perf(ios): reuse current focus for native text replacement

apple/runner/AgentDeviceRunner/AgentDeviceRunner/AgentDeviceRunnerAppFixtures.m

@@ -11,6 +11,7 @@
 #import "AgentDeviceRunnerApp.h"
 #import <UIKit/UIKit.h>
 #import <UserNotifications/UserNotifications.h>
+#import "UnitTests/AgentDeviceTextEntryFields.h"
 
 // Mirrors the App Tracking Transparency prompt's real button order: the affirmative button is
 // last (#3627). The answer label records which button the app actually took, so a test asserts
@@ -35,26 +36,6 @@
   return alert;
 }
 
-// A plain text field whose accessibility value is the masked rendering of its text. This is how
-// a React Native `secureTextEntry` input reaches XCTest: `.textField`, not `.secureTextField`,
-// with a value of bullets (tester-army/e2e#872).
-@interface AgentDeviceMaskedTextField : UITextField
-@end
-
-@implementation AgentDeviceMaskedTextField
-
-- (NSString *)accessibilityValue {
-  NSUInteger length = self.text.length;
-  if (length == 0) {
-    return [super accessibilityValue];
-  }
-  // Half the typed length, as observed in the report: the mask hides the count as well.
-  NSUInteger bullets = MAX((NSUInteger)1, length / 2);
-  return [@"" stringByPaddingToLength:bullets withString:@"\u2022" startingAtIndex:0];
-}
-
-@end
-
 @interface AgentDeviceRunnerFixtureViewController : AgentDeviceRunnerViewController
 @end
 
@@ -105,39 +86,6 @@ - (NSString *)accessibilityValue {
 }
 @end
 
-// XCTest reads every element's accessibility value each time it snapshots the app, which it does
-// for every element query and every attribute read from a live element. Counting this field's
-// value reads therefore counts the accessibility queries a command makes. The value is masked, so a
-// fill's read-back ends at its first read instead of polling through a stability window whose
-// poll count depends on how fast the host answers.
-static NSUInteger AgentDeviceTextEntryValueReads = 0;
-
-@interface AgentDeviceValueReadCountingTextField : AgentDeviceMaskedTextField
-@end
-
-@implementation AgentDeviceValueReadCountingTextField
-- (NSString *)accessibilityValue {
-  AgentDeviceTextEntryValueReads += 1;
-  return [super accessibilityValue];
-}
-@end
-
-// Reports the read count as its accessibility label and the counted field's text as its value, so
-// reading either changes no view state and the masked text stays checkable.
-@interface AgentDeviceValueReadCountLabel : UILabel
-@property(nonatomic, weak) UITextField *countedField;
-@end
-
-@implementation AgentDeviceValueReadCountLabel
-- (NSString *)accessibilityLabel {
-  return [NSString stringWithFormat:@"%lu", (unsigned long)AgentDeviceTextEntryValueReads];
-}
-
-- (NSString *)accessibilityValue {
-  return self.countedField.text;
-}
-@end
-
 // A field that reformats its own text on every edit into `## ### ## ##` digit groups, the way a
 // digit-grouping formatter in a Flutter or React Native input does. The accessibility value is
 // what the user sees — the formatted string — so it never equals the digits a `fill` sent (#2634).
@@ -712,6 +660,8 @@ - (void)viewDidLoad {
         [NSProcessInfo.processInfo.arguments containsObject:@"--agent-device-text-entry-masked-value"];
     BOOL countsValueReads =
         [NSProcessInfo.processInfo.arguments containsObject:@"--agent-device-text-entry-counts-value-reads"];
+    BOOL witnessesReplacement =
+        [NSProcessInfo.processInfo.arguments containsObject:@"--agent-device-text-entry-witness"];
     UITextField *textField;
     if (digitCountValue) {
       textField = [[AgentDeviceDigitCountTextField alloc] init];
@@ -724,6 +674,8 @@ - (void)viewDidLoad {
       textField = [[AgentDeviceSubmitClearingTextField alloc] init];
     } else if (maskedValue) {
       textField = [[AgentDeviceMaskedTextField alloc] init];
+    } else if (witnessesReplacement) {
+      textField = [[AgentDeviceTextEntryWitnessField alloc] init];
     } else if (countsValueReads) {
       textField = [[AgentDeviceValueReadCountingTextField alloc] init];
     } else {
@@ -761,7 +713,7 @@ - (void)viewDidLoad {
       [textField.widthAnchor constraintEqualToConstant:240],
       [textField.heightAnchor constraintEqualToConstant:AgentDeviceTextEntryFieldHeight],
     ]];
-    if (countsValueReads) {
+    if (countsValueReads || witnessesReplacement) {
       AgentDeviceValueReadCountLabel *valueReads = [[AgentDeviceValueReadCountLabel alloc] init];
       valueReads.countedField = textField;
       valueReads.accessibilityIdentifier = @"agent-device-text-entry-value-reads";

