fix(core): use unique symbol to indicate tags for cbor serialization

Author: kuhe

.changeset/sweet-planets-give.md

@@ -0,0 +1,5 @@
+---
+"@smithy/core": patch
+---
+
+make CBOR tags more distinct in JS

packages/core/src/submodules/cbor/cbor-decode.ts

@@ -26,6 +26,7 @@ import {
   specialNull,
   specialTrue,
   specialUndefined,
+  tagSymbol,
   Uint8,
   Uint32,
   Uint64,
@@ -122,7 +123,7 @@ export function decode(at: Uint32, to: Uint32): CborValueType {
         const valueOffset = _offset;
 
         _offset = offset + valueOffset;
-        return { tag: castBigInt(unsignedInt), value };
+        return { tag: castBigInt(unsignedInt), value, [tagSymbol]: true };
       }
     case majorUtf8String:
     case majorMap:

packages/core/src/submodules/cbor/cbor-encode.ts

@@ -9,12 +9,14 @@ import {
   majorMap,
   majorNegativeInt64,
   majorSpecial,
+  majorTag,
   majorUint64,
   majorUnstructuredByteString,
   majorUtf8String,
   specialFalse,
   specialNull,
   specialTrue,
+  tagSymbol,
   Uint64,
 } from "./cbor-types";
 import { alloc } from "./cbor-types";
@@ -179,6 +181,15 @@ export function encode(_input: any): void {
       cursor += input.byteLength;
       continue;
     } else if (typeof input === "object") {
+      if (input[tagSymbol]) {
+        if ("tag" in input && "value" in input) {
+          encodeStack.push(input.value);
+          encodeHeader(majorTag, input.tag);
+          continue;
+        } else {
+          throw new Error("tag encountered with missing fields, need 'tag' and 'value', found: " + JSON.stringify(input));
+        }
+      }
       const keys = Object.keys(input);
       for (let i = keys.length - 1; i >= 0; --i) {
         const key = keys[i];

packages/core/src/submodules/cbor/cbor-types.ts

@@ -66,3 +66,13 @@ export const minorIndefinite = 31; // 0b11111
 export function alloc(size: number) {
   return typeof Buffer !== "undefined" ? Buffer.alloc(size) : new Uint8Array(size);
 }
+
+/**
+ * @public
+ *
+ * The presence of this symbol as an object key indicates it should be considered a tag
+ * for CBOR serialization purposes.
+ *
+ * The object must also have the properties "tag" and "value".
+ */
+export const tagSymbol = Symbol("@smithy/core/cbor::tagSymbol");

packages/core/src/submodules/cbor/cbor.spec.ts

@@ -5,6 +5,8 @@ import { describe, expect, test as it } from "vitest";
 
 import { cbor } from "./cbor";
 import { bytesToFloat16 } from "./cbor-decode";
+import { tagSymbol } from "./cbor-types";
+import { dateToTag } from "./parseCborBody";
 
 // syntax is ESM but the test target is CJS.
 const here = __dirname;
@@ -179,6 +181,18 @@ describe("cbor", () => {
         161, 103, 109, 101, 115, 115, 97, 103, 101, 108, 104, 101, 108, 108, 111, 44, 32, 119, 111, 114, 108, 100,
       ]),
     },
+    {
+      name: "date=0",
+      data: dateToTag(new Date(0)),
+      // major tag (6 or 110), minor 1 (timestamp)
+      cbor: allocByteArray([0b11000001, 0]),
+    },
+    {
+      name: "date=turn of the millenium",
+      data: dateToTag(new Date(946684799999)),
+      // major tag (6 or 110), minor 1 (timestamp)
+      cbor: allocByteArray([0b11000001, 251, 65, 204, 54, 161, 191, 255, 223, 59]),
+    },
     {
       name: "complex object",
       data: {
@@ -202,7 +216,7 @@ describe("cbor", () => {
   ];
 
   const toBytes = (hex: string) => {
-    const bytes = [];
+    const bytes = [] as number[];
     hex.replace(/../g, (substr: string): string => {
       bytes.push(parseInt(substr, 16));
       return substr;
@@ -211,6 +225,19 @@ describe("cbor", () => {
   };
 
   describe("locally curated scenarios", () => {
+    it("should throw an error if serializing a tag with missing properties", () => {
+      expect(() =>
+        cbor.serialize({
+          myTag: {
+            [tagSymbol]: true,
+            tag: 1,
+            // value: undefined
+          },
+        })
+      ).toThrowError("tag encountered with missing fields, need 'tag' and 'value', found: {\"tag\":1}");
+      cbor.resizeEncodingBuffer(0);
+    });
+
     for (const { name, data, cbor: cbor_representation } of examples) {
       it(`should encode for ${name}`, async () => {
         const serialized = cbor.serialize(data);
@@ -292,6 +319,7 @@ describe("cbor", () => {
           return {
             tag: id,
             value: translateTestData(tagValue),
+            [tagSymbol]: true,
           };
         default:
           throw new Error(`Unrecognized test scenario <expect> type ${type}.`);

packages/core/src/submodules/cbor/cbor.ts

@@ -17,8 +17,13 @@ export const cbor = {
     return decode(0, payload.length);
   },
   serialize(input: any) {
-    encode(input);
-    return toUint8Array();
+    try {
+      encode(input);
+      return toUint8Array();
+    } catch (e) {
+      toUint8Array(); // resets cursor.
+      throw e;
+    }
   },
   /**
    * @public

packages/core/src/submodules/cbor/parseCborBody.ts

@@ -4,6 +4,7 @@ import { HeaderBag as __HeaderBag, HttpResponse, SerdeContext as __SerdeContext,
 import { calculateBodyLength } from "@smithy/util-body-length-browser";
 
 import { cbor } from "./cbor";
+import { tagSymbol } from "./cbor-types";
 
 /**
  * @internal
@@ -27,8 +28,9 @@ export const parseCborBody = (streamBody: any, context: SerdeContext): any => {
 /**
  * @internal
  */
-export const dateToTag = (date: Date): { tag: 1; value: number } => {
+export const dateToTag = (date: Date): { tag: 1; value: number; [tagSymbol]: true } => {
   return {
+    [tagSymbol]: true,
     tag: 1,
     value: date.getTime() / 1000,
   };