fix(typegen): prevent duplication of generator-owned methods (readValidator/writeValidator) for Edm.Boolean fields during regeneration (#119)

<!-- CURSOR_SUMMARY -->
> [!NOTE]
> **Medium Risk**
> Touches code generation logic that parses and rewrites method chains; mistakes could drop or misclassify user customizations in regenerated schemas, though new e2e coverage reduces risk.
> 
> **Overview**
> Prevents type regeneration from re-appending generator-owned chain methods (notably `readValidator`/`writeValidator` used for `Edm.Boolean`) when preserving existing field customizations.
> 
> This extends customization extraction to accept additional “standard” method names derived by parsing the generated `fieldBuilder` chain, and adds e2e tests asserting validators are not duplicated and a user-level `.transform()` is still preserved.
> 
> <sup>Written by [Cursor Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit 8203a33. This will update automatically on new commits. Configure [here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->

Author: eluce2

.changeset/fix-typegen-boolean-validator-duplication.md

@@ -0,0 +1,5 @@
+---
+"@proofkit/typegen": patch
+---
+
+Fix typegen duplicating readValidator/writeValidator on Edm.Boolean fields during regeneration

packages/typegen/src/fmodata/generateODataTypes.ts

@@ -315,7 +315,7 @@ function generateTableOccurrence(
 
     // Preserve user customizations from existing field
     if (matchedExistingField) {
-      line = preserveUserCustomizations(matchedExistingField, line);
+      line = preserveUserCustomizations(matchedExistingField, line, fieldBuilder);
     }
 
     // Add comma if not the last field
@@ -435,7 +435,11 @@ interface ParsedTableOccurrence {
 /**
  * Extracts user customizations (like .inputValidator() and .outputValidator()) from a method chain
  */
-function extractUserCustomizations(chainText: string, baseChainEnd: number): string {
+function extractUserCustomizations(
+  chainText: string,
+  baseChainEnd: number,
+  additionalStandardMethods?: Set<string>,
+): string {
   // We want to preserve user-added chained calls even if they were placed:
   // - before a standard method (e.g. textField().inputValidator(...).entityId(...))
   // - on fields that have no standard methods at all (possible when reduceMetadata is true)
@@ -447,6 +451,11 @@ function extractUserCustomizations(chainText: string, baseChainEnd: number): str
   // that can be appended to the regenerated chain.
 
   const standardMethodNames = new Set(["primaryKey", "readOnly", "notNull", "entityId", "comment"]);
+  if (additionalStandardMethods) {
+    for (const m of additionalStandardMethods) {
+      standardMethodNames.add(m);
+    }
+  }
 
   const start = Math.max(0, Math.min(baseChainEnd, chainText.length));
   const tail = chainText.slice(start);
@@ -824,10 +833,102 @@ function matchFieldByName(existingFields: Map<string, ParsedField>, fieldName: s
   return existingFields.get(fieldName) || null;
 }
 
+/**
+ * Extracts method names called in a chain expression.
+ * e.g. "numberField().readValidator(...).writeValidator(...)" => {"readValidator", "writeValidator"}
+ */
+function extractMethodNamesFromChain(chain: string): Set<string> {
+  const names = new Set<string>();
+  let depth = 0;
+  let i = 0;
+
+  function skipStringLiteral(quote: string): void {
+    i++; // skip opening quote
+    while (i < chain.length) {
+      if (chain[i] === "\\") {
+        i += 2;
+        continue;
+      }
+      if (chain[i] === quote) {
+        i++;
+        break;
+      }
+      i++;
+    }
+  }
+
+  while (i < chain.length) {
+    const ch = chain[i] ?? "";
+    // Skip string literals (handles quotes inside parens correctly)
+    if (ch === "'" || ch === '"' || ch === "`") {
+      skipStringLiteral(ch);
+      continue;
+    }
+    if (ch === "(") {
+      depth++;
+      i++;
+      continue;
+    }
+    if (ch === ")") {
+      depth--;
+      i++;
+      continue;
+    }
+    // Only match `.methodName(` at the top-level chain (depth 0)
+    if (ch === "." && depth === 0) {
+      i++;
+      const nameStart = i;
+      while (i < chain.length && REGEX_IDENT_CHAR.test(chain[i] ?? "")) {
+        i++;
+      }
+      if (i > nameStart) {
+        const name = chain.slice(nameStart, i);
+        while (i < chain.length && REGEX_WHITESPACE.test(chain[i] ?? "")) {
+          i++;
+        }
+        // Skip optional generic type args <...>
+        if (i < chain.length && chain[i] === "<") {
+          let angleDepth = 0;
+          while (i < chain.length) {
+            const c = chain[i] ?? "";
+            if (c === "'" || c === '"' || c === "`") {
+              skipStringLiteral(c);
+              continue;
+            }
+            if (c === "<") {
+              angleDepth++;
+            } else if (c === ">") {
+              angleDepth--;
+              if (angleDepth === 0) {
+                i++;
+                break;
+              }
+            }
+            i++;
+          }
+          while (i < chain.length && REGEX_WHITESPACE.test(chain[i] ?? "")) {
+            i++;
+          }
+        }
+        if (i < chain.length && chain[i] === "(") {
+          names.add(name);
+        }
+      }
+      continue;
+    }
+    i++;
+  }
+  return names;
+}
+
 /**
  * Preserves user customizations from an existing field chain
  */
-function preserveUserCustomizations(existingField: ParsedField | undefined, newChain: string): string {
+function preserveUserCustomizations(
+  existingField: ParsedField | undefined,
+  newChain: string,
+  fieldBuilder: string,
+): string {
   if (!existingField) {
     return newChain;
   }
@@ -848,7 +949,10 @@ function preserveUserCustomizations(existingField: ParsedField | undefined, newC
   const existingChainText = existingField.fullChainText;
   const existingBaseEnd = existingChainText.startsWith(baseBuilderPrefix) ? baseBuilderPrefix.length : 0;
 
