2258 Control Flow 786bf81a

Control Flow API Reference

Control cell execution flow with conditional stops, gating, and threading.

Conditional Execution

mo.stop

Halt cell execution based on a condition.

import marimo as mo

# Stop if condition is true
mo.stop(data is None)

# Stop with message
mo.stop(
    not data_loaded,
    mo.md("**Loading data...** Please wait.")
)

# Stop with UI element
mo.stop(
    form.value is None,
    mo.md("**Submit the form to continue.**")
)

When mo.stop() triggers:

1. A MarimoStopError is raised
2. Cell execution halts
3. The optional output is displayed
4. Downstream cells don't run (their definitions are removed)

Common Patterns

Gate on button click:

run_button = mo.ui.run_button(label="Run Analysis")

# In another cell
mo.stop(not run_button.value, mo.md("Click **Run Analysis** to proceed."))

# Expensive computation runs only after button click
result = expensive_analysis(data)

Gate on form submission:

form = mo.ui.text(label="Enter query").form()

# In another cell
mo.stop(form.value is None, mo.md("Submit a query to search."))

results = search(form.value)

Gate on data availability:

mo.stop(
    df is None or len(df) == 0,
    mo.callout("No data available. Upload a file to continue.", kind="warn")
)

# Process data
summary = df.describe()

Multiple conditions:

errors = []
if not username:
    errors.append("Username required")
if not password:
    errors.append("Password required")

mo.stop(
    len(errors) > 0,
    mo.callout(mo.md("\n".join(f"- {e}" for e in errors)), kind="danger")
)

Periodic Execution

mo.ui.refresh

Trigger cell re-execution at intervals.

# Auto-refresh every 5 seconds
refresh = mo.ui.refresh(interval=5)

# With options
refresh = mo.ui.refresh(
    options=["1s", "5s", "10s", "30s", "off"],
    default_interval="5s"
)

# In another cell - reference refresh to trigger updates
_ = refresh.value
current_time = datetime.now()
latest_data = fetch_data()

Runtime Configuration

Lazy Execution Mode

Configure via notebook settings to mark cells as stale instead of auto-running.

In the editor:

1. Open settings (gear icon)
2. Enable "Lazy execution"
3. Cells show "stale" indicator when dependencies change
4. Click to run manually

Useful for:

• Expensive computations
• Debugging
• Step-by-step execution

Disable Autorun

In notebook settings:

• Disable "Autorun on startup"
• Disable "Autorun on cell execution"

Threading

mo.Thread

Extended threading with frontend communication.

import marimo as mo

def worker():
    thread = mo.current_thread()
    for i in range(100):
        if thread.should_exit:
            # Cleanup and exit
            return
        # Do work
        mo.output.append(f"Progress: {i}%")

thread = mo.Thread(target=worker)
thread.start()

Properties:

• should_exit: Returns True when cell becomes invalid (re-run, deleted, interrupted)

Features:

• mo.output.append() and print() forward to frontend
• Must implement cleanup when should_exit is True

mo.current_thread

Get the current marimo thread context.

def worker():
    thread = mo.current_thread()
    while not thread.should_exit:
        # Process items
        pass

# Raises RuntimeError if not in a marimo thread

Error Handling

try:
    result = risky_operation()
except Exception as e:
    mo.stop(True, mo.callout(f"Error: {e}", kind="danger"))

# Continue with result
process(result)

Combining Control Flow

# Complex gating example
file = mo.ui.file(filetypes=[".csv"])
run_btn = mo.ui.run_button(label="Analyze")

mo.vstack([file, run_btn])

# In processing cell
mo.stop(file.value is None, mo.md("Upload a CSV file"))
mo.stop(not run_btn.value, mo.md("Click **Analyze** to process"))

# Both conditions met - process
df = pd.read_csv(io.BytesIO(file.value[0].contents))
analysis = analyze(df)

Best Practices

1. Use mo.stop() for prerequisites: Gate expensive operations until dependencies are ready
2. Provide clear feedback: Always include informative output in mo.stop()
3. Use run_button for expensive ops: Don't auto-run costly computations
4. Handle thread cleanup: Always check should_exit in long-running threads
5. Prefer reactivity over manual control: Let marimo's DAG handle most execution flow