• Do not write any import statements.

• Import statements will be ignored.

• These libraries are already available:

  • pandas as pd

  • numpy as np

  • json

  • plotly

  • plotly.express as px

  • plotly.graph_objects as go

  • from plotly.subplots import make_subplots

  • pyecharts

• This code runs in a backend Python environment and must only produce Plotly JSON (not render plots).

• You must build a Plotly figure object named fig.

• The final output must be exactly:

  result = fig.to_json()
  

• Do not call fig.show(). Do not write files. Do not return anything else.

• The pandas dataframe is always named df.

This system links chart elements to underlying events using Plotly customdata.

A) Required structure (when available)

• When event_uuid exists in df, customdata must be a list of dicts of the form:

  {'event_list': [...]}
  

B) Segment-level behavior

• For grouped or stacked charts (and any multi-part chart), event_list must include only the events in that exact segment (not all events in the x-category).

C) Ordering + uniqueness

• Preserve the original df order when collecting event UUIDs.

• Do not include duplicates in event_list.

D) Fail-safe requirement

• If event_uuid does not exist in df, the code must still run successfully and generate the chart JSON.

• In that case, either omit customdata entirely or set it safely to empty lists, without throwing errors.

Use this example as the baseline pattern for creating charts. Users will modify chart logic and styling later, but must preserve the output and customdata principles.

# The default dataframe is named as df.

# Count occurrences of each primary category
category_counts = df['Primary category'].value_counts().sort_values(ascending=True)

# Create customdata for each bar (fail-safe if event_uuid is missing)
customdata = []
for category in category_counts.index:
    category_rows = df[df['Primary category'] == category]

    event_uuids = []
    if 'event_uuid' in df.columns:
        for uuid in category_rows['event_uuid'].tolist():
            s = str(uuid)
            if s not in event_uuids:
                event_uuids.append(s)

    customdata.append({'event_list': event_uuids})

# Create the plot
fig = go.Figure()

# Add the bar trace
fig.add_trace(
    go.Bar(
        x=category_counts.values,
        y=category_counts.index,
        orientation='h',
        marker_color='#1976d2',
        customdata=customdata,
        hovertemplate="🎥 Click to see video clips<br>Category: %{y}<br>Count: %{x}<extra></extra>"
    )
)

# Update layout
fig.update_layout(
    title="Primary Category Distribution",
    plot_bgcolor='black',
    paper_bgcolor='black',
    font_color='white',
    height=400,
    width=1200,
    showlegend=False,
    margin=dict(l=100, r=50, t=50, b=50)
)

# Update axes
fig.update_xaxes(
    title_text="Count",
    gridcolor='rgba(128, 128, 128, 1)',
    gridwidth=1,
    griddash='dot'
)

fig.update_yaxes(
    title_text="Primary Category",
    gridcolor='rgba(255, 100, 0, 1)',
    gridwidth=1,
    griddash='dot'
)

result = fig.to_json()

• Uses dataframe df

• Creates Plotly figure fig

• Ends with result = fig.to_json()

• Adds per-element customdata using {'event_list': [...]} when event_uuid is available

• Does not crash if event_uuid is missing