@codeyam/codeyam-cli 0.1.0-staging.76566f9 → 0.1.0-staging.78c1cad

package/analyzer-template/packages/ai/src/lib/astScopes/methodSemantics.ts

@@ -139,6 +139,96 @@ export class ArrayFilterSemantics implements MethodSemantics {
   }
 }
 
+/**
+ * Array some method semantics
+ *
+ * some() returns true if at least one element passes the test, false otherwise
+ * The callback receives array elements to test
+ */
+export class ArraySomeSemantics implements MethodSemantics {
+  getReturnType(): string {
+    return 'boolean';
+  }
+
+  addEquivalences(
+    methodCallPath: StructuredPath,
+    sourcePath: StructuredPath,
+    context: AnalysisContext,
+  ): void {
+    // Mark source as array
+    context.addType(sourcePath, 'array');
+
+    // Mark return as boolean
+    context.addType(methodCallPath, 'boolean');
+
+    // Get the last function call segment to access the arguments
+    const lastFunctionCallSegment = methodCallPath.getLastFunctionCallSegment();
+
+    if (lastFunctionCallSegment && lastFunctionCallSegment.args.length > 0) {
+      // The first argument is the callback
+      const callbackPath = lastFunctionCallSegment.args[0];
+
+      // Add the callback to the structure
+      context.addType(callbackPath, 'function');
+
+      // Mark that the callback's first parameter receives array elements
+      context.addEquivalence(
+        callbackPath.withParameter(0),
+        sourcePath.withElement('*'),
+      );
+    }
+  }
+
+  isComplete(): boolean {
+    return true;
+  }
+}
+
+/**
+ * Array every method semantics
+ *
+ * every() returns true if all elements pass the test, false otherwise
+ * The callback receives array elements to test
+ */
+export class ArrayEverySemantics implements MethodSemantics {
+  getReturnType(): string {
+    return 'boolean';
+  }
+
+  addEquivalences(
+    methodCallPath: StructuredPath,
+    sourcePath: StructuredPath,
+    context: AnalysisContext,
+  ): void {
+    // Mark source as array
+    context.addType(sourcePath, 'array');
+
+    // Mark return as boolean
+    context.addType(methodCallPath, 'boolean');
+
+    // Get the last function call segment to access the arguments
+    const lastFunctionCallSegment = methodCallPath.getLastFunctionCallSegment();
+
+    if (lastFunctionCallSegment && lastFunctionCallSegment.args.length > 0) {
+      // The first argument is the callback
+      const callbackPath = lastFunctionCallSegment.args[0];
+
+      // Add the callback to the structure
+      context.addType(callbackPath, 'function');
+
+      // Mark that the callback's first parameter receives array elements
+      context.addEquivalence(
+        callbackPath.withParameter(0),
+        sourcePath.withElement('*'),
+      );
+    }
+  }
+
+  isComplete(): boolean {
+    return true;
+  }
+}
+
 /**
  * Array find method semantics
  *
@@ -173,12 +263,12 @@ export class ArrayFindSemantics implements MethodSemantics {
         sourcePath.withElement('*'),
       );
 
-      // THIS IS PRODUCING STRANGE RESULTS SUCH AS
-      // 'people.filter(cyScope2()).find(cyScope3())': 'people.filter(cyScope2()).functionCallReturnValue[]',
-      // 'personWithHome.places.find(cyScope5())': 'personWithHome.places[]'
-      // WHICH DO NOT MAKES SENSE
-      // REMOVING BUT LEAVING IN PLACE IN CASE THERE IS A VALID USE CASE
-      // The result is an element from the array that matches the callback condition
+      // NOTE: We intentionally do NOT create an equivalency from the find result
+      // to the array element here. While semantically correct (find returns an element),
+      // doing so causes type inference issues where the array element gets typed as
+      // 'function' instead of 'object'. The tracing back to source arrays happens
+      // through the || operator processing and null guard in addEquivalence.
+      // See: analyzeScopeNoAI1.test.ts for tests that break if this is uncommented.
       // context.addEquivalence(methodCallPath, sourcePath.withElement('*'));
     }
   }
@@ -239,11 +329,17 @@ export class ArrayReduceSemantics implements MethodSemantics {
 /**
  * Array concat method semantics
  *
- * concat() returns a new array combining the original array and additional items
+ * concat() returns a new array/string combining the original and additional items
+ *
+ * NOTE: We intentionally do NOT type the source as 'array' here because
+ * .concat() exists on both arrays AND strings. Typing the source prematurely
+ * causes type contamination when the source is equivalent to other paths
+ * via ?? chains. Type inference happens later in checkIfKnownType() based
+ * on accumulated evidence from all methods called on a path.
  */
 export class ArrayConcatSemantics implements MethodSemantics {
   getReturnType(): string {
-    return 'array';
+    return 'unknown'; // Could be array or string depending on source
   }
 
