matplotboard 1.3.0

Generate simple HTML dashboards using matplotlib

A utility to generate html dashboards using matplotlib. Matplotboard makes it easy to wrap your plotting functions and dump the plots into a searchable webpage or a markdown report. This is best demonstrated with an example.

import numpy as np
import matplotlib.pyplot as plt
import matplotboard as mpb

@mpb.decl_fig
def cool_fig():
    xs = np.linspace(-10, 10, 100)
    ys = xs**2
    plt.plot(xs, ys)

if __name__ == '__main__':
    figures = {
        'cool_fig': cool_fig(),
    }

    mpb.render(figures)
    mpb.generate_report(figures, 'Report')

You can view the results here. Let's walk through this one part at a time.

First, we import numpy and matplotlib for some calculations, and plotting, respectively. As well as matplotboard itself.

import numpy as np
import matplotlib.pyplot as plt
from src import matplotboard as mpb

matplotboard relies upon matplotlib for the underlying rendering engine so other plotting libraries are not supported. However, wrappers around matplotlib such as seaborn should work.

Next, we declare the function that is actually going to do the plotting.

@mpb.decl_fig
def cool_fig():
    xs = np.linspace(-10,10, 100)
    ys = xs**2
    plt.plot(xs, ys)

The decl_fig decorator modifies the function to work with matplotboard. A plotting function decorated with decl_fig must fulfill the following contract:

• A clean figure has been initiated and the function will do any plotting on that figure.
• It is free to subdivide the figure into as many axes as required, but shouldn't create additional Figure objects.
• The function can optionally return Markdown text that will be rendered along with the plot.
• The function shouldn't call savefig. This is handled by matplotboard automatically.

Finally, we declare the actual figures that we want to generate, and tell matplotboard to render the figures and assemble them into an interactive webpage.

if __name__ == '__main__':
    figures = {
        'cool_fig': cool_fig(),
    }

    mpb.render(figures)
    mpb.generate_report(figures, 'Report')

Both render, and generate_report take a dictionary as their first argument. The dictionary keys are strings that are interpreted as the individual figure names, and the dictionary values are the plots we want to generate. Note that the function is called before inserting it into the dictionary. Due to the modification of the original function by the decorator, this doesn't actually call the function yet, but bundles the function and any arguments together into a Figure object which it then returns for later processing by matplotboard.

By writing plotting functions with arguments, a single function can be reused to make many different plots. For example, you may have a dataset that is divided into several categories and you would like to plot some variable for each category. You could do this by writing one plotting function and calling it with different arguments to specify each of the categories.

Try running the example. If everything works, there should be a new folder in the current directory called dashboard, and within it an html file called report.html. Open it with your browser to see a dashboard containing a single plot. Try clicking on it for a zoomed view!

A single plot is not very interesting. Where matplotboard starts to really become useful is when you have lots of plots to generate. Check out the following example.

from itertools import product
import numpy as np
import matplotlib.pyplot as plt
import matplotboard as mpb

@mpb.decl_fig
def cool_fig(func, scale, color='b'):
    xs = np.linspace(-scale, scale, 100)
    f = {
        'sin': lambda xs: np.sin(xs),
        'tan': lambda xs: np.tan(xs),
        'exp': lambda xs: np.exp(xs),
    }[func]
    ys = f(xs)
    plt.plot(xs, ys, color=color)

if __name__ == '__main__':
    mpb.configure(multiprocess=True)
    figures = {}

    for color, function, scale in product('rbgk', ['sin', 'tan', 'exp'], np.linspace(1, 20, 20)):
        figures[f'{function}_{color}_{scale}'] = cool_fig(function, scale, color=color)


    mpb.render(figures)
    mpb.generate_report(figures, 'Report')

What's changed? You can view the page here

First of all, the plotting function has been enhanced to take a few arguments that modify it's behavior. You can now specify whether you would like to plot sin, tan, or exp as well as effectively set the x length scale.

Second, we now are programatically making all combinations of plotting color, function, and scale with the product function and declaring a plot for each combination. This comes down to 4*3*20=240 different plots. To speed things up a bit, this example also switches on matplotboard's multiprocessing support. Try running this example and open up the resulting web-page just as before. Note the pagination feature limiting the number of figures displayed at once. Also, try selecting a plot and moving through the figures on the page with the arrow keys. Finally, try out the filter box in the top right. A few interesting searches may be "sin_", "_r_", or "tan_g_9" to search for all sin plots, all red plots, and just the tan_g_9 plot, respectively.

For one final example, let's look at the support for writing reports the incorporate generated figures.

from itertools import product
import numpy as np
import matplotlib.pyplot as plt
import matplotboard as mpb


@mpb.decl_fig
def cool_fig(func, scale, color="b"):
    xs = np.linspace(-scale, scale, 100)
    f = {
        "sin": lambda xs: np.sin(xs),
        "tan": lambda xs: np.tan(xs),
        "exp": lambda xs: np.exp(xs),
    }[func]
    ys = f(xs)
    plt.plot(xs, ys, color=color)


report = """\
Authors: Will Hunting
Date: December 2, 1997

# Report On Functions

## Introduction

As we all know, there are many functions. An example is the sine function seen below.
fig::sin_b_1

## Other Functions

However, there are many other functions such as the tangent or exponential.

<div class="row">
<div class="col-md-6 row_fig">
fig::tan_r_1|The rugged tangent function
</div>
<div class="col-md-6 row_fig">
fig::exp_g_2|The majestic exponential function
</div>
</div>

The decision of which function is best is up to *you*!

## Local Figures

I happened to have a couple *really* fantastic figures on my computer that I
want to include as well. How do I include them? It's easy! Just add them to
the list of figures with the `loc_fig` function and they will be marked to be
copied to the output directory. Here are a couple examples:

<div class="row">
<div class="col-md-6 row_fig">
fig::image8
</div>
<div class="col-md-6 row_fig">
fig::image10
</div>
</div>
"""

if __name__ == "__main__":
    mpb.configure(multiprocess=True)
    figures = {}

    for color, function, scale in product(
        "rbgk", ["sin", "tan", "exp"], np.linspace(1, 5, 5)
    ):
        figures[f"{function}_{color}_{int(scale)}"] = cool_fig(
            function, scale, color=color
        )
    figures["image8"] = mpb.loc_fig("figures/image8.png")
    figures["image10"] = mpb.loc_fig("figures/image10.png")

    mpb.render(figures)
    mpb.generate_report(figures, "Report", body=report)

See result of this example here.

The generate_report function supports an optional body argument which signals matplotboard to render the markdown into a report, rather than making a simple plot dump. A special syntax is used for embedding generated figures.

fig::figure_name|Optional Caption

Bootstrap is included by default so multiple figures side-by-side are possible by use of a row div as shown in the example.

In addition to including generated figures via the fig:: construct, static figures (such as diagrams or photographs) can be included via the locfig:: (think local figure) construct, where instead of the figure name, you specify the path to the file. Finally, pictures out on the internet can be specified via extfig::.