Data Visualization
Best practices for creating charts, tables, and live data displays.
Plotly Charts
Creating a Chart
import plotly.express as px
# Create figure
df = pd.DataFrame({"x": [1, 2, 3], "y": [4, 5, 6]})
fig = px.line(df, x="x", y="y", title="My Chart")
# Show as widget
show_plotly(figure_json=fig.to_json(), title="Line Chart")
Updating a Chart
# Full update
new_fig = px.bar(df, x="category", y="value")
update_plotly(widget_id, figure_json=new_fig.to_json())
# Layout-only update (faster - doesn't re-render data)
layout_update = {"title": {"text": "Updated Title"}}
update_plotly(widget_id, figure_json=json.dumps({"layout": layout_update}), layout_only=True)
Theme Consistency
PyWry automatically switches Plotly charts between plotly_dark and plotly_white
templates when the theme toggles. To customize per-theme styles while preserving
automatic switching, use template_dark and template_light on PlotlyConfig:
from pywry import PlotlyConfig
config = PlotlyConfig(
template_dark={"layout": {"paper_bgcolor": "#1a1a2e", "plot_bgcolor": "#16213e"}},
template_light={"layout": {"paper_bgcolor": "#ffffff", "plot_bgcolor": "#f0f0f0"}},
)
show_plotly(figure_json=fig.to_json(), title="Themed Chart", config=config)
User values are deep-merged on top of the built-in base template — your overrides always win, and anything you don't set is inherited from the base.
Chart Sizing
# Let chart fill container
fig.update_layout(
autosize=True,
margin=dict(l=40, r=40, t=50, b=40),
)
# Or set explicit size
fig.update_layout(
width=600,
height=400,
)
AG Grid Tables
Creating a Table
import json
# Data as list of dicts
data = [
{"name": "Alice", "age": 30, "city": "NYC"},
{"name": "Bob", "age": 25, "city": "LA"},
]
# Show as AG Grid widget
show_dataframe(data_json=json.dumps(data), title="Users")
Theme Consistency
# AG Grid themes
# - ag-theme-quartz-dark (dark mode)
# - ag-theme-quartz (light mode)
# Widget auto-applies based on current theme
Large Datasets
# AG Grid handles large datasets efficiently via virtualization
# Pagination is built-in
data = [{"id": i, "value": random.random()} for i in range(10000)]
show_dataframe(data_json=json.dumps(data), title="Large Dataset")
# Grid virtualizes - only visible rows are rendered
Marquee for Live Data
Building Ticker Items
# Build individual ticker items
items = [
build_ticker_item(ticker="AAPL", text="AAPL $185.50"),
build_ticker_item(ticker="GOOGL", text="GOOGL $142.20"),
build_ticker_item(ticker="MSFT", text="MSFT $378.90"),
]
Creating Marquee
create_widget(
html="<div id='content'>Dashboard Content</div>",
toolbars=[{
"position": "top",
"items": [{
"type": "marquee",
"ticker_items": [
{"ticker": "AAPL", "text": "AAPL $185.50"},
{"ticker": "GOOGL", "text": "GOOGL $142.20"},
],
"speed": 50, # pixels per second
"direction": "left", # left or right
}]
}]
)
Updating Individual Tickers
# Update one ticker without affecting others
update_ticker_item(
widget_id,
ticker="AAPL",
text="AAPL $186.25 ▲",
styles={"color": "#22c55e"} # Green for up
)
update_ticker_item(
widget_id,
ticker="GOOGL",
text="GOOGL $141.50 ▼",
styles={"color": "#ef4444"} # Red for down
)
Updating All Tickers
# Replace all ticker items
update_marquee(widget_id, ticker_items=[
{"ticker": "AAPL", "text": "AAPL $186.25", "styles": {"color": "#22c55e"}},
{"ticker": "GOOGL", "text": "GOOGL $141.50", "styles": {"color": "#ef4444"}},
{"ticker": "NEW", "text": "NEW $100.00"}, # Add new ticker
])
Live Data Patterns
Pattern 1: Polling
import time
widget = create_widget(html=initial_html, toolbars=[...])
while True:
# Fetch new data
data = fetch_latest_data()
# Update chart
fig = create_chart(data)
update_plotly(widget.id, figure_json=fig.to_json())
# Update tickers
for ticker, price in data["prices"].items():
update_ticker_item(widget.id, ticker=ticker, text=f"{ticker} ${price:.2f}")
time.sleep(5) # Poll every 5 seconds
Pattern 2: Event-Driven
widget = create_widget(html=initial_html, toolbars=[
{
"position": "top",
"items": [
{"type": "button", "label": "Refresh", "event": "refresh"},
{"type": "select", "label": "Timeframe", "event": "timeframe",
"options": [
{"label": "1D", "value": "1d"},
{"label": "1W", "value": "1w"},
{"label": "1M", "value": "1m"},
]},
]
}
])
while True:
events = get_events(widget.id, clear=True)
for e in events:
if e["event_type"] == "refresh":
data = fetch_data()
update_chart(widget.id, data)
elif e["event_type"] == "timeframe":
timeframe = e["data"]["value"]
data = fetch_data(timeframe=timeframe)
update_chart(widget.id, data)
time.sleep(0.1)
Pattern 3: Streaming via Marquee
# Marquee is ideal for streaming price data
# - Continuous scrolling
- Individual ticker updates
- Visual indication of changes
def stream_prices():
while True:
for ticker, price, change in get_price_updates():
color = "#22c55e" if change >= 0 else "#ef4444"
arrow = "▲" if change >= 0 else "▼"
update_ticker_item(
widget_id,
ticker=ticker,
text=f"{ticker} ${price:.2f} {arrow}",
styles={"color": color}
)
time.sleep(0.1) # High-frequency updates OK for ticker
Dashboard Layout
Multi-Chart Layout
content = Div(
content="""
<div style="display: grid; grid-template-columns: 1fr 1fr; gap: 16px; padding: 16px;">
<div id="chart1" style="height: 300px;"></div>
<div id="chart2" style="height: 300px;"></div>
<div id="chart3" style="height: 300px;"></div>
<div id="chart4" style="height: 300px;"></div>
</div>
""",
component_id="dashboard",
)
widget = create_widget(html=content.build_html(), height=700)
# Embed Plotly charts in each div
# (Use Plotly.js directly for embedded charts, or use separate widgets)
Chart + Table Layout
content = Div(
content="""
<div style="display: flex; height: 100%; padding: 16px; gap: 16px;">
<div id="chart" style="flex: 2;"></div>
<div id="table" style="flex: 1; overflow: auto;"></div>
</div>
""",
component_id="layout",
)
Performance Tips
For Charts
- Use
layout_only=Truewhen only changing titles, axes, annotations - Reduce data points for line charts (downsample)
- Use WebGL for large scatter plots (
fig = px.scatter(..., render_mode='webgl'))
For Tables
- AG Grid virtualizes automatically - send all data
- Use server-side pagination for very large datasets (100k+ rows)
- Avoid complex cell renderers for large tables
For Marquee
- Keep ticker count reasonable (10-20 items)
- Update only changed tickers, not all
- Use reasonable update frequency (not faster than visual perception ~100ms)