   addEquivalences(
@@ -251,23 +347,24 @@ export class ArrayConcatSemantics implements MethodSemantics {
     sourcePath: StructuredPath,
     context: AnalysisContext,
   ): void {
-    // Mark source as array
-    context.addType(sourcePath, 'array');
+    // NOTE: We use 'unknown' instead of 'array' for the source because .concat()
+    // exists on both arrays AND strings. Using 'array' causes type contamination
+    // when the source is equivalent to other paths via ?? chains.
+    context.addType(sourcePath, 'unknown');
 
-    // Any arguments are also added to the array
+    // Any arguments are also added to the result
     const functionCallSegment = methodCallPath.getLastFunctionCallSegment();
 
     if (functionCallSegment) {
       // Each argument could be an array or single element
       functionCallSegment.args.forEach((argPath) => {
         // No need to guess the type - we'll discover it through normal processing
-        // But we can indicate that these elements will be in the result array
+        // But we can indicate that these elements will be in the result
         context.addEquivalence(methodCallPath, argPath);
       });
     }
 
-    // The original array elements are also in the result
-    context.addType(methodCallPath, 'array');
+    // The original elements are also in the result
     context.addEquivalence(methodCallPath, sourcePath.withElement('*'));
   }
 
@@ -279,11 +376,17 @@ export class ArrayConcatSemantics implements MethodSemantics {
 /**
  * Array slice method semantics
  *
- * slice() returns a shallow copy of a portion of an array
+ * slice() returns a shallow copy of a portion of an array (or string)
+ *
+ * NOTE: We intentionally do NOT type the source as 'array' here because
+ * .slice() exists on both arrays AND strings. Typing the source prematurely
+ * causes type contamination when the source is equivalent to other paths
+ * via ?? chains. Type inference happens later in checkIfKnownType() based
+ * on accumulated evidence from all methods called on a path.
  */
 export class ArraySliceSemantics implements MethodSemantics {
   getReturnType(): string {
-    return 'array';
+    return 'unknown'; // Could be array or string depending on source
   }
 
   addEquivalences(
@@ -291,12 +394,15 @@ export class ArraySliceSemantics implements MethodSemantics {
     sourcePath: StructuredPath,
     context: AnalysisContext,
   ): void {
-    // Mark source as array
-    context.addType(sourcePath, 'array');
+    // NOTE: We use 'unknown' instead of 'array' for the source because .slice()
+    // exists on both arrays AND strings. Using 'array' causes type contamination
+    // when the source is equivalent to other paths via ?? chains.
+    // Type inference happens later in checkIfKnownType() based on accumulated evidence.
+    context.addType(sourcePath, 'unknown');
 
-    // The elements in the result are the same type as the source array
+    // Still track the return value type for data structure tracking
     const methodCallPathWithReturnValues = methodCallPath.withReturnValues();
-    context.addType(methodCallPathWithReturnValues, 'array');
+    context.addType(methodCallPathWithReturnValues, 'unknown');
   }
 
   isComplete(): boolean {
@@ -518,6 +624,12 @@ export class ArrayFlatMapSemantics implements MethodSemantics {
  * Array at method semantics
  *
  * at() returns the element at the given index, supporting negative indices
+ *
+ * NOTE: We intentionally do NOT type the source as 'array' here because
+ * .at() exists on both arrays AND strings. Typing the source prematurely
+ * causes type contamination when the source is equivalent to other paths
+ * via ?? chains. Type inference happens later in checkIfKnownType() based
+ * on accumulated evidence from all methods called on a path.
  */
 export class ArrayAtSemantics implements MethodSemantics {
   getReturnType(): string {
@@ -529,10 +641,12 @@ export class ArrayAtSemantics implements MethodSemantics {
     sourcePath: StructuredPath,
     context: AnalysisContext,
   ): void {
-    // Mark source as array
-    context.addType(sourcePath, 'array');
+    // NOTE: We use 'unknown' instead of 'array' for the source because .at()
+    // exists on both arrays AND strings. Using 'array' causes type contamination
+    // when the source is equivalent to other paths via ?? chains.
+    context.addType(sourcePath, 'unknown');
 