apple/runner/AgentDeviceRunner/AgentDeviceRunner/UnitTests/AgentDeviceTextEntryFields.h

@@ -0,0 +1,18 @@
+#import <TargetConditionals.h>
+
+#if TARGET_OS_IOS && defined(AGENT_DEVICE_RUNNER_UNIT_TESTS)
+#import <UIKit/UIKit.h>
+
+@interface AgentDeviceMaskedTextField : UITextField
+@end
+
+@interface AgentDeviceValueReadCountingTextField : AgentDeviceMaskedTextField
+@end
+
+@interface AgentDeviceValueReadCountLabel : UILabel
+@property(nonatomic, weak) UITextField *countedField;
+@end
+
+@interface AgentDeviceTextEntryWitnessField : UITextField <UITextFieldDelegate, UIGestureRecognizerDelegate>
+@end
+#endif

apple/runner/AgentDeviceRunner/AgentDeviceRunner/UnitTests/AgentDeviceTextEntryFields.m

@@ -0,0 +1,99 @@
+#import "AgentDeviceTextEntryFields.h"
+
+#if TARGET_OS_IOS && defined(AGENT_DEVICE_RUNNER_UNIT_TESTS)
+@implementation AgentDeviceMaskedTextField
+- (NSString *)accessibilityValue {
+  NSUInteger length = self.text.length;
+  if (length == 0) {
+    return [super accessibilityValue];
+  }
+  NSUInteger bullets = MAX((NSUInteger)1, length / 2);
+  return [@"" stringByPaddingToLength:bullets withString:@"\u2022" startingAtIndex:0];
+}
+@end
+
+static NSUInteger AgentDeviceTextEntryValueReads = 0;
+
+@implementation AgentDeviceValueReadCountingTextField
+- (NSString *)accessibilityValue {
+  AgentDeviceTextEntryValueReads += 1;
+  return [super accessibilityValue];
+}
+@end
+
+@implementation AgentDeviceValueReadCountLabel
+- (NSString *)accessibilityLabel {
+  if ([self.countedField isKindOfClass:[AgentDeviceTextEntryWitnessField class]]) {
+    return self.countedField.accessibilityHint;
+  }
+  return [NSString stringWithFormat:@"%lu", (unsigned long)AgentDeviceTextEntryValueReads];
+}
+
+- (NSString *)accessibilityValue {
+  return self.countedField.text;
+}
+
+@end
+
+@interface AgentDeviceTextEntryWitnessField ()
+@property(nonatomic, assign) NSUInteger taps;
+@property(nonatomic, assign) NSUInteger selections;
+@property(nonatomic, assign) NSUInteger rejectedEdits;
+@property(nonatomic, assign) BOOL rejectNextSelectionDeletion;
+@end
+
+@implementation AgentDeviceTextEntryWitnessField
+- (instancetype)init {
+  self = [super init];
+  if (self) {
+    self.delegate = self;
+    self.autocorrectionType = UITextAutocorrectionTypeNo;
+    self.autocapitalizationType = UITextAutocapitalizationTypeNone;
+    UITapGestureRecognizer *tap = [[UITapGestureRecognizer alloc] initWithTarget:self action:@selector(recordTap:)];
+    tap.cancelsTouchesInView = NO;
+    tap.delegate = self;
+    [self addGestureRecognizer:tap];
+    if ([NSProcessInfo.processInfo.arguments containsObject:@"--agent-device-text-entry-witness-prefill"]) {
+      self.text = @"seedvalue";
+    }
+  }
+  return self;
+}
+
+- (void)recordTap:(UITapGestureRecognizer *)recognizer {
+  (void)recognizer;
+  self.taps += 1;
+}
+
+- (BOOL)gestureRecognizer:(UIGestureRecognizer *)gestureRecognizer shouldRecognizeSimultaneouslyWithGestureRecognizer:(UIGestureRecognizer *)otherGestureRecognizer {
+  (void)gestureRecognizer;
+  (void)otherGestureRecognizer;
+  return YES;
+}
+
+- (void)selectAll:(id)sender {
+  self.selections += 1;
+  self.rejectNextSelectionDeletion =
+      [NSProcessInfo.processInfo.arguments containsObject:@"--agent-device-text-entry-witness-refuse-selection"];
+  [super selectAll:sender];
+}
+
+- (BOOL)textField:(UITextField *)textField shouldChangeCharactersInRange:(NSRange)range replacementString:(NSString *)string {
+  NSArray<NSString *> *arguments = NSProcessInfo.processInfo.arguments;
+  if ([arguments containsObject:@"--agent-device-text-entry-witness-refuse-edits"] ||
+      (self.rejectNextSelectionDeletion && range.length > 1 && string.length == 0)) {
+    self.rejectNextSelectionDeletion = NO;
+    self.rejectedEdits += 1;
+    UITextPosition *start = textField.beginningOfDocument;
+    textField.selectedTextRange = [textField textRangeFromPosition:start toPosition:start];
+    return NO;
+  }
+  return YES;
+}
+
+- (NSString *)accessibilityHint {
+  return [NSString stringWithFormat:@"taps=%lu selections=%lu rejected=%lu",
+                   (unsigned long)self.taps, (unsigned long)self.selections, (unsigned long)self.rejectedEdits];
+}
+@end
+#endif

