Use Case: advanced dashboard for data exploration

[KonradUdoHannes]

In order to more easily explore the available data I would recommend a dashboard, that combines
functionality for partial text search with displaying of meta information and query results.
I believe a combination of the following features would allow good data exploration

1. Partial text search for the long description of Statistics, coupled to a display/selectability of the statistics that are filtered this way.
2. Display of the related Enums/Dimensions, their description and their members for a selected statistic.
3. Full text search for regions and/or a hierarchy coupled navigation of the regions
4. Ability to run a query based on the a selection of the previously specified elements
5. Display additional relevant meta information like for instance units

The ability to implement point 2. or something similar is coupled to issue #76

[elekt]

@KonradUdoHannes

I looked into this and indeed a panel dashboard would be a cool solution.
However I run into the same problem as before, what is out of the box interactive with ipywidgets is not working with panel.

panel dashboard notebook

Could you look at it? Maybe its something trivial.

[KonradUdoHannes]

I had a look and the main issue is that in the default panel functionality (the one used by interact) widgets modify panes. Replacing pn.widgets.DataFrame(df_search_long_desc ... to pn.pane.DataFrame(df_search_long_desc ... creates a working example for me. Generally panels interact is also quite good in automatically converting the arguments to a fitting pane. In this case it means that passing a function that simply returns a dataframe instead of a pane also works for me.

In order to create dynamic UIs I think we might also need other APIs instead or on top of interact.
I think both the Reactive functions API and the Parametrized class API allow for dynamically changing widgets.

[enryH]

If you change the global state of a ipython widget (say by another ipython widget) it will change "dynamacially". Sometimes constructing this can be a pain (I have an example for data-filtering based on several widgets somewhere).

I guess if this works with widgets, it should also work with panel?

[elekt]

Replacing widgets to pane is nice and makes it very dynamical, but the selecting functionality of pn.widgets.DataFrame is lost. That would be so good otherwise to actually select statistics from the queries.

[elekt]

I use multi select widgets instead. Would have been super elegant to use the row selection of the dataframe widget but this will do too