Saving and Loading Interactive Plots

This example shows how to save and reopen interactive plots created with tsunami_ip_utils. This is useful for sharing plots with collaborators or saving plots for later use.

Non-Custom Interactive Plots

The interactive plots generated in Visualizing Contributions can all be saved using the write_html method of the plotly figure object. This method writes the plot to an HTML file, which can be opened in a web browser to view the interactive plot. Even for tsunami_ip_utils.viz.pie_plot.InteractivePieLegend plots (which aren’t plotly figures, but rather a custom wrapper around a plotly figure), the same write_html method can be used.

Saving Plots

However, for plots that implement interactive python callbacks, such as tsunami_ip_utils.viz.scatter_plot.InteractiveScatterLegend plots, the interactivity cannot fully be saved to html because the callbacks are implemented in python (though the current static html state of the plot can be written to html with the tsunami_ip_utils.viz.scatter_plot.InteractiveScatterLegend.write_html() method). So, to save these plots, a different approach is taken. These plots are serialized as python pickle objects, which can be reloaded later by anyone with tsunami_ip_utils installed. To serialize an interactive scatter plot, the tsunami_ip_utils.viz.scatter_plot.InteractiveScatterLegend.save_state() method is used. This method just takes a path to the output pickle file to save the plot state to.

First we need to get some data to create an interactive scatter plot. We’ll use the same code as in Contribution Correlation Plots

from tsunami_ip_utils.viz.viz import correlation_plot
from tsunami_ip_utils.integral_indices import get_uncertainty_contributions
from paths import EXAMPLES
import matplotlib.pyplot as plt

application_files = [ EXAMPLES / 'data' / 'example_sdfs' / 'MCT' / f'MIX-COMP-THERM-001-001.sdf' ]
experiment_files = [ EXAMPLES / 'data' / 'example_sdfs' / 'MCT' / f'MIX-COMP-THERM-004-009.sdf' ]

contributions_nuclide, contributions_nuclide_reaction = get_uncertainty_contributions(
    application_files,
    experiment_files,
)

# Now get the application and experiment contributions
application_contributions = contributions_nuclide_reaction['application'][0]
experiment_contributions  = contributions_nuclide_reaction['experiment'][0]

fig = correlation_plot(
    application_contributions,
    experiment_contributions,
    integral_index_name='%(Δk/k)^2', # Note that Plotly plots don't support pylatex in labels, so we just use a plain string
    plot_type='interactive_scatter',
    interactive_legend=True,
)
fig.save_state( EXAMPLES / 'scatter_plot.pkl' )

Loading Plots

Now, that we’ve serialized the plot to a .pkl file, we can reload it using the tsunami_ip_utils.viz.scatter_plot.load_interactive_scatter_plot() function.

from tsunami_ip_utils.viz.scatter_plot import load_interactive_scatter_plot

fig = load_interactive_scatter_plot( EXAMPLES / 'scatter_plot.pkl' )
fig.show()

and as the above shows, the full interactivity of the scatter plot (including the legend updates with the recalculated summary statistics and linear regression) is retained.

We can also save the plot to an image

fig = load_interactive_scatter_plot( EXAMPLES / 'scatter_plot.pkl' )
fig.to_image( EXAMPLES / '_static' / 'loaded_scatter_plot.png' )

# sphinx_gallery_thumbnail_path = '../../examples/_static/loaded_scatter_plot.png'

Other Interactive Plots

The only other polts that have to be serialized like above to retain their interactivity are matrix plots (i.e. those created by tsunami_ip_utils.viz.viz.matrix_plot`()). These figure objects also have a save state method (i.e. tsunami_ip_utils.viz.matrix_plot.InteractiveMatrixPlot.save_state()), which can be used to save the plot state to a pickle file. And these plots can be reloaded, likewise, using the tsunami_ip_utils.viz.matrix_plot.load_interactive_matrix_plot() function. An example is shown below for a matrix that is very time consuming to generate.

from tsunami_ip_utils.viz.matrix_plot import load_interactive_matrix_plot

fig = load_interactive_matrix_plot( EXAMPLES / 'data' / 'hmf_contribution_comparisons_nuclide_reaction.pkl' )
fig.show()

- Zoom Out + Zoom In