apple/runner/AgentDeviceRunner/AgentDeviceRunnerUITests/RunnerTests+TextEntry.swift

@@ -161,6 +161,7 @@ extension RunnerTests {
     let prefersFocusedElement: Bool
     let contentTranslation: TextEntryContentTranslation
     let fromTapWitness: Bool
+    let reusesCurrentFocus: Bool
     /// The input the first resolved element was. Once bound, resolution refuses any other input.
     let boundIdentity: TextEntryInputIdentity?
     /// The text input that sat under `refreshPoint` before the focus tap, which binds the target to
@@ -179,6 +180,7 @@ extension RunnerTests {
       prefersFocusedElement: Bool,
       contentTranslation: TextEntryContentTranslation,
       fromTapWitness: Bool = false,
+      reusesCurrentFocus: Bool = false,
       boundIdentity: TextEntryInputIdentity? = nil,
       inputAtRefreshPoint: TextInputAtPoint? = nil,
       postFocusRead: ResolvedTextEntryInput? = nil
@@ -188,6 +190,7 @@ extension RunnerTests {
       self.prefersFocusedElement = prefersFocusedElement
       self.contentTranslation = contentTranslation
       self.fromTapWitness = fromTapWitness
+      self.reusesCurrentFocus = reusesCurrentFocus
       self.boundIdentity = boundIdentity ?? inputAtRefreshPoint?.identity
       self.inputAtRefreshPoint = inputAtRefreshPoint
       self.postFocusRead = postFocusRead
@@ -206,6 +209,7 @@ extension RunnerTests {
         prefersFocusedElement: prefersFocusedElement,
         contentTranslation: contentTranslation,
         fromTapWitness: fromTapWitness,
+        reusesCurrentFocus: reusesCurrentFocus,
         boundIdentity: boundIdentity ?? input.identity,
         inputAtRefreshPoint: inputAtRefreshPoint
       )
@@ -411,11 +415,13 @@ extension RunnerTests {
     if existing?.isEmpty == true {
       return
     }
-    // Read before the caret tap, which dispatches to the app.
+    let defersCaretTap = reader.target.reusesCurrentFocus
     var valueLength = normalizedElementText(input.attributes.value).count
     // The tap also focuses the field, which a clear can reach with focus unconfirmed.
 #if !os(tvOS)
-    try reader.dispatch { try moveCaretToEnd(app: app, input: input) }
+    if !defersCaretTap {
+      try reader.dispatch { try moveCaretToEnd(app: app, input: input) }
+    }
 #endif
     // An input that stops answering once focused (#3060) cannot be read back, so it gets one burst.
     guard try reader.input() != nil else {
@@ -456,6 +462,19 @@ extension RunnerTests {
       valueLength = remaining.count
     }
 #endif
+    if defersCaretTap {
+      guard let fallbackInput = try reader.input() else {
+        return
+      }
+#if !os(tvOS)
+      try reader.dispatch { try moveCaretToEnd(app: app, input: fallbackInput) }
+#endif
+      guard try reader.input() != nil else {
+        let burst = Self.textClearBurst(valueLength: valueLength)
+        reader.dispatch { fallbackInput.element.typeText(String(repeating: XCUIKeyboardKey.delete.rawValue, count: burst)) }
+        return
+      }
+    }
     for _ in 0...(valueLength / Self.maxTextClearBurst + 1) {
       guard let input = try reader.input() else {
         return

