Drop 24-hex ObjectId guess; EJSON.parse handles ObjectId via $oid form

Author: tnaum-ms

src/documentdb/utils/parseDocumentId.test.ts

@@ -42,18 +42,20 @@ describe('parseDocumentId', () => {
         outputChannelError.mockClear();
     });
 
-    it('parses an ObjectId from canonical EJSON', () => {
+    it('parses an ObjectId from its EJSON $oid form (no hex guessing needed)', () => {
         const oid = new ObjectId();
+        // This is exactly how the UI serializes an ObjectId _id: {"$oid":"..."}.
+        expect(canonical(oid)).toBe(`{"$oid":"${oid.toHexString()}"}`);
+
         const result = parseDocumentId(canonical(oid));
         expect(result).toBeInstanceOf(ObjectId);
         expect((result as ObjectId).toHexString()).toBe(oid.toHexString());
     });
 
-    it('parses a bare 24-character hex string as an ObjectId (legacy shape)', () => {
+    it('does NOT guess: a bare 24-character hex string (no $oid) throws', () => {
         const hex = 'a'.repeat(24);
-        const result = parseDocumentId(hex);
-        expect(result).toBeInstanceOf(ObjectId);
-        expect((result as ObjectId).toHexString()).toBe(hex);
+        expect(() => parseDocumentId(hex)).toThrow(/Unable to parse the document _id/);
+        expect(outputChannelError).toHaveBeenCalledTimes(1);
     });
 
     it('keeps a string _id as a string', () => {

src/documentdb/utils/parseDocumentId.ts

@@ -5,7 +5,6 @@
 
 import * as l10n from '@vscode/l10n';
 import { EJSON } from 'bson';
-import { ObjectId } from 'mongodb';
 import { ext } from '../../extensionVariables';
 
 /**
@@ -22,30 +21,24 @@ const MAX_LOGGED_ID_LENGTH = 256;
  * The id is expected to be Extended JSON (EJSON) produced by `EJSON.stringify`,
  * which is the BSON library's own canonical/relaxed serialization. `EJSON.parse`
  * is therefore the ground truth: it round-trips every `_id` type the driver
- * supports — `ObjectId`, string, number, `UUID`, `Date`, `Decimal128`,
- * **embedded documents** (e.g. `{ author: 'John', userId: 2343345 }`) and
- * arrays — preserving field order, which is significant when the driver matches
- * an embedded-document `_id`.
+ * supports — `ObjectId` (as `{"$oid":"…"}`), string, number, `UUID`, `Date`,
+ * `Decimal128`, **embedded documents** (e.g. `{ author: 'John', userId: 2343345 }`)
+ * and arrays — preserving field order, which is significant when the driver
+ * matches an embedded-document `_id`.
  *
- * As a single deterministic concession we also accept a bare 24-character hex
- * string and construct an `ObjectId` from it (a legacy, non-EJSON id shape).
- *
- * If the id maps to neither, we **throw** rather than guess. Silently coercing
- * an unrecognized id into a raw string would change which document the driver
- * matches and could read or delete the wrong record, so we fail loudly instead.
- * The offending value (length-capped) is written to the output channel for
- * follow-up.
+ * If the id cannot be parsed, we **throw** rather than guess. Silently coercing
+ * an unrecognized id into a raw string (or assuming it is an ObjectId) would
+ * change which document the driver matches and could read or delete the wrong
+ * record, so we fail loudly instead. The offending value (length-capped) is
+ * written to the output channel for follow-up.
  *
  * @throws Error if the id cannot be interpreted as a BSON value the driver understands.
  */
 export function parseDocumentId(id: string): unknown {
     try {
         return EJSON.parse(id);
     } catch {
-        // Not Extended JSON. The only other shape we accept is a bare 24-char hex ObjectId.
-        if (ObjectId.isValid(id)) {
-            return new ObjectId(id);
-        }
+        // Not Extended JSON — fall through to fail loudly below rather than guessing.
     }
 
     const preview = id.length > MAX_LOGGED_ID_LENGTH ? id.slice(0, MAX_LOGGED_ID_LENGTH) : id;
@@ -55,8 +48,8 @@ export function parseDocumentId(id: string): unknown {
 
     ext.outputChannel.error(
         'Unable to parse document _id. The value could not be interpreted as a BSON type understood ' +
-            'by the driver (expected Extended JSON — e.g. {"$oid":"…"}, {"$numberInt":"…"}, an embedded ' +
-            `document — or a 24-character hex ObjectId). Received _id: ${detail}`,
+            'by the driver (expected Extended JSON — e.g. {"$oid":"…"}, {"$numberInt":"…"}, or an ' +
+            `embedded document). Received _id: ${detail}`,
     );
 
     throw new Error(l10n.t('Unable to parse the document _id. See the output channel for details.'));