User Guide

Installation

To install use pip:

# with imgui and jupyterlab
pip install -U "fastplotlib[notebook,imgui]"

# minimal install, install glfw, pyqt6 or pyside6 separately
pip install -U fastplotlib

# with imgui
pip install -U "fastplotlib[imgui]"

# to use in jupyterlab, no imgui
pip install -U "fastplotlib[notebook]"

We strongly recommend installing simplejpeg for use in notebooks, you must first install libjpeg-turbo.

• If you use conda, you can get libjpeg-turbo through conda.

• If you are on linux you can get it through your distro’s package manager.

• For Windows and Mac compiled binaries are available on their release page: libjpeg-turbo/libjpeg-turbo

Once you have libjpeg-turbo:

pip install simplejpeg

What is fastplotlib?

fastplotlib is a cutting-edge plotting library built using the pygfx rendering engine. The lower-level details of the rendering process (i.e. defining a scene, camera, renderer, etc.) are abstracted away, allowing users to focus on their data. The fundamental goal of fastplotlib is to provide a high-level, expressive API that promotes large-scale explorative scientific visualization. We want to make it easy and intuitive to produce interactive visualizations that are as performant and vibrant as a modern video game 😄

How to use fastplotlib

Before giving a detailed overview of the library, here is a minimal example:

import fastplotlib as fpl
import imageio.v3 as iio

# create a `Figure`
fig = fpl.Figure()

# read data
data = iio.imread("imageio:astronaut.png")

# add image graphic
image_graphic = fig[0, 0].add_image(data=data)

# show the plot
fig.show()

if __name__ == "__main__":
    fpl.run()

This is just a simple example of how the fastplotlib API works to create a plot, add some image data to the plot, and then visualize it. However, we are just scratching the surface of what is possible with fastplotlib. Next, let’s take a look at the building blocks of fastplotlib and how they can be used to create more complex visualizations.

Figure

The starting point for creating any visualization in fastplotlib is a Figure object. This can be a single plot or a grid of subplots. The Figure object houses and takes care of the underlying rendering components such as the camera, controller, renderer, and canvas. Most users won’t need to use these directly; however, the ability to directly interact with the rendering engine is still available if needed.

By default, if no shape argument is provided when creating a Figure, there will be a single subplot. All subplots in a Figure can be accessed using indexing (i.e. fig_object[i ,j]).

After defining a Figure, we can begin to add Graphic objects.

Graphics

A Graphic can be an image, a line, a scatter, a collection of lines, and more. All graphics can also be given a convenient name. This allows graphics to be easily accessed from figures:

# create a `Figure`
fig = fpl.Figure()

# read data
data = iio.imread("imageio:astronaut.png")

add image graphic
image_graphic = fig[0, 0].add_image(data=data, name="astronaut")

# show plot
fig.show()

# index plot to get graphic
fig[0, 0]["astronaut"]

See the examples gallery for examples on how to create and interactive with all the various types of graphics.

Graphics also have mutable properties that can be linked to events. Some of these properties, such as the data or colors of a line can even be indexed, allowing for the creation of very powerful visualizations.

1. Common properties that all graphics have

Feature Name	Description
name	Graphic name
offset	Offset position of the graphic, [x, y, z]
rotation	Graphic rotation quaternion
visible	Access or change the visibility
deleted	Used when a graphic is deleted, triggers events that can be useful to indicate this graphic has been deleted

2. Graphic-Specific properties

   1. ImageGraphic

   Feature Name	Description
   data	Underlying image data
   vmin	Lower contrast limit of an image
   vmax	Upper contrast limit of an image
   cmap	Colormap of an image

   2. LineGraphic, LineCollection, LineStack

   Feature Name	Description
   data	underlying data of the line(s)
   colors	colors of the line(s)
   cmap	colormap of the line(s)
   thickness	thickness of the line(s)

   3. ScatterGraphic

   Feature Name	Description
   data	underlying data of the scatter points
   colors	colors of the scatter points
   cmap	colormap of the scatter points
   sizes	size of the scatter points

   4. TextGraphic

   Feature Name	Description
   text	data of the text
   font_size	size of the text
   face_color	color of the text face
   outline_color	color of the text outline
   outline_thickness	thickness of the text

Using our example from above: once we add a Graphic to the figure, we can then begin to change its properties.

image_graphic.vmax = 150

Graphic properties also support numpy-like slicing for getting and setting data. For example

# basic numpy-like slicing, set the top right corner
image_graphic.data[:150, -150:] = 0

Fancy indexing is also supported!

bool_array = np.random.choice([True, False], size=(512, 512), p=[0.1, 0.9])
image_graphic.data[bool_array] = 254

Selectors

A primary feature of fastplotlib is the ability to easily interact with your data. Two extremely helpful tools that can be used in order to facilitate this process are a LinearSelector and LinearRegionSelector.