-  const userCustomizations = extractUserCustomizations(existingChainText, existingBaseEnd);
+  // Methods in the generated field builder (e.g. readValidator, writeValidator for
+  // Edm.Boolean) are generator-owned and must not be duplicated as user customizations.
+  const generatorMethods = extractMethodNamesFromChain(fieldBuilder);
+  const userCustomizations = extractUserCustomizations(existingChainText, existingBaseEnd, generatorMethods);
 
   if (!userCustomizations) {
     return newChain;

packages/typegen/tests/e2e/fmodata-preserve-customizations.test.ts

@@ -129,6 +129,102 @@ describe("fmodata generateODataTypes preserves user customizations", () => {
     }
   });
 
+  it("does not duplicate generator-owned methods for Edm.Boolean fields", async () => {
+    const tmpDir = await fs.mkdtemp(path.join(os.tmpdir(), "proofkit-fmodata-preserve-"));
+
+    try {
+      const entitySetName = "MyTable";
+      const entityTypeName = "NS.MyTable";
+      const metadata = makeMetadata({
+        entitySetName,
+        entityTypeName,
+        fields: [{ name: "is_active", type: "Edm.Boolean", fieldId: "F1" }],
+      });
+
+      const existingFilePath = path.join(tmpDir, "MyTable.ts");
+      await fs.writeFile(
+        existingFilePath,
+        [
+          `import { fmTableOccurrence, numberField } from "@proofkit/fmodata";`,
+          `import { z } from "zod/v4";`,
+          "",
+          `export const MyTable = fmTableOccurrence("MyTable", {`,
+          `  is_active: numberField().readValidator(z.coerce.boolean()).writeValidator(z.boolean().transform((v) => (v ? 1 : 0))).entityId("F1"),`,
+          "}, {",
+          `  entityId: "T1",`,
+          "});",
+          "",
+        ].join("\n"),
+        "utf8",
+      );
+
+      await generateODataTypes(metadata, {
+        type: "fmodata",
+        path: tmpDir,
+        clearOldFiles: false,
+        tables: [{ tableName: "MyTable" }],
+      });
+
+      const regenerated = await fs.readFile(existingFilePath, "utf8");
+      // readValidator and writeValidator should each appear exactly once
+      const readValidatorCount = (regenerated.match(/readValidator/g) || []).length;
+      const writeValidatorCount = (regenerated.match(/writeValidator/g) || []).length;
+      expect(readValidatorCount).toBe(1);
+      expect(writeValidatorCount).toBe(1);
+    } finally {
+      await fs.rm(tmpDir, { recursive: true, force: true });
+    }
+  });
+
+  it("preserves user .transform() on Edm.Boolean fields (not confused with nested .transform)", async () => {
+    const tmpDir = await fs.mkdtemp(path.join(os.tmpdir(), "proofkit-fmodata-preserve-"));
+
+    try {
+      const entitySetName = "MyTable";
+      const entityTypeName = "NS.MyTable";
+      const metadata = makeMetadata({
+        entitySetName,
+        entityTypeName,
+        fields: [{ name: "is_active", type: "Edm.Boolean", fieldId: "F1" }],
+      });
+
+      const existingFilePath = path.join(tmpDir, "MyTable.ts");
+      await fs.writeFile(
+        existingFilePath,
+        [
+          `import { fmTableOccurrence, numberField } from "@proofkit/fmodata";`,
+          `import { z } from "zod/v4";`,
+          "",
+          `export const MyTable = fmTableOccurrence("MyTable", {`,
+          `  is_active: numberField().readValidator(z.coerce.boolean()).writeValidator(z.boolean().transform((v) => (v ? 1 : 0))).entityId("F1").transform((v) => !!v),`,
+          "}, {",
+          `  entityId: "T1",`,
+          "});",
+          "",
+        ].join("\n"),
+        "utf8",
+      );
+
+      await generateODataTypes(metadata, {
+        type: "fmodata",
+        path: tmpDir,
+        clearOldFiles: false,
+        tables: [{ tableName: "MyTable" }],
+      });
+
+      const regenerated = await fs.readFile(existingFilePath, "utf8");
+      // The user-added .transform() should be preserved
+      expect(regenerated).toContain(".transform((v) => !!v)");
+      // readValidator and writeValidator should each appear exactly once
+      const readValidatorCount = (regenerated.match(/readValidator/g) || []).length;
+      const writeValidatorCount = (regenerated.match(/writeValidator/g) || []).length;
+      expect(readValidatorCount).toBe(1);
+      expect(writeValidatorCount).toBe(1);
+    } finally {
+      await fs.rm(tmpDir, { recursive: true, force: true });
+    }
+  });
+
   it("preserves aliased imports when regenerating files", async () => {
     const tmpDir = await fs.mkdtemp(path.join(os.tmpdir(), "proofkit-fmodata-preserve-"));