docs: spec language reference page + writing-specs rewrite (#46)

* docs(architecture): add device/emulator node and XCTest edge to diagram

* docs(manual): rewrite writing-specs with accurate API and updated examples

* docs(manual): add spec-language reference page

* docs(sidebar): add spec-language entry to sidebar nav
This commit is contained in:
pj authored and GitHub committed 2026-04-26 13:07:31 +07:00
1 parent 34fb73d6cb
commit 28909d954b
5 files changed
+427 -80

No files matched your search

+1
View File
@@ -6,5 +6,6 @@ title: Manual
- [Getting started](./getting-started/)
- [Writing specs](./writing-specs/)
- [Spec language reference](./spec-language/)
- [Runs](./runs/)
- [CLI reference](./cli/)
+237
View File
@@ -0,0 +1,237 @@
---
title: Spec language reference
---
# Spec language reference
## Module structure
A spec is a TypeScript module evaluated by the Go runner each step. It must export `properties` and `actionsRoot` on `globalThis` (the bundler entry point does this automatically via the final two lines):
```ts
import { ... } from "@sanderling/spec";
export const properties = { ... };
export const actionsRoot = weighted(...);
(globalThis as { actions?: unknown }).actions = actionsRoot;
(globalThis as { properties?: unknown }).properties = properties;
```
## State
Every extractor callback receives a `State`:
```ts
interface State {
ax: AccessibilityTree;
snapshots: Record<string, unknown>;
lastAction: Action | null;
logs: readonly LogEntry[];
exceptions: readonly ExceptionRecord[];
time: number; // ms since run start
}
```
| Field | Description |
|---|---|
| `ax` | Live UI hierarchy for this step |
| `snapshots` | Key-value data pushed by the app SDK (empty if SDK not integrated) |
| `lastAction` | The action dispatched in the previous step, or `null` on the first step |
| `logs` | Log entries collected since the previous step |
| `exceptions` | Uncaught exceptions or `Sanderling.reportError()` calls since the previous step |
| `time` | Milliseconds elapsed since the run started |
## Selectors
Selectors are passed to `ax.find()`, `ax.findAll()`, and element-scoped `.find()` / `.findAll()`.
### String selectors
| Form | Matches |
|---|---|
| `id:<value>` | Exact match on resource-id, or element whose resource-id ends with `:id/<value>` (Android) |
| `text:<value>` | Substring match on text content |
| `desc:<value>` | Exact match on accessibility description; also matches when description starts with `<value>, ` (iOS merged labels) |
| `descPrefix:<prefix>` | Starts-with match on accessibility description |
| `<attr>:<value>` | Substring match on any raw attribute by name |
Boolean attributes (`"true"` / `"false"`) use exact match rather than substring.
### Object selectors
Pass an object to apply multiple attribute filters with AND semantics:
```ts
s.ax.find({ accessibilityText: "LoginScreen" })
s.ax.find({ accessibilityText: "account_card", clickable: "true" })
```
Every key-value pair must match. Substring and boolean rules apply per attribute.
### Path queries
Chains of string selectors separated by ` > ` scope each segment to the subtree of the previous match. Path queries are only supported on the tree root (`ax.find`, `ax.findAll`), not on element-scoped `.find`/`.findAll`.
```ts
s.ax.find("id:HomeScreen > descPrefix:account_card:")
s.ax.find("id:LedgerScreen > desc:ledger_balance_display")
```
### Cross-platform aliases
These key aliases are resolved automatically so selectors work across platforms without changes:
| Write this | Also checks |
|---|---|
| `content-desc` | `accessibilityText` |
| `accessibilityText` | `content-desc` |
| `label` | `accessibilityText` |
| `accessibilityLabel` | `accessibilityText` |
| `identifier` | `resource-id` |
| `accessibilityIdentifier` | `resource-id` |
## AccessibilityElement fields
Fields available on every element returned by `find` / `findAll`:
| Field | Type | Description |
|---|---|---|
| `id` | `string` | resource-id (Android) or accessibility identifier (iOS) |
| `text` | `string` | Visible text content |
| `desc` | `string` | Accessibility description (`content-desc` / `accessibilityText`) |
| `class` | `string` | View class (Android), element type (iOS), or HTML tag (web) |
| `clickable` | `boolean` | Element is interactive |
| `enabled` | `boolean` | Element is enabled |
| `checked` | `boolean` | Checkbox or toggle state |
| `focused` | `boolean` | Element has input focus |
| `selected` | `boolean` | Selection state |
| `bounds` | `{ left, top, right, bottom }` | Bounding box in device pixels |
| `x` | `number` | Center X (derived from bounds) |
| `y` | `number` | Center Y (derived from bounds) |
| `attrs` | `Record<string, string>` | All raw attributes from the driver |
## Platform notes
### Android
- `id` maps to the Android resource-id (e.g., `com.example:id/button`). The `id:<value>` selector matches by suffix after `:id/`, so `id:button` matches `com.example:id/button`.