A LinearSelector is a horizontal or vertical line slider. This tool allows you to very easily select different points in your data. Let’s look at an example:

import fastplotlib as fpl
import numpy as np

# generate data
xs = np.linspace(-10, 10, 100)
ys = np.sin(xs)
sine = np.column_stack([xs, ys])

fig = fpl.Figure()

sine_graphic = fig[0, 0].add_line(data=sine, colors="w")

# add a linear selector the sine wave
selector = sine_graphic.add_linear_selector()

fig.show(maintain_aspect=False)

A LinearRegionSelector is very similar to a LinearSelector but as opposed to selecting a singular point of your data, you are able to select an entire region.

See the examples gallery for more in-depth examples with selector tools.

Now we have the basics of creating a Figure, adding Graphics to a Figure, and working with Graphic properties to dynamically change or alter them. Let’s take a look at how we can define events to link Graphics and their properties together.

Events

Events can be a multitude of things: traditional events such as mouse or keyboard events, or events related to Graphic properties.

There are two ways to add events in fastplotlib.

1. Use the method add_event_handler()

   def event_handler(ev):
       pass
   
   graphic.add_event_handler(event_handler, "event_type")
   

2. or a decorator

   @graphic.add_event_handler("event_type")
   def event_handler(ev):
       pass
   

The event_handler is a user-defined function that accepts an event instance as the first and only positional argument. Information about the structure of event instances are described below. The "event_type" is a string that identifies the type of event; this can be either a pygfx.Event or a Graphic property event.

graphic.supported_events will return a tuple of all event_type strings that this graphic supports.

When an event occurs, the user-defined event handler will receive an event object. Depending on the type of event, the event object will have relevant information that can be used in the callback. See below for event tables.

Graphic property events

All Graphic events have the following attributes:

attribute	type	description
type	str	“colors” - name of the event
graphic	Graphic	graphic instance that the event is from
info	dict	event info dictionary
target	WorldObject	pygfx rendering engine object for the graphic
time_stamp	float	time when the event occurred, in ms

The info attribute will house additional information for different Graphic property events:

event_type: “colors”

Vertex Colors

info dict

dict key	value type	value description
key	int | slice | np.ndarray[int | bool] | tuple[slice, …]	key at which colors were indexed/sliced
value	np.ndarray	new color values for points that were changed, shape is [n_points_changed, RGBA]
user_value	str | np.ndarray | tuple[float] | list[float] | list[str]	user input value that was parsed into the RGBA array

Uniform Colors

info dict

dict key	value type	value description
value	np.ndarray	new color values for points that were changed, shape is [n_points_changed, RGBA]

event_type: “sizes”

info dict

dict key	value type	value description
key	int | slice | np.ndarray[int | bool] | tuple[slice, …]	key at which vertex positions data were indexed/sliced
value	np.ndarray | float | list[float]	new data values for points that were changed, shape depends on the indices that were set

event_type: “data”

info dict

dict key	value type	value description
key	int | slice | np.ndarray[int | bool] | tuple[slice, …]	key at which vertex positions data were indexed/sliced
value	np.ndarray | float | list[float]	new data values for points that were changed, shape depends on the indices that were set

event_type: “thickness”

info dict

dict key	value type	value description
value	float	new thickness value

event_type: “cmap”

info dict

dict key	value type	value description
value	string	new colormap value

event_type: “selection”

LinearSelector

additional event attributes:

attribute	type	description
get_selected_index	callable	returns indices under the selector

info dict:

dict key	value type	value description
value	np.ndarray	new x or y value of selection

LinearRegionSelector

additional event attributes:

attribute	type	description
get_selected_indices	callable	returns indices under the selector
get_selected_data	callable	returns data under the selector

info dict:

dict key	value type	value description
value	np.ndarray	new [min, max] of selection

Rendering engine events from a Graphic

Rendering engine event handlers can be added to a graphic or to a Figure (see next section). Here is a description of all rendering engine events and their attributes.

