Support for updating datasets?

[SamLau95]

Hey all, I'd like to use altair to make interactive demos for teaching data science concepts. For teaching purposes, it would be really helpful to have some way of updating a chart's dataset after the initial render.

For example, we'd like to show a single histogram of sample means becoming narrower as the sample size #increases. We can do this using ipywidgets, but as described in #333 the chart flickers significantly. My hope is that just focusing on dataset changes rather than chart updates in general will be better scoped and easier to implement (I can offer my own time to work on this).

Based on Vega-lite's view.change method, I'm imagining something like:

data = alt.Data(values=[{'x': 'A', 'y': 5},
                    {'x': 'B', 'y': 3},
                    {'x': 'C', 'y': 6},
                    {'x': 'D', 'y': 7},
                    {'x': 'E', 'y': 2}])
chart = alt.Chart(data).mark_bar().encode(
    x='x:O',  # specify ordinal data
    y='y:Q',  # specify quantitative data
)

# later...
chart.change(alt.ChangeSet().insert({'x': 'F', 'y': 1}))

# and using ipywidgets:
def update(new_height):
    chart.change(alt.ChangeSet().insert({'x': 'F', 'y': new_height}))
interact(update, new_height=(1, 10))

Any opinions on this? I'm peeking at the altair and ipyvega code and this appears doable but not super easy. Maybe you'll have a better idea of how hard this will be.

cc @CalebS97

[ellisonbg]

We can likely get this to work with jupyter's new output updating...

…

On Mon, Oct 23, 2017 at 3:54 PM, Sam Lau ***@***.***> wrote: Hey all, I'd like to altair to make interactive demos for teaching data science concepts. For teaching purposes, it would be really helpful to have some way of updating a chart's dataset after the initial render. For example, we'd like to create a histogram of sample means becoming narrower as the sample size increases. We can do this using ipywidgets, but as described in #333 <#333> the chart flickers significantly. My hope is that just focusing on dataset changes rather than chart updates in general will be better scoped and easier to implement (I can offer my own time to work on this). Based on Vega-lite's view.change method <https://vega.github.io/vega/docs/api/view/#view_change>, I'm imagining something like: data = alt.Data(values=[{'x': 'A', 'y': 5}, {'x': 'B', 'y': 3}, {'x': 'C', 'y': 6}, {'x': 'D', 'y': 7}, {'x': 'E', 'y': 2}]) chart = alt.Chart(data).mark_bar().encode( x='x:O', # specify ordinal data y='y:Q', # specify quantitative data ) # later... chart.change(alt.ChangeSet().insert({'x': 'F', 'y': 1})) # and using ipywidgets:def update(new_height): chart.change(alt.ChangeSet().insert({'x': 'F', 'y': new_height})) interact(update, new_height=(1, 10)) cc @CalebS97 <https://github.com/calebs97> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#426>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABr0N-HPK1MyQx3egOxXX5a7XVfIuXcks5svRkmgaJpZM4QDm1Z> .

-- Brian E. Granger Associate Professor of Physics and Data Science Cal Poly State University, San Luis Obispo @ellisonbg on Twitter and GitHub bgranger@calpoly.edu and ellisonbg@gmail.com

[tonylee3399]

I actually stumbled upon this needs. Would this feature be considered in the updates in the future? Thanks in advance!

[jakevdp]

This would definitely be considered; what it lacks right now is a motivated person to work on it. Are you interested?

[dangotbanned]

Related

• public facing method to update a dataset in place #3318