-    // The return value is equivalent to an array element
+    // The return value is equivalent to an element
     context.addEquivalence(
       methodCallPath.withReturnValues(),
       sourcePath.withElement('*'),
@@ -616,6 +730,45 @@ export class ObjectFromEntriesSemantics implements MethodSemantics {
   }
 }
 
+/**
+ * String split method semantics
+ *
+ * split() returns an array of strings by splitting the source string on a delimiter.
+ * This is commonly used for URL path parsing: splat?.split('/') → segments[]
+ *
+ * Important for controllability: if the source string is controllable (e.g., from useParams()),
+ * then the resulting array elements are also controllable.
+ */
+export class StringSplitSemantics implements MethodSemantics {
+  getReturnType(): string {
+    return 'string[]';
+  }
+
+  addEquivalences(
+    methodCallPath: StructuredPath,
+    sourcePath: StructuredPath,
+    context: AnalysisContext,
+  ): void {
+    // Mark source as string
+    context.addType(sourcePath, 'string');
+
+    // Mark the result as a string array
+    context.addType(methodCallPath, 'string[]');
+
+    // Establish equivalence: each element of the result derives from the source string
+    // This enables controllability tracking: if sourcePath is controllable,
+    // then methodCallPath[] elements are controllable via the same mock
+    context.addEquivalence(
+      methodCallPath.withReturnValues().withElement('*'),
+      sourcePath,
+    );
+  }
+
+  isComplete(): boolean {
+    return true;
+  }
+}
+
 /**
  * Promise.then method semantics
  *
@@ -742,6 +895,132 @@ export class UseStateSemantics implements MethodSemantics {
   }
 }
 
+// ─── Set/Map method semantics ──────────────────────────────────────────
+//
+// Set/Map membership operations (.has, .delete, .add, .get, .set, .clear)
+// don't create data flow equivalencies — they're membership checks or
+// mutations, not transformations that propagate schema structure.
+// Without these semantics, code like `filterRatings.has(articleId)` creates
+// spurious equivalencies (332 entries in the Margo LibraryPage case).
+
+/**
+ * Set.has() / Map.has() — membership check, returns boolean
+ */
+export class CollectionHasSemantics implements MethodSemantics {
+  getReturnType(): string {
+    return 'boolean';
+  }
+
+  addEquivalences(
+    methodCallPath: StructuredPath,
+    _sourcePath: StructuredPath,
+    context: AnalysisContext,
+  ): void {
+    context.addType(methodCallPath, 'boolean');
+  }
+
+  isComplete(): boolean {
+    return true;
+  }
+}
+
+/**
+ * Set.delete() / Map.delete() — removal, returns boolean
+ */
+export class CollectionDeleteSemantics implements MethodSemantics {
+  getReturnType(): string {
+    return 'boolean';
+  }
+
+  addEquivalences(
+    methodCallPath: StructuredPath,
+    _sourcePath: StructuredPath,
+    context: AnalysisContext,
+  ): void {
+    context.addType(methodCallPath, 'boolean');
+  }
+
+  isComplete(): boolean {
+    return true;
+  }
+}
+
+/**
+ * Set.add() — adds an element, returns the Set (for chaining)
+ */
+export class SetAddSemantics implements MethodSemantics {
+  getReturnType(): string {
+    return 'object'; // Returns the Set itself
+  }
+
+  addEquivalences(
+    _methodCallPath: StructuredPath,
+    _sourcePath: StructuredPath,
+    _context: AnalysisContext,
+  ): void {
+    // No equivalencies — add() doesn't create data flow
+  }
+
+  isComplete(): boolean {
+    return true;
+  }
+}
+
+/**
+ * Set.clear() / Map.clear() — removes all elements, returns void
+ */
+export class CollectionClearSemantics implements MethodSemantics {
+  getReturnType(): string {
+    return 'void';
+  }
+
+  addEquivalences(): void {
+    // No equivalencies
+  }
+
+  isComplete(): boolean {
+    return true;
+  }
+}
+
+/**
+ * Map.get() — retrieves a value by key, returns unknown (value type)
+ */
+export class MapGetSemantics implements MethodSemantics {
+  getReturnType(): string {
+    return 'unknown';
+  }
+
+  addEquivalences(
+    _methodCallPath: StructuredPath,
+    _sourcePath: StructuredPath,
+    _context: AnalysisContext,
+  ): void {
+    // No equivalencies — the return value type depends on the map contents
+  }
+
+  isComplete(): boolean {
+    return true;
+  }
+}
+
+/**
+ * Map.set() — sets a key-value pair, returns the Map (for chaining)
+ */
+export class MapSetSemantics implements MethodSemantics {
+  getReturnType(): string {
+    return 'object'; // Returns the Map itself
+  }
+
+  addEquivalences(): void {
+    // No equivalencies
+  }
+
+  isComplete(): boolean {
+    return true;
+  }
+}
+
 /**
  * Create and populate the registry with method semantics
  */
@@ -761,10 +1040,15 @@ export function createMethodRegistry(): MethodSemanticsRegistry {
   registry.register('slice', new ArraySliceSemantics(), 'Array');
   registry.register('splice', new ArraySpliceSemantics(), 'Array');
   registry.register('push', new ArrayPushSemantics(), 'Array');
+  registry.register('some', new ArraySomeSemantics(), 'Array');
+  registry.register('every', new ArrayEverySemantics(), 'Array');
 