- `desc` maps to `content-desc`.
- `class` is the Java view class name (e.g., `android.widget.TextView`).
- `attrs` contains raw UIAutomator attributes: `package`, `scrollable`, `checkable`, etc.
### iOS
- `id` maps to the `accessibilityIdentifier` set via `.accessibilityIdentifier` in SwiftUI/UIKit.
- `desc` maps to `accessibilityText`, which Maestro builds by merging `accessibilityLabel` and the element's value (e.g., `"Close, icon description"`). The `desc:<value>` selector handles this by also matching when the description starts with `<value>, `.
- `class` is the XCUITest element type (e.g., `XCUIElementTypeButton`).
- `attrs` contains raw XCUITest attributes: `title`, `placeholderValue`, `hasFocus`, etc.
### Web (Chrome)
- `id` maps to the HTML `id` attribute.
- `desc` is derived from `aria-label`, `alt`, or `title`.
- `class` is the lowercase HTML tag name (e.g., `button`, `input`).
- `attrs` contains all HTML attributes available to CDP.
### KMP (Kotlin Multiplatform)
KMP apps are tested identically to native apps. An Android KMP build uses the Android driver; an iOS KMP build uses the iOS driver. There is no separate KMP driver. The accessibility tree structure reflects the target platform, so the same selector portability rules apply.
## Extractors
```ts
const loggedIn = extract((s) => !!s.ax.find("id:home-tab-bar"));
loggedIn.current // T - value from the current step
loggedIn.previous // T | undefined - value from the previous step, undefined on first step
```
Extractors are evaluated before properties and action generators. Use `.previous` to detect transitions between steps.
## LTL operators
| Function | Meaning |
|---|---|
| `always(f)` | `f` must hold at every step |
| `eventually(f).within(n, unit)` | `f` must hold at some step within `n` `"milliseconds"`, `"seconds"`, or `"steps"` |
| `now(f)` | `f` evaluated at the current step (for use inside `always`/`next` bodies) |
| `next(f)` | `f` evaluated at the step immediately after the current one |
**Formula combinators** - available on every `Formula`:
| Method | Meaning |
|---|---|
| `.implies(other)` | If `this` holds, `other` must also hold |
| `.and(other)` | Both must hold |
| `.or(other)` | At least one must hold |
| `.not()` | Negation |
## Actions
### Constructors
```ts
Tap({ on: element | string })
InputText({ into: element | string, text: string })
Swipe({ from: element | Point, to: element | Point, durationMillis?: number })
PressKey({ key: Key })
Wait({ durationMillis: number })
```
`Key` values: `"back"`, `"home"`, `"enter"`, `"tab"`, `"up"`, `"down"`, `"left"`, `"right"`.
On web, `"back"` maps to Backspace and `"home"` is not supported. All other keys work on all platforms.
### Built-in generators
| Generator | Behaviour |
|---|---|
| `taps` | Random tap on a clickable element |
| `swipes` | Random swipe gesture |
| `waitOnce` | Idles one step |
| `pressKey` | Presses a random supported key |
### `actions(generator)`
Wraps a callback that returns `Action[]`. The callback runs each step the generator is eligible.
```ts
const doLogin = actions(() => {
if (loggedIn.current) return [];
const submit = loginSubmit.current;
return submit ? [Tap({ on: submit })] : [];
});
```
### `weighted(...entries)`
Assembles a weighted tree. Each entry is `[weight, generator]`. Weights are relative within the tree.
```ts
export const actionsRoot = weighted(
[50, doLogin],
[10, taps],
[2, swipes],
);
```
### `from(items)`
Returns a `Sampler<T>` that cycles through a fixed list. Use `.generate()` to pick an item.
```ts
const names = from(["Checking", "Savings", "Travel"]);
// inside an actions() callback:
InputText({ into: nameField, text: names.generate() })
```
## Default properties
```ts
import { noUncaughtExceptions, noLogcatErrors } from "@sanderling/spec/defaults/properties";
```
| Property | Fails when |
|---|---|
| `noUncaughtExceptions` | An uncaught exception or `Sanderling.reportError()` call is captured |
| `noLogcatErrors` | Logcat emits any error-level (`E`) lines since the previous step |
+186 -80
View File
@@ -9,16 +9,13 @@ A spec has three parts: extractors, properties, and actions.
```ts
import { extract, always, now, actions, weighted, Tap, taps, swipes } from "@sanderling/spec";
// 1. Extractors pull values from each observed state.
const loggedIn = extract((s) => !!s.ax.find("id:home-tab-bar"));
// 2. Properties are LTL formulas evaluated every step.
export const properties = {
cartNeverNegative: always(() => cartCount.current >= 0),
};
// 3. Actions are a weighted tree of what sanderling is allowed to do.
export const actions = weighted(
export const actionsRoot = weighted(
[10, taps],
[2, swipes],
);
@@ -28,84 +25,207 @@ The Go runner calls into the JS runtime each step. Extractors re-read the curren
## The `State` object
What extractors see:
What extractors receive:
```ts
interface State {
ax: AccessibilityTree; // view hierarchy
screen: { id: string; hash: string };
ax: AccessibilityTree;
snapshots: Record<string, unknown>;
lastAction: Action | null;
logs: LogEntry[]; // since previous state
exceptions: Exception[];
time: number; // ms since run start
logs: readonly LogEntry[];
exceptions: readonly ExceptionRecord[];
time: number; // ms since run start
}
```
`ax.find("text:Click me")`, `ax.find("id:login-form")`, `ax.findAll("role:todo-row")` are the common accessors. Prefer stable testID-style identifiers over positional selectors, for the same reason you would in Espresso or XCUITest.
`ax` is the live UI hierarchy. `snapshots` carries any key-value data pushed by the app SDK. `logs` and `exceptions` contain entries collected since the previous step.
## Pattern: preconditions (login, onboarding)
## Extractors
sanderling has no setup phase and no fixtures. Preconditions are action generators with two properties:
`extract()` wraps a getter that runs against every new state. The returned object exposes `.current` (this step's value) and `.previous` (last step's, or `undefined` on the first step).
1. High weight, so they fire whenever applicable.
2. Gated on a state extractor, so they return an empty tree when not applicable and self-disable once the precondition is met.
```ts
const loggedIn = extract((s) => !!s.ax.find("id:home-tab-bar"));
const balance = extract<number>((s) => s.snapshots["account.balance"] as number ?? 0);
// Inside a property or action:
loggedIn.current // boolean
loggedIn.previous // boolean | undefined
```
Extractors are cheap. Prefer one extractor per concept and reuse it across properties and action generators.
## Finding elements
`ax.find(selector)` returns the first matching `AccessibilityElement`, or `undefined`. `ax.findAll(selector)` returns all matches. Both are available on the tree root and on any element (scoped to its subtree).
**String selectors:**
| Form | Match rule |
|---|---|
| `id:<value>` | Exact match on resource-id, or suffix after `:id/` (Android) |
| `text:<value>` | Substring match on text content |
| `desc:<value>` | Exact match on accessibility description, or starts-with for iOS merged labels |
| `descPrefix:<prefix>` | Starts-with on accessibility description |
| `<attr>:<value>` | Substring match on any raw attribute by name |
**Object selectors** (AND of all given attributes):
```ts
s.ax.find({ accessibilityText: "LoginScreen" })
s.ax.find({ accessibilityText: "login_email" })
```
**Path queries** (global only):
```ts
s.ax.find("id:HomeScreen > descPrefix:account_card:")
```
Each segment is matched within the subtree of the previous match.
**Cross-platform aliases** are resolved automatically. `label` and `accessibilityLabel` both resolve to `accessibilityText`; `content-desc` and `accessibilityText` are interchangeable; `identifier` and `accessibilityIdentifier` resolve to `resource-id`.
See the [Spec language reference](./spec-language/) for the complete selector grammar and per-platform field availability.
## Properties
Properties are named LTL formulas exported from the spec. The verifier evaluates each one every step and fails the run when a formula is violated.
```ts
export const properties = {
balanceNeverNegative: always(() => balance.current >= 0),
loginReachable: eventually(() => loggedIn.current).within(30, "seconds"),
};
```
**Operators:**
- `always(f)` - `f` must hold at every step.
- `eventually(f).within(n, unit)` - `f` must hold at some step within `n` milliseconds, seconds, or steps.
- `now(f)` - evaluates `f` at the current step (used for implication antecedents).
- `next(f)` - evaluates `f` at the next step.
**Combinators** - available on any formula:
```ts
now(() => loggedIn.current).implies(now(() => cartCount.current !== undefined))
formulaA.and(formulaB)
formulaA.or(formulaB)
formulaA.not()
```
`implies`, `and`, `or`, and `not` compose freely.
## Actions
Action generators return a list of actions to perform. The runner samples one from the weighted tree and dispatches it through the driver.
**Built-in generators** (pass directly to `weighted`):
- `taps` - autonomous random taps on clickable elements.
- `swipes` - autonomous random swipe gestures.
- `waitOnce` - idles one step.
- `pressKey` - presses a random supported key.
**Action constructors:**
```ts
Tap({ on: element }) // tap an element or selector string
InputText({ into: element, text: "hello" }) // clear and type into a field
Swipe({ from: elementOrPoint, to: elementOrPoint, durationMillis?: number })
PressKey({ key: "back" | "home" | "enter" | "tab" | "up" | "down" | "left" | "right" })
Wait({ durationMillis: number })
```
**Samplers** - cycle over a fixed list:
```ts
const names = from(["Checking", "Savings", "Travel"]);
names.generate() // picks from the list
```
**Custom generators:**
```ts
const doLogin = actions(() => {
if (loggedIn.current) return [];
const emailField = loginEmail.current;
const submit = loginSubmit.current;
if (!emailField || !submit) return [];
return [InputText({ into: emailField, text: "[email protected]" }), Tap({ on: submit })];
});
```
**Weighted trees:**
```ts
export const actionsRoot = weighted(
[100, dismissOnboarding],
[50, doLogin],
[10, taps],
[2, swipes],
[1, weighted(
[3, openDeepLink("app://home")],
[1, openDeepLink("app://settings")],
)],
);
```
Weights are relative within each tree. Nested trees get their own local budget.
## Default properties
`@sanderling/spec/defaults/properties` exports ready-made properties:
```ts
import { noUncaughtExceptions, noLogcatErrors } from "@sanderling/spec/defaults/properties";
export const properties = {
noUncaughtExceptions, // fails if the app throws an uncaught exception
noLogcatErrors, // fails if logcat emits any error-level lines
};
```
## Pattern: preconditions
sanderling has no setup phase. Preconditions are action generators with high weight that self-disable once the condition is satisfied.
```ts
const onLoginScreen = extract((s) => !!s.ax.find("id:login-form"));
const loginEmailField = extract((s) => s.ax.find("id:email-field"));
const loginSubmit = extract((s) => s.ax.find("id:sign-in-button"));
const doLogin = actions(() => {
if (!onLoginScreen.current) return [];
const emailField = state.ax.find("id:email-field");
const signInButton = state.ax.find("id:sign-in-button");
if (!emailField || !signInButton) return [];
const email = loginEmailField.current;
const submit = loginSubmit.current;
if (!email || !submit) return [];
return [
InputText({ into: emailField, text: "[email protected]" }),
Tap({ on: signInButton }),
InputText({ into: email, text: "[email protected]" }),
Tap({ on: submit }),
];
});
```
Stack these for onboarding, consent dialogs, cold-start flows:
Stack these for onboarding, consent dialogs, and cold-start flows:
```ts
const dismissOnboarding = actions(() => {
const skip = state.ax.find("text:Skip");
return skip ? [Tap({ on: skip })] : [];
});
export const actions = weighted(
[100, dismissOnboarding], // clear the path first
[50, doLogin], // log in when the login screen appears
[10, taps], // exploration
export const actionsRoot = weighted(
[100, dismissOnboarding],
[50, doLogin],
[10, taps],
[2, swipes],
);
```
Lifecycle of a run:
```
Step 1: fresh install, onboarding visible
eligible: dismissOnboarding (weight 100)
picks: Tap "Skip"
Step 2-3: login screen visible
eligible: doLogin (weight 50)
picks: InputText / Tap to sign in
Step 4+: home screen, onboarding and login generators return []
eligible: taps, swipes
picks: autonomous exploration
```
Session state (tokens, keychain, prefs) persists through the rest of the run. If the app logs the user out mid-run, `doLogin` re-fires automatically. No retry logic, no special-casing.
