Errors on Notebook logs when refreshing a page with urth widgets about declarativewidgets

[wellecks]

I'm starting to investigate why we are seeing these errors

[parente]

Have seen this often in dashboards that use widgets. You probably know it, but the JS left over on the page seems to execute using channels that no longer exist. Have never seen it cause a problem other than the noise in the log.

[lbustelo]

I traces it a while back to local storage and ipywidgets widgetmanager. When we create a widgetmodel, the comm info gets stored in local storage to support page refreshes. On a refresh, widgetmanager tries to reconnect.

Now, this error I think happens after rebooting the whole notebook and still having the old comm info associated with the page's url.

We should investigate if refresh works for us and open a separate issue if not.

Gino B.

On Oct 2, 2015, at 10:58 AM, Peter Parente [email protected] wrote:

Have seen this often in dashboards that use widgets. You probably know it, but the JS left over on the page seems to execute using channels that no longer exist. Have never seen it cause a problem other than the noise in the log.

—
Reply to this email directly or view it on GitHub.

[wellecks]

Right - it looks like the notebook saves the widget manager's state (which include each widget's state) in local storage. It then requests the kernel-side state (via comm message) on refresh using the state in local storage, which may have comms that no longer exist in the kernel

[wellecks]

So there's an extra step in reproducing it - in between steps 3 and 4, restart the kernel. This clears out the kernel's comms, while the comm's are still around in the local storage

[wellecks]

It turns out that the same error message occurs with IPython widgets. In a fresh notebook:

from IPython.html import widgets
from IPython.display import display

w = widgets.FloatSlider(value=0,min=0,max=200,step=1)
display(w)

then run cells, save, restart the kernel, and refresh

[wellecks]

It looks like they may have fixed this in ipywidgets 4.0: jupyter-widgets/ipywidgets#62 (see 'request_state is not called for invalid ids'). We're using the IPython 3.0 widgets here though.

I think since it is/was a known issue that is independent of our widgets, we should close this. Thoughts?

[lbustelo]

I agree with closing.

Gino B.

On Oct 2, 2015, at 1:27 PM, Sean Welleck [email protected] wrote:

It looks like they may have fixed this in ipywidgets 4.0: jupyter-widgets/ipywidgets#62 (see 'request_state is not called for invalid ids'). We're using the IPython 3.0 widgets here though.

I think since it is/was a known issue that is independent of our widgets, we should close this. Thoughts?

—
Reply to this email directly or view it on GitHub.