• pointer_down: emitted when the user interacts with mouse, touch or other pointer devices, by pressing it down.

  • x: horizontal position of the pointer within the widget.

  • y: vertical position of the pointer within the widget.

  • button: the button to which this event applies. See “Mouse buttons” section below for details.

  • buttons: a tuple of buttons being pressed down.

  • modifiers: a tuple of modifier keys being pressed down. See section below for details.

  • ntouches: the number of simultaneous pointers being down.

  • touches: a dict with int keys (pointer id’s), and values that are dicts that contain “x”, “y”, and “pressure”.

  • time_stamp: a timestamp in seconds.

• pointer_up: emitted when the user releases a pointer. This event has the same keys as the pointer down event.

• pointer_move: emitted when the user moves a pointer. This event has the same keys as the pointer down event. This event is throttled.

• double_click: emitted on a double-click. This event looks like a pointer event, but without the touches.

• wheel: emitted when the mouse-wheel is used (scrolling), or when scrolling/pinching on the touchpad/touchscreen.

  Similar to the JS wheel event, the values of the deltas depend on the platform and whether the mouse-wheel, trackpad or a touch-gesture is used. Also, scrolling can be linear or have inertia. As a rule of thumb, one “wheel action” results in a cumulative dy of around 100. Positive values of dy are associated with scrolling down and zooming out. Positive values of dx are associated with scrolling to the right. (A note for Qt users: the sign of the deltas is (usually) reversed compared to the QWheelEvent.)

  On MacOS, using the mouse-wheel while holding shift results in horizontal scrolling. In applications where the scroll dimension does not matter, it is therefore recommended to use delta = event[‘dy’] or event[‘dx’].

  • dx: the horizontal scroll delta (positive means scroll right).

  • dy: the vertical scroll delta (positive means scroll down or zoom out).

  • x: the mouse horizontal position during the scroll.

  • y: the mouse vertical position during the scroll.

  • buttons: a tuple of buttons being pressed down.

  • modifiers: a tuple of modifier keys being pressed down.

  • time_stamp: a timestamp in seconds.

• key_down: emitted when a key is pressed down.

  • key: the key being pressed as a string. See section below for details.

  • modifiers: a tuple of modifier keys being pressed down.

  • time_stamp: a timestamp in seconds.

• key_up: emitted when a key is released. This event has the same keys as the key down event.

Time stamps

Since the time origin of time_stamp values is undefined, time stamp values only make sense in relation to other time stamps.

Mouse buttons

• 0: No button.

• 1: Left button.

• 2: Right button.

• 3: Middle button

• 4-9: etc.

Keys

The key names follow the browser spec.

• Keys that represent a character are simply denoted as such. For these the case matters: “a”, “A”, “z”, “Z” “3”, “7”, “&”, “ “ (space), etc.

• The modifier keys are: “Shift”, “Control”, “Alt”, “Meta”.

• Some example keys that do not represent a character: “ArrowDown”, “ArrowUp”, “ArrowLeft”, “ArrowRight”, “F1”, “Backspace”, etc.

Add renderer event handlers to a Figure

You can add event handlers to a Figure object’s renderer. For example, this is useful for defining click events where you want to map click positions to the nearest graphic object. See the previous section for a description of all the renderer events.

Renderer event handlers can be added using a method or a decorator.

For example:

import fastplotlib as fpl
import numpy as np

# generate some circles
def make_circle(center, radius: float, n_points: int = 75) -> np.ndarray:
    theta = np.linspace(0, 2 * np.pi, n_points)
    xs = radius * np.sin(theta)
    ys = radius * np.cos(theta)

    return np.column_stack([xs, ys]) + center

# this makes 5 circles, so we can create 5 cmap values, so it will use these values to set the
# color of the line based by using the cmap as a LUT with the corresponding cmap_value
circles = list()
for x in range(0, 50, 10):
    circles.append(make_circle(center=(x, 0), radius=4, n_points=100))

# create figure
fig = fpl.Figure()

# add circles to plot
circles_graphic = fig[0,0].add_line_collection(data=circles, cmap="tab10", thickness=10)

# get the nearest graphic that is clicked and change the color
@fig.renderer.add_event_handler("click")
def click_event(ev):
    # reset colors
    circles_graphic.cmap = "tab10"

    # map the click position to world coordinates
    xy = fig[0, 0].map_screen_to_world(ev)[:-1]

    # get the nearest graphic to the position
    nearest = fpl.utils.get_nearest_graphics(xy, circles_graphic)[0]

    # change the closest graphic color to white
    nearest.colors = "w"

fig.show()

ImageWidget

Often times, developing UIs for interacting with multi-dimension image data can be tedious and repetitive. In order to aid with common image and video visualization requirements the ImageWidget automatically generates sliders to easily navigate through different dimensions of your data. The image widget supports 2D, 3D and 4D arrays.

