Getting Started¶
Create a project¶
This scaffolds:
app/with a minimalmain.pypythonnative.jsonproject configrequirements.txt.gitignore
A minimal app/main.py looks like:
import pythonnative as pn
Stack = pn.create_stack_navigator()
@pn.component
def HomeScreen():
nav = pn.use_navigation()
count, set_count = pn.use_state(0)
return pn.Column(
pn.Text(f"Count: {count}", style={"font_size": 24}),
pn.Button("Tap me", on_click=lambda: set_count(count + 1)),
pn.Button("Open details", on_click=lambda: nav.navigate("Detail", {"count": count})),
style={"spacing": 12, "padding": 16},
)
@pn.component
def DetailScreen():
route = pn.use_route()
return pn.Text(f"Count was {route.params.get('count', 0)}", style={"padding": 16})
@pn.component
def App():
return pn.NavigationContainer(
Stack.Navigator(
Stack.Screen("Home", HomeScreen, options={"title": "Home"}),
Stack.Screen("Detail", DetailScreen, options={"title": "Detail"}),
initial_route="Home",
)
)
Key ideas:
@pn.componentmarks a function as a PythonNative component. The function returns an element tree describing the UI. PythonNative creates and updates native views automatically.pn.use_state(initial)creates local component state. Call the setter to update it and the UI re-renders automatically.pn.create_stack_navigator()returns aStackwith.Navigatorand.Screenfactories. Wrap them inpn.NavigationContainerto enablepn.use_navigation()andpn.use_route()anywhere below.- The
Appfunction is the entry point. The Android and iOS templates importapp.main, look up its top-levelAppattribute, and start rendering. If you'd rather expose a differently-named component, configure your templates to load an explicit dotted path like"app.main.RootScreen". style={...}passes visual and layout properties as a dict (or list of dicts) to any component.- Element functions like
pn.Text(...),pn.Button(...),pn.Column(...)create lightweight descriptions, not native objects.
When the root Stack.Navigator is rendered inside the host's first screen, navigate(...) and go_back() drive the native navigation controller (UINavigationController on iOS, AndroidX Navigation Component on Android). Each pushed screen runs in its own reconciler host, so state on the previous screen is preserved by the platform stack.
Run on a platform¶
- Uses bundled templates (no network required for scaffolding)
- Copies your
app/into the generated project
If you just want to scaffold the platform project without building, use:
This stages files under build/ so you can open them in Android Studio or Xcode.
Hot reload while developing¶
For day-to-day UI work, run with --hot-reload:
The first run still builds and launches the native app. After that,
edits under app/ are copied into the running app's writable source
overlay and the active page refreshes without a full rebuild.
PythonNative prefers a Fast Refresh path: each
@pn.component function is matched by
qualified name across the reloaded module, the live VNode tree's
function references are swapped in place, and the next render reuses
the existing hook state. So edits to the body of a component preserve
in-memory state (counters, scroll positions, etc.). When Fast Refresh
cannot find a clean swap — for example, after deeper structural
edits — PythonNative falls back to a full remount of the active page
so you never get stuck with a stale tree.
This works best for Python UI changes; native template changes (Kotlin, Swift, manifests) still require a normal rebuild.
Viewing logs¶
After the app launches, pn run attaches to the app's stdout/stderr so Python
print() output and tracebacks stream back into your terminal until you press
Ctrl+C:
import pythonnative as pn
@pn.component
def App():
count, set_count = pn.use_state(0)
print(f"[App] render count={count}")
return pn.Column(
pn.Text(f"Count: {count}"),
pn.Button("Tap me", on_click=lambda: set_count(count + 1)),
)
- On Android, logs are streamed via
adb logcatfiltered to thepython.stdout/python.stderrtags (that Chaquopy redirectsprint()to) plus the template's Kotlin tags. - On iOS Simulator, the app is launched via
xcrun simctl launch --console-pty, which forwards the Python process's standard streams to your terminal.
Pass --no-logs if you'd rather run fire-and-forget:
Clean¶
Remove the build artifacts safely: