refactor: converge Maestro input handling

Author: thymikee

src/compat/maestro/__tests__/replay-flow.test.ts

@@ -223,7 +223,7 @@ test('parseMaestroReplayFlow keeps focused inputText and pressKey Enter as separ
   assert.deepEqual(parsed.actionLines, [3, 4, 5]);
 });
 
-test('parseMaestroReplayFlow marks tapOn before inputText for snapshot tap focus', () => {
+test('parseMaestroReplayFlow coalesces tapOn inputText through native fill', () => {
   const parsed = parseMaestroReplayFlow(`appId: com.callstack.agentdevicelab
 ---
 - tapOn:
@@ -234,11 +234,12 @@ test('parseMaestroReplayFlow marks tapOn before inputText for snapshot tap focus
   assert.deepEqual(
     parsed.actions.map((entry) => [entry.command, entry.positionals]),
     [
-      ['__maestroTapOn', ['id="editableNameInput"']],
-      ['type', ['Saved list']],
+      ['wait', ['id="editableNameInput"', '30000']],
+      ['fill', ['id="editableNameInput"', 'Saved list']],
     ],
   );
-  assert.equal(parsed.actions[0]?.flags?.maestro?.allowNonHittableCoordinateFallback, undefined);
+  assert.deepEqual(parsed.actionLines, [3, 3]);
+  assert.equal(parsed.actions[1]?.flags?.maestro?.allowNonHittableCoordinateFallback, true);
 });
 
 test('parseMaestroReplayFlow coalesces tapOn inputText while preserving pressKey Enter submit', () => {
@@ -281,6 +282,27 @@ test('parseMaestroReplayFlow does not coalesce text entry for non-input-looking
   assert.equal(parsed.actions[0]?.flags?.maestro?.allowNonHittableCoordinateFallback, undefined);
 });
 
+test('parseMaestroReplayFlow maps focused input commands to native type and keyboard actions', () => {
+  const parsed = parseMaestroReplayFlow(`appId: com.callstack.agentdevicelab
+---
+- inputText: hello
+- eraseText:
+    charactersToErase: 3
+- pasteText: pasted
+- pressKey: Return
+`);
+
+  assert.deepEqual(
+    parsed.actions.map((entry) => [entry.command, entry.positionals]),
+    [
+      ['type', ['hello']],
+      ['type', ['\b'.repeat(3)]],
+      ['type', ['pasted']],
+      ['__maestroPressEnter', []],
+    ],
+  );
+});
+
 test('parseMaestroReplayFlow rejects relative runScript paths without source path', () => {
   assert.throws(
     () =>
@@ -658,10 +680,10 @@ test('parseMaestroReplayFlow parses the test-app Maestro suite fixture', () => {
       '__maestroAssertVisible',
       '__maestroTapOn',
       '__maestroAssertVisible',
-      '__maestroTapOn',
-      'type',
-      '__maestroTapOn',
-      'type',
+      'wait',
+      'fill',
+      'wait',
+      'fill',
       '__maestroTapOn',
       '__maestroAssertVisible',
       '__maestroAssertVisible',

src/compat/maestro/replay-flow.ts

@@ -114,9 +114,7 @@ function optimizeTypedAfterTap(
     return { actions: [clearMaestroNonHittableTap(action)], actionLines: [line], consumed: 1 };
   }
   const pressEnterAction = actions[index + 2];
-  if (pressEnterAction?.command !== MAESTRO_RUNTIME_COMMAND.pressEnter) {
-    return { actions: [clearMaestroNonHittableTap(action)], actionLines: [line], consumed: 1 };
-  }
+  const shouldKeepEnter = pressEnterAction?.command === MAESTRO_RUNTIME_COMMAND.pressEnter;
   return {
     actions: [
       {
@@ -130,10 +128,10 @@ function optimizeTypedAfterTap(
         positionals: [tapSelector, typedAfterTap],
         flags: action.flags,
       },
-      pressEnterAction,
+      ...(shouldKeepEnter ? [pressEnterAction] : []),
     ],
-    actionLines: [line, line, actionLines[index + 2] ?? line],
-    consumed: 3,
+    actionLines: [line, line, ...(shouldKeepEnter ? [actionLines[index + 2] ?? line] : [])],
+    consumed: shouldKeepEnter ? 3 : 2,
   };
 }
 