apple/runner/AgentDeviceRunner/AgentDeviceRunnerUITests/RunnerTests+TextEntryFocus.swift

@@ -104,7 +104,7 @@ extension RunnerTests {
   }
 
   @MainActor
-  func focusTextInputForTextEntry(app: XCUIApplication, x: Double?, y: Double?) throws -> TextEntryTarget {
+  func focusTextInputForTextEntry(app: XCUIApplication, x: Double?, y: Double?, reusingFocusedInput: Bool = false) throws -> TextEntryTarget {
     let contentTranslation = liveContentTranslation(app: app)
     guard let x, let y else {
       let softwareKeyboardVisible = isKeyboardVisible(app: app)
@@ -135,8 +135,23 @@ extension RunnerTests {
       return TextEntryTarget(element: focused, refreshPoint: nil, prefersFocusedElement: true, contentTranslation: contentTranslation)
     }
 
-    let keyboardVisibleBeforeTap = isKeyboardVisible(app: app)
     let input = try textInputCandidatesAt(app: app, point: CGPoint(x: x, y: y), contentTranslation: contentTranslation).first
+#if os(iOS)
+    if reusingFocusedInput, let input, let identity = input.identity, identity.isDistinguishable,
+       let snapshot = input.snapshot, snapshotHasKeyboardFocus(snapshot) {
+      NSLog("%@current-input-reports-keyboard-focus polls=0", TextEntryFocusEvidence.logMarker)
+      return TextEntryTarget(
+        element: input.element,
+        refreshPoint: textEntryRefreshPoint(frame: input.frameOnScreen, fallbackTo: CGPoint(x: x, y: y)),
+        prefersFocusedElement: false,
+        contentTranslation: contentTranslation,
+        reusesCurrentFocus: true,
+        boundIdentity: identity,
+        postFocusRead: input
+      )
+    }
+#endif
+    let keyboardVisibleBeforeTap = isKeyboardVisible(app: app)
     let target = input?.element
     let requestedPoint = CGPoint(x: x, y: y)
     // `tapAt` takes a screen point: with no screen box, tap the caller's own point (#3899, #3060).

apple/runner/AgentDeviceRunner/AgentDeviceRunnerUITests/RunnerTests+TypeExecution.swift

@@ -65,7 +65,7 @@ extension RunnerTests {
     if let resolvedCoordinateTarget {
       target = resolvedCoordinateTarget
     } else {
-      target = try focusTextInputForTextEntry(app: activeApp, x: command.x, y: command.y)
+      target = try focusTextInputForTextEntry(app: activeApp, x: command.x, y: command.y, reusingFocusedInput: textEntryMode == .replacement)
     }
     NSLog(
       "AGENT_DEVICE_RUNNER_TEXT_ENTRY_PHASE commandId=%@ phase=focus durationMs=%.1f chars=%d mode=%@",