ipython_blocking

ipython_blocking is a context manager for capturing cell execution messages in a Jupyter notebook, along with magic commands %block and %blockrun for convenience. The primary use-case for blocking notebook execution is to wait for users to interact with ipywidgets and then reference the values entered in those Widgets.

Install

ipython_blocking is on PyPI, install with pip.

pipinstallipython_blocking

Usage

Try out the demo notebooks in Binder to see ipython_blocking in action. The most common way to use ipython_blocking is with the %blockrun magic and running a notebook with "cell -> run all". %blockrun button stops the cell execution messages from the initial "cell -> run all", and attaches a "cell -> run all below" handler to the button so that a notebook can be run in a linear fashion without callback functions after a user has filled out other Widget values.

### cell #1importipywidgetsaswidgetsimportipython_blocking# enables %block and %blockrun magictext=widgets.Text()
dropdown=widgets.Dropdown(options=['', 'foo', 'bar', 'baz'])
button=widgets.Button(description='Run')
box=widgets.VBox(children=[text, dropdown, button])
box### cell #2%blockrunbutton### cell #3 -- doesn't execute until the 'Run' button is pressed### This gives the user a chance to interact with the Text and Dropdown widgetsprint(text.value)
print(dropdown.value)

(The dropdown menu doesn't appear in this .gif because its treated as a "separate window" in Windows screen capture, sorry!)

CaptureExecution

The way ipython_blocking "blocks" cell execution is by creating a context manager that changes the behavior of the IPython.shell.kernel['execute_request'] handler. When you execute a cell in a Jupyter notebook, it sends a execute_request comms message to the kernel with that code.

While the CaptureExecution manager is "blocking", it stores those messages in a list instead of actually executing them. When the context manager exits, it resets the handler to its original behavior and then either replays the stored messages or drops them.

importipython_blockingctx=ipython_blocking.CaptureExecution(replay=True)
withctx:
whileTrue:
ifbreak_function():
breakctx.step() # handles all other messages that aren't 'execute_request' including widget value changes

%block

The %block magic is enabled upon importing ipython_blocking. It takes either a function or widget object and creates the CaptureExecution manager to block until that function returns True or the widget value changes.

# cell 1importipywidgetsaswidgetsimportipython_blockingdd=widgets.Dropdown(options=['', 'foo', 'bar', 'baz'])
dd# cell 2%blockdd# cell 3# Won't actually be executed until the user chooses an option in the dd widgetprint(dd.value)

%blockrun

The %blockrun magic is similar to %block but it only accepts an ipywidgets.Button target and it attaches a "cell -> run all below" handler to the button. If you expect the application logic of your Notebook to be run more than once (and/or don't want to reinitialize the Widgets because the user might only change one of many options), then %blockrun is the better magic to use.

It is often handy to make the target Button unclickable when the Notebook first renders, then add .observe handlers on other Widgets that can make the Button clickable once some input validation has happened.

# cell 1importipywidgetsaswidgetsimportipython_blockingtext=widgets.Text()
dd=widgets.Dropdown(options=['', 'foo', 'bar', 'baz'])
button=widgets.Button(description='Run', disabled=True)
defvalidation(ev):
"make button clickable if user has put in more than 5 characters and chosen a dropdown option"iflen(text.value) >5anddd.value:
button.disabled=Falseelse:
button.disabled=Truetext.observe(validation)
dd.observe(validation)
box=widgets.VBox(children=[text, dd, button])
box# cell 2%blockrunbutton# cell 3print(text.value)
print(dd.value)

Alternatives

The other ways to get the value of a widget after a user has interacted with the widget is to structure your notebook with event callbacks or to write your code asynchronously.

I believe there are major benefits to writing the application logic of a Jupyter Notebook in a linear and synchronous fashion, with as many variables as possible in the notebook global scope. Those benefits include:

• Better introspection and comprehension of the workflow (without littering your code with global and print statements)
• More direct debug when something goes wrong
• Easier to break code into small blocks/cells