embed-pyqtgraph-tutorial

This is a walkthrough of embedding pyqtgraph content in a PyQt application you're designing with Qt Designer. Here's an overview of the steps involved:

• Create the layout of the UI in Qt Designer (generates template.ui)
• Compile the XML template to Python source (uses build_ui.py to generate template.py)
• Write a custom QWidget class which takes on the UI and populates the embedded pyqtgraph widget (application.py)

The example provided displays a PlotWidget with some data and a QCheckBox to enable/disable mouse control of the plot. It's a simple example, but once you've seen the process of adding pyqtgraph widgets to your UI, you can create arbitrarily complex applications without having to write a ton of boilerplate layout code.

Prerequisites

I'm going to assume you have some way of creating Qt applications with Python, and the instructions here should not be much (if at all) different between the ways to accomplish that (i.e. PyQt4, PyQt5, PySide, PySide2). For reference, I'm using PyQt5 and Python 3, so the following works for setting up an environment for running the example:

$ python -m venv .venv
$ source .venv/bin/activate
(.venv) $ pip install pyqt5 pyqtgraph
(.venv) $ python application.py

If you're using a different setup (e.g. PyQt4, PySide), you'll need to fix some imports before being able to run the example application.

I'm also going to assume you have have some knowledge of how to use pyqtgraph. The pyqtgraph examples cover its capabilities pretty well.

Tutorial

Creating the UI Layout

┌────────────┐   ┌─────────────┐   ┌─────────────┐
│ creativity │ → │ Qt Designer │ → │ template.ui │
└────────────┘   └─────────────┘   └─────────────┘

Generally, you can use custom widgets in Qt Designer through the promote mechanism. This means you add an item of the base class of your custom item to your UI layout, right click on it, then go to "Promote to ...". Since we're using Python instead of C++, the "Header file" in this case is going to be the Python namespace at which you would access the class as though you were importing it. For example, if you had a custom widget class, where an import statement might look like:

from myproject.widgets import CustomDialog

you would put myproject.widgets in the "Header file" field and CustomDialog in the "Promoted class name" field. The basic Qt class you inherit from (e.g. QDialog, QWidget, etc.) is then the "Base class name."

pyqtgraph actually provides some documentation for how to embed plotting objects into a Qt UI using Qt Designer (see embed pyqtgraph), but it's a bit lacking. In our case, add a QGraphicsView to the layout where you want the plot to be, then promote it to a PlotWidget. Most pyqtgraph classes are available under the pyqtgraph namespace, so just put "pyqtgraph" in the header file field.

If you inspect the template.ui file, you can see towards the bottom there is a customwidget section specifying everything we just covered:

class	PlotWidget
base class	QGraphicsView
namespace	pyqtgraph

Compiling the Layout

┌─────────────┐   ┌─────────────┐   ┌─────────────┐
│ template.ui │ → │ build_ui.py │ → │ template.py │
└─────────────┘   └─────────────┘   └─────────────┘

There are a few ways to turn your UI template into Python code. The first (my preferred) way of doing it is to write a little script to grab all of your *.ui files and use the pyuic module to write the corresponding Python code to the desired destination. An example is given here (build_ui.py). You can modify the SRC and DEST variables if you want to arrange your project structure differently. My recommendation is to put your UI files in something like res/ and the compiled Python files in your project source to make importing simple (see the next step).

You could instead use pyuic directly from the command line. For single UI templates, this is fine, but you'll want a scripted approach if you end up with more. For reference, here's how you'd convert the template given here:

(.venv) $ pyuic5 template.ui -o template.py

The third option would be to use the UI template directly from your hand-written Python code (i.e. using uic.loadUI()). I tend to avoid this to make packaging the project easier, but it is probably not all that difficult to package the UI templates. If you're already dealing with packaging custom icon files and such, this would be a non-issue.

Now, if you take a look at the template.py file, you'll see there's a class called Ui_CustomWidget that has, primarily, a setupUi method to create and lay out all of the content we designed in Qt Designer. This is a pretty simple example, but with complex layouts with many items, the use of Qt Designer and automatic generation of this code really shines.

Also notice the import line at the bottom of the file:

from pyqtgraph import PlotWidget

This all comes straight from that bottom section of the template.ui file, where we specified in Qt Designer that we have a custom item of a specific class and where it can be found.

Wrapping the UI in a Custom QWidget

┌─────────────┐   ┌────────────────┐   ┌─────────┐
│ template.py │ → │ application.py │ → │ success │
└─────────────┘   └────────────────┘   └─────────┘

Now for the difficult part. PyQt provides some documentation for using the generated code, though it is a little sparse if you don't have the high-level overview of the whole process. Essentially, you write a custom class (usually QWidget, but really it can be any QWidget-like class such as QMainWindow, QDialog, etc.), then use the compiled Python code to access and interact with the layout you specified. The key here is that you are not modifying the generated Python code. That means in our case that we don't want to mess with template.py by hand. If you make changes to it, you can no longer make changes to your UI from Qt Designer because re-generating the template will overwrite your changes!

The nice thing about this workflow overall is that it cleans up your custom class code a lot, since you're not creating container layouts all over the place, specifying sizes, size policies, etc.

Our implementation is in application.py. It starts out with importing Qt stuff so we can set up a QApplication and run it. We also import the template module so we have access to the template class (PyQt refers to it as the form class). We then write a custom QWidget implementation which instantiates the form class and calls its setupUi method. I typically assign the form class to an attribute called ui. Once you call setupUi, you now can access the items in your UI through the names they were given in Qt Designer.

So in our example, we access the plotWidget attribute of the form class object, which is a pyqtgraph PlotWidget object. Now the pyqtgraph API applies, so we can plot some content and whatever else we need. In this case, we connect up the checkbox to a method that toggles mouse functionality on the PlotWidget.

Note: It is possible to make CustomWidget inherit from UI_CustomWidget (so you'd call ``self.setupUi()``), but I prefer to explicitly set up the form class object as an attribute of the custom widget -- mostly because it's very clear when you're accessing UI elements specified by Designer rather than widgets you might add programmatically.

Other Notes

Feel free to open an issue if anything here isn't clear.

I originally wrote this in response to seeing questions on the pyqtgraph mailing list about using pyqtgraph functionality in an application designed with Qt Designer. It's not obvious at all that you can use the promote mechanism with PyQt (as opposed to C++ Qt), so I wrote this to help people out. Feel free to use these materials in a pull request to improve pyqtgraph documentation.