How to convert a FigureResampler to a json format

[erwanmq]

Hi!

I am using plotly_resampler with manual mode to plot hundred of thousands points in a plotly graph. I want the user to add or remove events from this plot. To add the events, it's quite simple, I have to call .add_trace(). But to remove the events 2 options are available. Either I use .visible = False or I transform the FigureResampler to a json format and remove the last data value.
But when I do this, the json format loses all the points that are not plotted in the graph. So I can lost 80-90% of my points and I can't get them back.

Is there a solution to erase a trace from a FigureResampler without losing all points and losing the state of 'FigureResampler'?

I should add that I would like to avoid .visible = False for the sake of performance. Thank you!

[jonasvdd]

Hi,

Regarding conversion to a json format

FigureResampler only renders aggregated data of the high-frequency traces.
This is achieved by storing this aggregated data in plotly's default data property and storing the raw, high-frequency, data in the (non-plotly) _hf_data property.

As such, when the plotly's to_json() is invoked, only the aggregated data, stored within the data property, will be serialized into the .json file. So there is no convenient method to store the high frequency data into a .json file. Also remark that .json is an inefficient manner to store (large amounts of) data as e.g. floats will be encoded as strings.

However, FigureResampler figures can be serialized into pickle objects (which stores the data-as-is (i.e. in binary format). Maybe you could use this format to save your FigureResampler figures?

plotly-resampler/tests/test_serialization.py

Lines 58 to 65 in 0a68cb7

	fig = FigureResampler(default_n_shown_samples=50, show_dash_kwargs=dict(port=8051))
	for i in range(nb_traces):
	fig.add_trace(go.Scattergl(name=f"trace--{i}"), **_get_hf_props(nb_samples))
	assert fig._show_dash_kwargs["port"] == 8051
	pickle.dump(fig, open(pickle_figure, "wb"))
	fig_pickle = pickle.load(open(pickle_figure, "rb"))

Regarding erasing a trace

I quickly glanced at the plotly-community forum, but found no straightforward solutions to remove a trace.

After toying around with this in a notebook I came to a (not 100% clean) solution.

1. Create a figure with multiple traces:

import plotly.graph_objects as go
from plotly_resampler import FigureResampler

import numpy as np

labels = list(range(0, 3))
np.random.seed(0xDEADBEEF)
x = np.arange(100000)
y = np.random.normal(size=100000)

fig = FigureResampler()
fig.add_trace(go.Scatter(x=x, y=y, mode='lines', line_color='lightblue'))
fig.add_trace(go.Scatter(x=x, y=y + 10, mode='lines+markers', line_color='salmon'))
fig.add_trace(go.Scatter(x=x, y=y + 20, mode='lines', line_color='lightseagreen'))
fig.update_layout(height=500, margin=dict(l=0, r=0, t=0, b=0))
fig.show_dash()

2. Delete a trace from the above created figure:

trace_index = 1  # specify the trace index

# delete the `hf_data` entry
fig._hf_data.pop(fig.data[trace_index].uid)  # the hf_data is a dict keyed by trace uid

#  delete the trace itself in the `data` list
fig.data[1].visible = False
fig.data[1].x = []
fig.data[1].y = []

Regarding removing a data-value (e.g. the last value) from a trace

Could you provide me with a minimal code example regarding your use-case?

---

I hope this helps you any further!
Kind regards,
Jonas