Let’s look at an example:

import fastplotlib as fpl
import imageio.v3 as iio

movie = iio.imread("imageio:cockatoo.mp4")

# convert RGB movie to grayscale
gray_movie = np.dot(movie[..., :3], [0.299, 0.587, 0.114])

iw_movie = ImageWidget(
data=gray_movie,
cmap="gray"
)

iw_movie.show()

Animations

An animation function is a user-defined function that gets called on every rendering cycle. Let’s look at an example:

import fastplotlib as fpl
import numpy as np

# generate some data
start, stop = 0, 2 * np.pi
increment = (2 * np.pi) / 50

# make a simple sine wave
xs = np.linspace(start, stop, 100)
ys = np.sin(xs)

figure = fpl.Figure(size=(700, 560))

# plot the image data
sine = figure[0, 0].add_line(ys, name="sine", colors="r")


# increment along the x-axis on each render loop :D
def update_line(subplot):
    global increment, start, stop
    xs = np.linspace(start + increment, stop + increment, 100)
    ys = np.sin(xs)

    start += increment
    stop += increment

    # change only the y-axis values of the line
    subplot["sine"].data[:, 1] = ys


figure[0, 0].add_animations(update_line)

figure.show(maintain_aspect=False)

Here we are defining a function that updates the data of the LineGraphic in the plot with new data. When adding an animation function, the user-defined function will receive a subplot instance as an argument when it is called.

Spaces

There are several spaces to consider when using fastplotlib:

1. World Space

   World space is the 3D space in which graphical objects live. Objects and the camera can exist anywhere in this space.

2. Data Space

   Data space is simply the world space plus any offset or rotation that has been applied to an object.

Note

World space does not always correspond directly to data space, you may have to adjust for any offset or rotation of the Graphic.

3. Screen Space

   Screen space is the 2D space in which your screen pixels reside. This space is constrained by the screen width and height in pixels. In the rendering process, the camera is responsible for projecting the world space into screen space.

Note

When interacting with Graphic objects, there is a very helpful function for mapping screen space to world space (Figure.map_screen_to_world(pos=(x, y))). This can be particularly useful when working with click events where click positions are returned in screen space but Graphic objects that you may want to interact with exist in world space.

For more information on the various spaces used by rendering engines please see this article

Imgui

Fastplotlib uses imgui_bundle to provide within-canvas UI elemenents if you installed fastplotlib using the imgui toggle, i.e. fastplotlib[imgui], or installed imgui_bundle afterwards.

Fastplotlib comes built-in with imgui UIs for subplot toolbars and a standard right-click menu with a number of options. You can also make custom GUIs and embed them within the canvas, see the examples gallery for detailed examples.

Note

Imgui is optional, you can use other GUI frameworks such at Qt or ipywidgets with fastplotlib. You can also of course use imgui and Qt or ipywidgets.

Using fastplotlib in an interactive shell

There are multiple ways to use fastplotlib in interactive shells, such as ipython.

1. Jupyter

On jupyter lab the jupyter backend (i.e. jupyter_rfb) is normally selected. This works via client-server rendering. Images generated on the server are streamed to the client (Jupyter) via a jpeg byte stream. Events (such as mouse or keyboard events) are then streamed in the opposite direction prompting new images to be generated by the server if necessary. This remote-frame-buffer approach makes the rendering process very fast. fastplotlib viusalizations can be displayed in cell output or on the side using sidecar.

A Qt backend can also optionally be used as well. If %gui qt is selected before importing fastplotlib then this backend will be used instead.

Lastly, users can also force using glfw by specifying this as an argument when instantiating a Figure (i.e. Figure(canvas="gflw").

Note

Do not mix between gui backends. For example, if you start the notebook using Qt, do not attempt to force using another backend such as jupyter_rfb later.

2. IPython

Users can select between using a Qt backend or gflw using the same methods as above.