How to use Matplotlib

This guide shows you how to use Matplotlib to create plots in Deephaven.

By default, Deephaven does not come with Matplotlib, so you can either use our Deephaven+Matplotlib base repository or extend the Deephaven Dockerized set up. Both options are documented below.

Quickstart

To get up and running with Matplotlib, clone the Deephaven+Matplotlib base repo, enter its directory, and then run docker-compose up -d as usual:

git clone https://github.com/deephaven-examples/deephaven-matplotlib-base.git
cd deephaven-matplotlib-base
docker-compose pull
docker-compose up  --build -d

This starts the Deephaven IDE with all the needed packages.

Now, you're ready to use these plotting libraries. Open the IDE and get plotting!

Extend Deephaven

If instead you wish to extend the Deephaven build, start by following the tutorial to Launch Deephaven from pre-built images.

Once you've completed the steps in the tutorial, you can extend Deephaven. The Deephaven deployment typically only comes with a docker-compose.yml file. This file will need one modification to the following code block:

services:
  deephaven:
    image: ghcr.io/deephaven/server:latest

This needs to be changed to:

services:
  deephaven:
    build: ./server

Once that's done, you'll need to create a new directory called server, from which docker-compose will build the necessary Docker image and dependencies.

mkdir server
cd server

The server folder will contain a Dockerfile. The Dockerfile defines how the Deephaven server image will be extended for matplotlib support.

Here's what Dockerfile should look like:

# syntax=docker/dockerfile:1.4

FROM ghcr.io/deephaven/web-plugin-packager:latest as build
RUN ./pack-plugins.sh @deephaven/js-plugin-matplotlib

FROM ghcr.io/deephaven/server:latest
RUN pip install --no-cache-dir deephaven-plugin-matplotlib matplotlib
COPY --link --from=build js-plugins /opt/deephaven/config/js-plugins/

Everything's ready to go! The following shell commands will spin up Deephaven with matplotlib support:

cd ..
docker compose up --build

Matplotlib examples

Here is the basic usage of Matplotlib to show one figure:

import matplotlib.pyplot as plt


x = [0, 2, 4, 6]
y = [1, 3, 4, 8]
plt.figure()
plt.plot(x, y)
plt.xlabel('x values')
plt.ylabel('y values')
plt.title('plotted x and y values')
plt.legend(['line 1'])

m_figure=plt.gcf()

The full functionaly of Matplotlib is avilable inside the Deephaven IDE:

import matplotlib.pyplot as plt
import numpy as np
import math

# Get the angles from 0 to 2 pie (360 degree) in narray object
X = np.arange(0, math.pi*2, 0.05)

# Using built-in trigonometric function we can directly plot
# the given cosine wave for the given angles
Y1 = np.sin(X)
Y2 = np.cos(X)
Y3 = np.tan(X)
Y4 = np.tanh(X)

# Initialise the subplot function using number of rows and columns
figure, axis = plt.subplots(2, 2)

# For Sine Function
axis[0, 0].plot(X, Y1)
axis[0, 0].set_title("Sine Function")

# For Cosine Function
axis[0, 1].plot(X, Y2)
axis[0, 1].set_title("Cosine Function")

# For Tangent Function
axis[1, 0].plot(X, Y3)
axis[1, 0].set_title("Tangent Function")

# For Tanh Function
axis[1, 1].plot(X, Y4)
axis[1, 1].set_title("Tanh Function")

Here are some 3D examples from Rashida Nasrin Sucky. The data, available from Kaggle, needs to be placed in the data directory. For more information see our guide on Docker data volumes.

import pandas as pd
import numpy as np
from mpl_toolkits import mplot3d
import matplotlib.pyplot as plt
df = pd.read_csv("/data/auto_clean.csv")


fig = plt.figure(figsize=(10, 10))
ax = plt.axes(projection="3d")
ax.scatter3D(df['length'], df['width'], df['height'],
             c = df['peak-rpm'], s = df['price']/50, alpha = 0.4)
ax.set_xlabel("Length")
ax.set_ylabel("Width")
ax.set_zlabel("Height")
ax.set_title("Relationship between height, weight, and length")
plt.show()

df['body-style'].unique()

df['body_style1'] = df['body-style'].replace({"convertible": 1,
                                             "hatchback": 2,
                                             "sedan": 3,
                                             "wagon": 4,
                                             "hardtop": 5})

gr = df.groupby("body_style1")['peak-rpm', 'price'].agg('mean')
x = gr.index
y = gr['peak-rpm']
z = [0]*5
colors = ["b", "g", "crimson", 'r', 'pink']
dx = 0.3 * np.ones_like(z)
dy = [30]*5
dz = gr['price']
fig = plt.figure(figsize=(10, 8))
ax = fig.add_subplot(111, projection="3d")
ax.set_xticklabels(['convertible', 'hatchback', 'sedan', 'wagon', 'hardtop'])
ax.set_xlabel("Body Style", labelpad = 7)
ax.set_yticks(np.linspace(5000, 5250, 6))
ax.set_ylabel("Peak Rpm", labelpad=10)
ax.set_zlabel("Price")
ax.set_zticks(np.linspace(7000, 22250, 6))
ax.set_title("Change of Price with Body_style and Peak RPM")
ax.bar3d(x, y, z, dx, dy, dz)

def z_function(x, y):
    return np.sin(np.sqrt(x**2 + y**2))
plt.figure(figsize=(8, 8))
ax = plt.axes(projection="3d")
x = df['peak-rpm']
y = df['city-mpg']
z = z_function(x, y)
ax.plot_trisurf(x, y, z,
                cmap='viridis', edgecolor='none');
ax.set_xlabel("Peak RPM")
ax.set_ylabel("City-MPG")
ax.set_title("Peak RPM vs City-MPG")
ax.view_init(60, 25)
plt.show()

tri_surf=plt.gcf()

Matplotlib examples with TableAnimation

Deephaven's matplotlib plug-in also enables you to visualize real-time data.

