fix(core): detect and resolve stringified number tokens

Number tokens are encoded as a range of very large negative numbers (for
example: -1.888154589709072e+289). When these are naively stringified,
the `resolve()` method doesn't recognize and translate them anymore,
and these numbers end up in the target template in a confusing way.

However, recognizing them is actually not that hard and can be done
using a regex. We can then do the token resolution appropriately, making
it so that construct authors do not have to call
`Tokenization.stringifyNumber()` anymore in order to support
stringification of number values.

Fixes #19546.

packages/@aws-cdk/aws-stepfunctions-tasks/test/batch/submit-job.test.ts

@@ -169,6 +169,33 @@ test('supports tokens', () => {
   });
 });
 
+test('container overrides are tokens', () => {
+  // WHEN
+  const task = new BatchSubmitJob(stack, 'Task', {
+    jobDefinitionArn: batchJobDefinition.jobDefinitionArn,
+    jobName: 'JobName',
+    jobQueueArn: batchJobQueue.jobQueueArn,
+    containerOverrides: {
+      memory: cdk.Size.mebibytes(sfn.JsonPath.numberAt('$.asdf')),
+    },
+  });
+
+  // THEN
+  expect(stack.resolve(task.toStateJson())).toEqual({
+    Type: 'Task',
+    Resource: { 'Fn::Join': ['', ['arn:', { Ref: 'AWS::Partition' }, ':states:::batch:submitJob.sync']] },
+    End: true,
+    Parameters: {
+      JobDefinition: { Ref: 'JobDefinition24FFE3ED' },
+      JobName: 'JobName',
+      JobQueue: { Ref: 'JobQueueEE3AD499' },
+      ContainerOverrides: {
+        ResourceRequirements: [{ 'Type': 'MEMORY', 'Value.$': '$.asdf' }],
+      },
+    },
+  });
+});
+
 test('supports passing task input into payload', () => {
   // WHEN
   const task = new BatchSubmitJob(stack, 'Task', {

packages/@aws-cdk/core/lib/private/cloudformation-lang.ts

@@ -43,8 +43,8 @@ export class CloudFormationLang {
 
     // Some case analysis to produce minimal expressions
     if (parts.length === 1) { return parts[0]; }
-    if (parts.length === 2 && typeof parts[0] === 'string' && typeof parts[1] === 'string') {
-      return parts[0] + parts[1];
+    if (parts.length === 2 && ['string', 'number'].includes(typeof parts[0]) && ['string', 'number'].includes(typeof parts[1])) {
+      return `${parts[0]}${parts[1]}`;
     }
 
     // Otherwise return a Join intrinsic (already in the target document language to avoid taking
@@ -323,8 +323,8 @@ export function minimalCloudFormationJoin(delimiter: string, values: any[]): any
     const el = values[i];
     if (isSplicableFnJoinIntrinsic(el)) {
       values.splice(i, 1, ...el['Fn::Join'][1]);
-    } else if (i > 0 && isPlainString(values[i - 1]) && isPlainString(values[i])) {
-      values[i - 1] += delimiter + values[i];
+    } else if (i > 0 && isConcatable(values[i - 1]) && isConcatable(values[i])) {
+      values[i - 1] = `${values[i-1]}${delimiter}${values[i]}`;
       values.splice(i, 1);
     } else {
       i += 1;
@@ -333,8 +333,8 @@ export function minimalCloudFormationJoin(delimiter: string, values: any[]): any
 
   return values;
 
-  function isPlainString(obj: any): boolean {
-    return typeof obj === 'string' && !Token.isUnresolved(obj);
+  function isConcatable(obj: any): boolean {
+    return ['string', 'number'].includes(typeof obj) && !Token.isUnresolved(obj);
   }
 
   function isSplicableFnJoinIntrinsic(obj: any): boolean {

packages/@aws-cdk/core/lib/private/encoding.ts

@@ -14,7 +14,9 @@ const QUOTED_BEGIN_STRING_TOKEN_MARKER = regexQuote(BEGIN_STRING_TOKEN_MARKER);
 const QUOTED_BEGIN_LIST_TOKEN_MARKER = regexQuote(BEGIN_LIST_TOKEN_MARKER);
 const QUOTED_END_TOKEN_MARKER = regexQuote(END_TOKEN_MARKER);
 
-const STRING_TOKEN_REGEX = new RegExp(`${QUOTED_BEGIN_STRING_TOKEN_MARKER}([${VALID_KEY_CHARS}]+)${QUOTED_END_TOKEN_MARKER}`, 'g');
+const STRINGIFIED_NUMBER_PATTERN = '-1\\.\\d{16}e\\+289';
+
+const STRING_TOKEN_REGEX = new RegExp(`${QUOTED_BEGIN_STRING_TOKEN_MARKER}([${VALID_KEY_CHARS}]+)${QUOTED_END_TOKEN_MARKER}|(${STRINGIFIED_NUMBER_PATTERN})`, 'g');
 const LIST_TOKEN_REGEX = new RegExp(`${QUOTED_BEGIN_LIST_TOKEN_MARKER}([${VALID_KEY_CHARS}]+)${QUOTED_END_TOKEN_MARKER}`, 'g');
 
 /**
@@ -52,7 +54,7 @@ export class TokenString {
         ret.addLiteral(this.str.substring(rest, m.index));
       }
 