   // Register Object methods
   registry.register('fromEntries', new ObjectFromEntriesSemantics(), 'Object');
 
+  // Register String methods
+  registry.register('split', new StringSplitSemantics(), 'String');
+
   // Register Promise methods
   registry.register('then', new PromiseThenSemantics(), 'Promise');
 
@@ -772,6 +1056,15 @@ export function createMethodRegistry(): MethodSemanticsRegistry {
   registry.register('useState', new UseStateSemantics(), 'React');
   registry.register('useMemo', new UseMemoSemantics(), 'React');
 
+  // Register Set/Map collection methods.
+  // Only .has() and .delete() are safe to register by method name — they always
+  // return boolean regardless of the collection type (Set, Map, WeakSet, WeakMap,
+  // FormData, URLSearchParams, etc.).
+  // .add(), .clear(), .get(), .set() are too ambiguous across different types
+  // and can't be registered until the lookup code supports class-aware resolution.
+  registry.register('has', new CollectionHasSemantics());
+  registry.register('delete', new CollectionDeleteSemantics());
+
   return registry;
 }
 

package/analyzer-template/packages/ai/src/lib/astScopes/nodeToSource.ts

@@ -54,6 +54,25 @@ export function nodeToSource(
   //   return serializeBindingElement(node, sourceFile, options);
   // }
 
+  // Object/array/function literals should be marked as literal paths.
+  // This prevents them from creating sourceEquivalencies when used as
+  // function call arguments (e.g., setUndoEntry({ label: '...', undo: ... })).
+  // Without this, the full source text of the literal (300-500 chars for complex
+  // objects) gets embedded in path strings, creating expensive-to-parse
+  // equivalency entries that don't contribute useful schema information.
+  if (ts.isObjectLiteralExpression(node)) {
+    return StructuredPath.createLiteralPath(node.getText(sourceFile), 'object');
+  }
+  if (ts.isArrayLiteralExpression(node)) {
+    return StructuredPath.createLiteralPath(node.getText(sourceFile), 'array');
+  }
+  if (ts.isArrowFunction(node) || ts.isFunctionExpression(node)) {
+    return StructuredPath.createLiteralPath(
+      node.getText(sourceFile),
+      'function',
+    );
+  }
+
   // Fallback for unsupported node types
   return StructuredPath.fromBase(node.getText(sourceFile));
 }

package/analyzer-template/packages/ai/src/lib/astScopes/paths.ts

@@ -406,11 +406,18 @@ export class StructuredPath {
           } else {
             // Convert each argument to its string representation
             const argsStr = funcSegment.args
-              .map((arg) =>
-                arg.toStringWithSegments(arg.base, arg.segments, {
+              .map((arg) => {
+                // For literal arguments (object/array/function literals),
+                // use a compact placeholder instead of the full source text.
+                // This prevents 300-500 char inline objects from being embedded
+                // in call signature paths, which creates expensive equivalencies.
+                if (arg.isLiteral()) {
+                  return `{...}`;
+                }
+                return arg.toStringWithSegments(arg.base, arg.segments, {
                   withReturnValues: doRenderNestedFunctionCallReturnValue,
-                }),
-              )
+                });
+              })
               .join(', ');
             result += `(${argsStr})`;
           }

package/analyzer-template/packages/ai/src/lib/astScopes/patterns/forInStatementHandler.ts

