Plotly Dash: Render a Graph only if certain data in a Pandas DataFrame exists

1
votes

I currently use Plotly Dash to create a webapp to visualize data. The syntax involves using dcc.Graph to render an HTML Div as follows. Below the html.Div is the code to render the plot figure. The x and y axis values come from a Pandas DF. Can anyone recommend a way to use an IF statement to check if a given column exists and then proceed to return the following code. Plotly Dash gives an error when a 'fig' variable does not exist

html.Div([
    html.Div([
        html.H3('Chart'),
        dcc.Graph(id='Chart',figure = chart_fig
)



sensor3_trace_day1 = Scatter(
y = sensor_3_day1['BG Calibration (mg/dl)'],
x = sensor_3_day1['Current (nA)'],
mode = 'markers',
name = 'Sensor 3 Day 1'
)    
data = [sensor1_trace]

layout = Layout(showlegend=True, height = 600, width = 600, title='Day     1: BG Cal vs Current',xaxis={'title':'Current [nA]'},yaxis={'title':'BG Calibration (mg/dl)'})

chart_fig = dict(data=data, layout=layout)

iplot(chart_fig)
2
pythonplotlyplotly-dash

2 Answers

2
votes

Another possible solution is to create a callback and show an empty plot, under certain conditions. This example is very similar to the one by @lwileczek

def empty_plot(label_annotation):
    '''
    Returns an empty plot with a centered text.
    '''

    trace1 = go.Scatter(
        x=[],
        y=[]
    )

    data = [trace1]

    layout = go.Layout(
        showlegend=False,
        xaxis=dict(
            autorange=True,
            showgrid=False,
            zeroline=False,
            showline=False,
            ticks='',
            showticklabels=False
        ),
        yaxis=dict(
            autorange=True,
            showgrid=False,
            zeroline=False,
            showline=False,
            ticks='',
            showticklabels=False
        ),
        annotations=[
            dict(
                x=0,
                y=0,
                xref='x',
                yref='y',
                text=label_annotation,
                showarrow=True,
                arrowhead=7,
                ax=0,
                ay=0
            )
        ]
    )

    fig = go.Figure(data=data, layout=layout)
    # END
    return fig

# Layout
html.Div([
    html.Div([
        # Here it goes the dropdown menu
        # ....

        # The graph is defined below
        html.H3('Chart'),
        dcc.Graph(id='chart')
         ]),
     ])

# Callback
@app.callback(Output('chart','figure'),
              [Input('dropdown','value')])
def update_fig(drop_val):
    if drop_val in df.columns:
        # fig = (...)  Here it goes your figure
    else:
        fig = empty_plot('Nothing to display') # Here it goes the empty plot
    return fig
1
votes

What you can do is use two different @app.callbacks(). The first would return a figure and the second would return a "style" dictionary for the dcc.Graph() object. If the figure is empty you don't want to show the div. If you have a dropdown or a button that you use to update the information i.e. read in/ create the dataframe you can use that as input for your callbacks. For example:

html.Div([
    html.Div([
        html.H3('Chart'),
        dcc.Graph(id='chart')
         ]),
     ])

@app.callback(Output('chart','figure'),
              [Input('dropdown','value')])
def update_fig(drop_val):
    if "col" in df.columns:
        fig = {'data':data,'layout':layout}
    else:
        fig = {'data':[],'layout':layout}
    return fig

@app.callback(Output('chart','style'),
              [Input('dropdown','value')])
def update_style(drop_val):
    if "col" in df.columns:
        style = {'borderStyle':solid,'display':'inline-block'}
    else:
        style = {'display':'none'}
    return style

In this case I have some imaginary dropdown that may flip between which df is being read in or something and then outputs the display and a figure each time. It must return a workable figure no matter what so make up 

data = [{'x':[1,2,4], 'y':[4,5,6], 'type':'bar'}]

or something similar and just set display to "none" so it wont be rendered.