Once `onLoginScreen.current` is false, `doLogin` returns `[]` and drops out of the eligible set automatically.
## Pattern: conditional properties
Use gating extractors the same way inside properties. Express "only check X when Y holds" with `now(...).implies(...)`:
Gate a property so it only applies when a precondition holds:
```ts
const loggedIn = extract((s) => !!s.ax.find("id:home-tab-bar"));
export const properties = {
cartPersistsWhenLoggedIn: always(
now(() => loggedIn.current).implies(now(() => cartCount.current !== undefined)),
@@ -113,44 +233,30 @@ export const properties = {
};
```
`implies`, `and`, `or`, and `not` are methods on any formula. Combine them freely.
## Pattern: step-to-step invariants
## Pattern: eventually
`always` asserts something holds at every step. `eventually` asserts it holds at some step, usually with a time bound:
Use `next()` to express invariants that span two consecutive steps:
```ts
loginSucceedsWithin30s: eventually(() => loggedIn.current).within(30, "seconds"),
```
`within` takes `"milliseconds"`, `"seconds"`, or `"steps"`. Useful for liveness checks: the loading spinner eventually goes away, the deep link eventually lands on `/home`.
## Pattern: weighted exploration sub-trees
Nest `weighted` to group related actions and tune their collective rate:
```ts
export const actions = weighted(
[100, dismissOnboarding],
[50, doLogin],
[10, taps],
[2, swipes],
[1, weighted(
[3, openLink("todos://home")],
[1, openLink("todos://settings")],
[1, openLink("todos://item/42/edit")],
)],
const newAccountBalanceIsZero = always(
next(() => {
const prev = accounts.previous ?? [];
const curr = accounts.current;
if (prev.length === 0 || curr.length === 0) return true;
const prevIds = new Set(prev.map((a) => a.id));
return curr.filter((a) => !prevIds.has(a.id)).every((a) => a.balance === 0);
}),
);
```
Weights are relative within a tree, so nested trees get their own local budget. This is how you keep low-frequency but high-value actions (deep links, background/foreground, rotate) from drowning out normal tapping.
## Anti-patterns
**Positional taps.** `Tap({ on: { x: 100, y: 200 } })` works for a demo but breaks on any layout change. Always prefer an `ax.find("id:...")` reference.
**Accessing `state` outside of `extract`.** The `state` argument exists only inside the `extract()` callback. Use extractors and `.current` everywhere else.
**Sleep or wait-for-time.** `Wait(3000)` inside an action generator is a smell. If you need to wait for a condition, use an extractor and gate the next action on it.
**Positional taps.** `Tap({ on: { x: 100, y: 200 } })` breaks on any layout change. Always prefer an `ax.find("id:...")` reference.
**Retry logic inside generators.** Generators should be pure: given the same state they produce the same actions. Retry is the runner's responsibility.
**Unbounded `eventually`.** Without `.within(...)`, `eventually` never fails within a finite run. Almost always you want a bound.
**Unbounded `eventually`.** Without a `.within(...)`, `eventually` never fails within a finite run. It just stays residual. Almost always you want a bound.
**`Wait()` inside generators.** Waiting for a condition belongs in an extractor guard, not inside a generator.
**Retry logic inside generators.** Generators must be pure. Given the same state they produce the same actions. Retry is the runner's responsibility.