@@ -22,11 +22,8 @@ export class ForInStatementHandler implements PatternHandler {
     // Process the expression being iterated over first (e.g., object)
     context.processExpression(node.expression);
 
-    // Get the path of the object being iterated
-    const expressionPath = context.getStructuralPath(node.expression);
-
-    // If the initializer is a variable declaration with an identifier, establish equivalence
-    // to indicate it's iterating over the keys of the expression
+    // If the initializer is a variable declaration with an identifier,
+    // type it as 'string' since for-in keys are always strings
     if (ts.isVariableDeclarationList(node.initializer)) {
       const declarations = node.initializer.declarations;
       if (
@@ -36,25 +33,21 @@ export class ForInStatementHandler implements PatternHandler {
       ) {
         const variableName = StructuredPath.fromBase(declarations[0].name.text);
 
-        // Add type for the key variable (string for object keys)
+        // Type the key variable as 'string' — for-in keys are always strings.
+        // Note: We intentionally do NOT equivalence the key variable to
+        // expressionPath[*key*]. The key variable is the KEY (a string),
+        // while [*key*] in the schema represents the VALUE at a dynamic key.
+        // Creating that equivalence would cause the 'string' type to propagate
+        // to the value path. The value access (e.g., obj[key]) is handled
+        // correctly by the element access handler in processExpression.
         context.addType(variableName, 'string');
-
-        if (expressionPath) {
-          // For for...in loops, the variable is equivalent to the keys of the object
-          // We represent this with a *key* suffix to indicate it's a key, not a value
-          context.addEquivalence(variableName, expressionPath.withKey('key'));
-        }
       }
     } else if (ts.isIdentifier(node.initializer)) {
       // If the initializer is directly an identifier (rare but possible)
       const variableName = StructuredPath.fromBase(node.initializer.text);
 
-      // Add type for the key variable
+      // See comment above — type as string but don't equivalence to value path
       context.addType(variableName, 'string');
-
-      if (expressionPath) {
-        context.addEquivalence(variableName, expressionPath.withKey('key'));
-      }
     } else {
       context.processExpression(node.initializer);
     }

package/analyzer-template/packages/ai/src/lib/astScopes/patterns/ifStatementHandler.ts

@@ -5,6 +5,10 @@ import {
   markConditionVariablesAsNullable,
   extractConditionalUsage,
 } from '../processExpression';
+import {
+  extractConditionalEffectsFromIfStatement,
+  findUseStateSetters,
+} from '../conditionalEffectsExtractor';
 
 /**
  * Handler for if statements
@@ -13,6 +17,7 @@ import {
  * - The condition expression
  * - The then block
  * - The else block if present
+ * - Conditional effects (setter calls inside conditionals)
  */
 export class IfStatementHandler implements PatternHandler {
   canHandle(node: ts.Node): boolean {
@@ -31,6 +36,19 @@ export class IfStatementHandler implements PatternHandler {
     // Process the condition expression first
     context.processExpression(node.expression);
 
+    // Extract conditional effects (setter calls inside this if statement)
+    const knownSetters = findUseStateSetters(context.sourceFile);
+
+    const effects = extractConditionalEffectsFromIfStatement(
+      node,
+      context,
+      knownSetters,
+    );
+
+    for (const effect of effects) {
+      context.addConditionalEffect(effect);
+    }
+
     // Process the 'then' statement block
     context.processNode(node.thenStatement);
 

package/analyzer-template/packages/ai/src/lib/astScopes/patterns/switchStatementHandler.ts

@@ -2,6 +2,10 @@ import ts from 'typescript';
 import { AnalysisContext, ConditionalUsage } from '../types';
 import { PatternHandler } from './patternHandler';
 import { StructuredPath } from '../paths';
+import {
+  extractConditionalEffectsFromSwitchStatement,
+  findUseStateSetters,
+} from '../conditionalEffectsExtractor';
 
 /**
  * Helper to extract source location from an AST node
@@ -80,6 +84,17 @@ export class SwitchStatementHandler implements PatternHandler {
       }
     }
 
+    // Extract conditional effects (setter calls inside switch cases)
+    const knownSetters = findUseStateSetters(context.sourceFile);
+    const effects = extractConditionalEffectsFromSwitchStatement(
+      node,
+      context,
+      knownSetters,
+    );
+    for (const effect of effects) {
+      context.addConditionalEffect(effect);
+    }
+
     // Process each clause in the switch statement
     for (const clause of node.caseBlock.clauses) {
       // Process the case expression if it's a case clause (not default)