src/daemon/handlers/__tests__/session-replay-vars.test.ts

@@ -27,7 +27,7 @@ const LOC = { file: 'test.ad', line: 1 };
 type CapturedInvocation = {
   command: string;
   positionals?: string[];
-  flags?: Record<string, unknown>;
+  flags?: CommandFlags;
 };
 
 async function runReplayFixture(params: {
@@ -1247,7 +1247,7 @@ test('runReplayScriptFile lets snapshot id tap handle Maestro one-point edge con
   );
 });
 
-test('runReplayScriptFile keeps Maestro text-entry tapOn on the snapshot path', async () => {
+test('runReplayScriptFile coalesces Maestro text-entry tapOn into native fill', async () => {
   const calls: CapturedInvocation[] = [];
   const { response } = await runReplayFixture({
     label: 'maestro-tap-input-text-snapshot',
@@ -1284,12 +1284,11 @@ test('runReplayScriptFile keeps Maestro text-entry tapOn on the snapshot path',
   assert.deepEqual(
     calls.map((call) => [call.command, call.positionals]),
     [
-      ['snapshot', []],
-      ['click', ['120', '120']],
-      ['type', ['Saved list']],
+      ['wait', ['id="editableNameInput"', '30000']],
+      ['fill', ['id="editableNameInput"', 'Saved list']],
     ],
   );
-  assert.equal(calls[0]?.flags?.noRecord, true);
+  assert.equal(calls[1]?.flags?.maestro?.allowNonHittableCoordinateFallback, true);
 });
 
 test('runReplayScriptFile resolves Maestro swipe.label from a labeled element rect', async () => {

test/integration/provider-scenarios/android-test-suite.test.ts

@@ -126,6 +126,62 @@ test('Provider-backed integration Android Maestro replay uses fresh selector sna
   );
 });
 
+test('Provider-backed integration Android Maestro coalesces tapOn inputText and pressKey Enter through native paths', async () => {
+  await withProviderScenarioResource(
+    async () => await createAndroidSettingsWorld({ nativeTextInjection: true }),
+    async (world) => {
+      const client = world.daemon.client();
+      const suiteRoot = path.join(world.tempRoot, 'suite-maestro-input');
+      fs.mkdirSync(suiteRoot, { recursive: true });
+      const flowPath = path.join(suiteRoot, 'input-submit.yaml');
+      fs.writeFileSync(
+        flowPath,
+        [
+          'appId: com.android.settings',
+          '---',
+          '- launchApp',
+          '- tapOn: Search',
+          '- inputText: "Łódź café"',
+          '- pressKey: Enter',
+          '',
+        ].join('\n'),
+      );
+
+      const suite = await client.replay.test({
+        paths: [flowPath],
+        backend: 'maestro',
+        artifactsDir: path.join(suiteRoot, 'artifacts'),
+        timeoutMs: 30000,
+        ...world.selection,
+      });
+
+      assert.equal(suite.total, 1, JSON.stringify(suite));
+      assert.equal(suite.passed, 1, JSON.stringify(suite));
+      assert.equal(suite.failed, 0, JSON.stringify(suite));
+      assert.deepEqual(world.textInjectionCalls, [
+        {
+          action: 'fill',
+          target: { x: 195, y: 52 },
+          text: 'Łódź café',
+          delayMs: 0,
+        },
+      ]);
+      assert.equal(
+        world.adbCalls.some(
+          (call) => call[0] === 'shell' && call[1] === 'input' && call[2] === 'text',
+        ),
+        false,
+        JSON.stringify(world.adbCalls),
+      );
+      assert.deepEqual(
+        world.adbCalls.find((call) => call.slice(0, 4).join(' ') === 'shell input keyevent ENTER'),
+        ['shell', 'input', 'keyevent', 'ENTER'],
+      );
+      world.assertNoHostAdbCalls();
+    },
+  );
+});
+
 function androidMaestroReplayXml(searchBounds: string): string {
   return [
     '<?xml version="1.0" encoding="UTF-8"?>',