution for websites. apc.mpl.save_figure( str(output_dirpath / "example-figure-web.pdf"), size="full_square", context="web" ) # Save a figure with the resolution for print. apc.mpl.save_figure( str(output_dirpath / "example-figure-print.pdf"), size="full_square", context="print" )

'created' timestamp seems very low; regarding as unix timestamp

'created' timestamp seems very low; regarding as unix timestamp

You can also generate figures with multiple different file types. For example, you can save a figure as both a PNG and a PDF file.

In [10]:

apc.mpl.save_figure(
    str(output_dirpath / "example-figure"),
    size="full_square",
    filetypes=["pdf", "png"],
    context="web",
)

apc.mpl.save_figure( str(output_dirpath / "example-figure"), size="full_square", filetypes=["pdf", "png"], context="web", )

Invalid filetype ''. Skipping.

<Figure size 460.8x345.6 with 0 Axes>