To install the plug-in, run:

pip install deephaven-plugin-matplotlib

Making a live animation requires two imports:

from deephaven.plugin.matplotlib import TableAnimation
import matplotlib.pyplot as plt

The TableAnimation class will re-draw a plot every time a ticking table is updated. It needs three inputs:

1. The figure containing data (in the examples below, it's fig).
2. The table containing data (in the examples below, it's tt or tt_sorted_top).
3. The function defining how the plot is animated (update_fig).

See the examples below.

Line plot

import matplotlib.pyplot as plt
from deephaven import time_table
from deephaven.plugin.matplotlib import TableAnimation

# Create a ticking table with the sin function
tt = time_table("00:00:01").update(["x=i", "y=Math.sin(x)"])

fig = plt.figure()      # Create a new figure
ax = fig.subplots()     # Add an axes to the figure
line, = ax.plot([],[])  # Plot a line. Start with empty data, will get updated with table updates.

# Define our update function. We only look at `data` here as the data is already stored in the format we want
def update_fig(data, update):
    line.set_data([data['x'], data['y']])

    # Resize and scale the axes. Our data may have expanded and we don't want it to appear off screen.
    ax.relim()
    ax.autoscale_view(True, True, True)

# Create our animation. It will listen for updates on `tt` and call `update_fig` whenever there is an update
ani = TableAnimation(fig, tt, update_fig)

Bar plot

import matplotlib.pyplot as plt
from deephaven import time_table
from deephaven.plugin.matplotlib import TableAnimation
from deephaven import SortDirection

top_n = 5
# Create a ticking table with the linear function y = x
tt = time_table("00:00:01").update(["x=i", "y=i"])
tt_sorted = tt.sort(order_by=["y"], order=[SortDirection.DESCENDING])
# Create a table with 5 largest values for the chart
tt_sorted_top = tt_sorted.head(top_n)

fig, ax = plt.subplots()
rects = ax.bar(range(top_n), [0] * top_n)
def update_fig(data, update):
    # update heights of the columns
    for rect, h in zip(rects, data["y"]):
        rect.set_height(h)
    # update labels
    ax.set_xticklabels(data["x"])
    ax.relim()
    ax.autoscale_view(True, True, True)
bar_plot_ani = TableAnimation(fig, tt_sorted_top, update_fig)

Scatter plot

Scatter plots require data in a different format than line plots, so we need to pass in the data differently:

import matplotlib.pyplot as plt
from deephaven import time_table
from deephaven.plugin.matplotlib import TableAnimation

tt = time_table("00:00:01").update(["x=Math.random()", "y=Math.random()", "z=Math.random()*50"])

fig = plt.figure()
ax = fig.subplots()
ax.set_xlim(0, 1)
ax.set_ylim(0, 1)
scat = ax.scatter([],[])    # Provide empty data initially
scatter_offsets = []        # Store separate arrays for offsets and sizes
scatter_sizes = []

def update_fig(data, update):
    # This assumes that table is always increasing. Otherwise need to look at other
    # properties in update for creates and removed items
    added = update.added()
    for i in range(0, len(added['x'])):
        # Append new data to the sources
        scatter_offsets.append([added['x'][i], added['y'][i]])
        scatter_sizes.append(added['z'][i])

    # Update the figure
    scat.set_offsets(scatter_offsets)
    scat.set_sizes(scatter_sizes)

ani = TableAnimation(fig, tt, update_fig)

Multiple series

It's possible to have multiple kinds of series in the same figure. Here is an example of a line and a scatter plot:

import matplotlib.pyplot as plt
from deephaven import time_table
from deephaven.plugin.matplotlib import TableAnimation

tt = time_table("00:00:01").update(["x=i", "y=Math.sin(x)", "z=Math.cos(x)", "r=Math.random()", "s=Math.random()*100"])

fig = plt.figure()
ax = fig.subplots()
line1, = ax.plot([],[])
line2, = ax.plot([],[])
scat = ax.scatter([], [])
scatter_offsets = []
scatter_sizes = []

def update_fig(data, update):
    line1.set_data([data['x'], data['y']])
    line2.set_data([data['x'], data['z']])
    added = update.added()
    for i in range(0, len(added['x'])):
        scatter_offsets.append([added['x'][i], added['r'][i]])
        scatter_sizes.append(added['s'][i])
    scat.set_offsets(scatter_offsets)
    scat.set_sizes(scatter_sizes)
    ax.relim()
    ax.autoscale_view(True, True, True)

ani = TableAnimation(fig, tt, update_fig)

Related documentation

• How to use Seaborn
• How to create XY series plots
• How to create category plots
• How to create histograms
• How to create category histograms
• How to use the Chart Builder
• How to use plug-ins
• Arrays