-      ret.addToken(lookup(m[1]));
+      ret.addToken(lookup(m[1] ?? m[2]));
 
       rest = this.re.lastIndex;
       m = this.re.exec(this.str);
@@ -218,3 +220,12 @@ export function extractTokenDouble(encoded: number): number | undefined {
   return ints[0] + shl32(ints[1] & 0xFFFF);
   /* eslint-enable no-bitwise */
 }
+
+const STRINGIFIED_NUMBER_REGEX = new RegExp(STRINGIFIED_NUMBER_PATTERN);
+
+/**
+ * Return whether the given string contains accidentally stringified number tokens
+ */
+export function stringContainsNumberTokens(x: string) {
+  return !!x.match(STRINGIFIED_NUMBER_REGEX);
+}

packages/@aws-cdk/core/lib/private/token-map.ts

@@ -177,8 +177,12 @@ export class TokenMap {
 
   private registerNumberKey(token: IResolvable): number {
     const counter = this.tokenCounter++;
+    const dbl = createTokenDouble(counter);
+    // Register in the number map, as well as a string representation of that token
+    // in the string map.
     this.numberTokenMap.set(counter, token);
-    return createTokenDouble(counter);
+    this.stringTokenMap.set(`${dbl}`, token);
+    return dbl;
   }
 }
 

packages/@aws-cdk/core/lib/string-fragments.ts

@@ -1,5 +1,5 @@
 import { IFragmentConcatenator, IResolvable } from './resolvable';
-import { isResolvableObject } from './token';
+import { isResolvableObject, Token } from './token';
 
 /**
  * Result of the split of a string with Tokens
@@ -71,8 +71,10 @@ export class TokenizedStringFragments {
           const mapped = mapper.mapToken(f.token);
           if (isResolvableObject(mapped)) {
             ret.addToken(mapped);
-          } else {
+          } else if (Token.isUnresolved(mapped)) {
             ret.addIntrinsic(mapped);
+          } else {
+            ret.addLiteral(mapped);
           }
           break;
         case 'intrinsic':

packages/@aws-cdk/core/test/tokens.test.ts

@@ -1,5 +1,5 @@
-import { Fn, isResolvableObject, Lazy, Stack, Token, Tokenization } from '../lib';
-import { createTokenDouble, extractTokenDouble } from '../lib/private/encoding';
+import { CfnResource, Fn, isResolvableObject, Lazy, Stack, Token, Tokenization } from '../lib';
+import { createTokenDouble, extractTokenDouble, stringContainsNumberTokens } from '../lib/private/encoding';
 import { Intrinsic } from '../lib/private/intrinsic';
 import { findTokens } from '../lib/private/resolve';
 import { IResolvable } from '../lib/resolvable';
@@ -482,15 +482,12 @@ describe('tokens', () => {
       expect(() => {
         resolve({ value: encoded[0] });
       }).toThrow(/Found an encoded list/);
-
-
     });
   });
 
   describe('number encoding', () => {
     test('basic integer encoding works', () => {
       expect(16).toEqual(extractTokenDouble(createTokenDouble(16)));
-
     });
 
     test('arbitrary integers can be encoded, stringified, and recovered', () => {
@@ -504,16 +501,12 @@ describe('tokens', () => {
         const decoded = extractTokenDouble(roundtripped);
         expect(decoded).toEqual(x);
       }
-
-
     });
 
     test('arbitrary numbers are correctly detected as not being tokens', () => {
       expect(undefined).toEqual(extractTokenDouble(0));
       expect(undefined).toEqual(extractTokenDouble(1243));
       expect(undefined).toEqual(extractTokenDouble(4835e+532));
-
-
     });
 
     test('can number-encode and resolve Token objects', () => {
@@ -528,8 +521,34 @@ describe('tokens', () => {
       // THEN
       const resolved = resolve({ value: encoded });
       expect(resolved).toEqual({ value: 123 });
+    });
+
+    test('regex detects all stringifications of encoded tokens', () => {
+      expect(stringContainsNumberTokens(`${createTokenDouble(0)}`)).toBeTruthy();
+      expect(stringContainsNumberTokens(`${createTokenDouble(Math.pow(2, 48) - 1)}`)).toBeTruthy(); // MAX_ENCODABLE_INTEGER
+      expect(stringContainsNumberTokens('1234')).toBeFalsy();
+    });
+
+    test('handle stringified number token', () => {
+      // GIVEN
+      const tok = `the answer is: ${Lazy.number({ produce: () => 86 })}`;
 
+      // THEN
+      expect(resolve({ value: `${tok}` })).toEqual({
+        value: 'the answer is: 86',
+      });
+    });
 
+    test('handle stringified number reference', () => {
+      const stack = new Stack();
+      const res = new CfnResource(stack, 'Resource', { type: 'My::Resource' });
+      // GIVEN
+      const tok = `the answer is: ${Token.asNumber(res.ref)}`;
+
+      // THEN
+      expect(resolve({ value: `${tok}` })).toEqual({
+        value: { 'Fn::Join': ['', ['the answer is: ', { Ref: 'Resource' }]] },
+      });
     });
   });
 
@@ -694,25 +713,21 @@ describe('tokens', () => {
   describe('stringifyNumber', () => {
     test('converts number to string', () => {
       expect(Tokenization.stringifyNumber(100)).toEqual('100');
-
     });
 
     test('converts tokenized number to string', () => {
       expect(resolve(Tokenization.stringifyNumber({
         resolve: () => 100,
       } as any))).toEqual('100');
-
     });
 
     test('string remains the same', () => {
       expect(Tokenization.stringifyNumber('123' as any)).toEqual('123');
-
     });
 
     test('Ref remains the same', () => {
       const val = { Ref: 'SomeLogicalId' };
       expect(Tokenization.stringifyNumber(val as any)).toEqual(val);
-
     });
 
     test('lazy Ref remains the same', () => {
@@ -791,3 +806,4 @@ function tokensThatResolveTo(value: any): Token[] {
 function resolve(x: any) {
   return new Stack().resolve(x);
 }
+