feat(cdk): allow Tokens to be encoded as lists (#1144)

Add the ability for Tokens to be encoded into lists using the
`token.toList()` method.

This is useful for Tokens that intrinsically represent lists of strings,
so we can pass them around in the type system as `string[]` (and
interchange them with literal `string[]` values).

The encoding is reversible just like string encodings are reversible.
Contrary to strings encodings, concatenation operations are not
allowed (they cannot be applied after decoding since CloudFormation
does not have any list concatenation operators).

This change does not change any CloudFormation resources yet to
take advantage of the new encoding.

Implements the most important remaining part of #744.

Author: rix0rrr

packages/@aws-cdk/cdk/lib/cloudformation/fn.ts

@@ -87,6 +87,7 @@ export class FnJoin extends Fn {
   private readonly listOfValues: any[];
   // Cache for the result of resolveValues() - since it otherwise would be computed several times
   private _resolvedValues?: any[];
+  private canOptimize: boolean;
 
   /**
    * Creates an ``Fn::Join`` function.
@@ -103,11 +104,13 @@ export class FnJoin extends Fn {
     super('Fn::Join', [ delimiter, new Token(() => this.resolveValues()) ]);
     this.delimiter = delimiter;
     this.listOfValues = listOfValues;
+    this.canOptimize = true;
   }
 
   public resolve(): any {
-    if (this.resolveValues().length === 1) {
-      return this.resolveValues()[0];
+    const resolved = this.resolveValues();
+    if (this.canOptimize && resolved.length === 1) {
+      return resolved[0];
     }
     return super.resolve();
   }
@@ -120,6 +123,12 @@ export class FnJoin extends Fn {
   private resolveValues() {
     if (this._resolvedValues) { return this._resolvedValues; }
 
+    if (unresolved(this.listOfValues)) {
+      // This is a list token, don't resolve and also don't optimize.
+      this.canOptimize = false;
+      return this._resolvedValues = this.listOfValues;
+    }
+
     const resolvedValues = [...this.listOfValues.map(e => resolve(e))];
     let i = 0;
     while (i < resolvedValues.length) {

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

@@ -17,7 +17,8 @@ export const RESOLVE_METHOD = 'resolve';
  * semantics.
  */
 export class Token {
-  private tokenKey?: string;
+  private tokenStringification?: string;
+  private tokenListification?: string[];
 
   /**
    * Creates a token that resolves to `value`.
@@ -72,10 +73,10 @@ export class Token {
       return this.valueOrFunction.toString();
     }
 
-    if (this.tokenKey === undefined) {
-      this.tokenKey = TOKEN_STRING_MAP.register(this, this.displayName);
+    if (this.tokenStringification === undefined) {
+      this.tokenStringification = TOKEN_MAP.registerString(this, this.displayName);
     }
-    return this.tokenKey;
+    return this.tokenStringification;
   }
 
   /**
@@ -89,6 +90,30 @@ export class Token {
     throw new Error('JSON.stringify() cannot be applied to structure with a Token in it. Use a document-specific stringification method instead.');
   }
 
+  /**
+   * Return a string list representation of this token
+   *
+   * Call this if the Token intrinsically evaluates to a list of strings.
+   * If so, you can represent the Token in a similar way in the type
+   * system.
+   *
+   * Note that even though the Token is represented as a list of strings, you
+   * still cannot do any operations on it such as concatenation, indexing,
+   * or taking its length. The only useful operations you can do to these lists
+   * is constructing a `FnJoin` or a `FnSelect` on it.
+   */
+  public toList(): string[] {
+    const valueType = typeof this.valueOrFunction;
+    if (valueType === 'string' || valueType === 'number' || valueType === 'boolean') {
+      throw new Error('Got a literal Token value; cannot be encoded as a list.');
+    }
+
+    if (this.tokenListification === undefined) {
+      this.tokenListification = TOKEN_MAP.registerList(this, this.displayName);
+    }
+    return this.tokenListification;
+  }
+
   /**
    * Return a concated version of this Token in a string context
    *
@@ -103,12 +128,15 @@ export class Token {
 
 /**
  * Returns true if obj is a token (i.e. has the resolve() method or is a string
- * that includes token markers).
+ * that includes token markers), or it's a listifictaion of a Token string.
+ *
  * @param obj The object to test.
  */
 export function unresolved(obj: any): boolean {
   if (typeof(obj) === 'string') {
-    return TOKEN_STRING_MAP.createTokenString(obj).test();
+    return TOKEN_MAP.createStringTokenString(obj).test();
+  } else if (Array.isArray(obj) && obj.length === 1) {
+    return isListToken(obj[0]);
   } else {
     return typeof(obj[RESOLVE_METHOD]) === 'function';
   }
@@ -158,7 +186,7 @@ export function resolve(obj: any, prefix?: string[]): any {
   // string - potentially replace all stringified Tokens
   //
   if (typeof(obj) === 'string') {
-    return TOKEN_STRING_MAP.resolveMarkers(obj as string);
+    return TOKEN_MAP.resolveStringTokens(obj as string);
   }
 
   //
@@ -169,27 +197,31 @@ export function resolve(obj: any, prefix?: string[]): any {
     return obj;
   }
 
-  //
-  // tokens - invoke 'resolve' and continue to resolve recursively
-  //
-
-  if (unresolved(obj)) {
-    const value = obj[RESOLVE_METHOD]();
-    return resolve(value, path);
-  }
-
   //
   // arrays - resolve all values, remove undefined and remove empty arrays
   //
 
   if (Array.isArray(obj)) {
+    if (containsListToken(obj)) {
+      return TOKEN_MAP.resolveListTokens(obj);
+    }
+
     const arr = obj
       .map((x, i) => resolve(x, path.concat(i.toString())))
       .filter(x => typeof(x) !== 'undefined');
 
     return arr;
   }
 
+  //
+  // tokens - invoke 'resolve' and continue to resolve recursively
+  //
+
+  if (unresolved(obj)) {
+    const value = obj[RESOLVE_METHOD]();
+    return resolve(value, path);
+  }
+
   //
   // objects - deep-resolve all values
   //
@@ -221,6 +253,14 @@ export function resolve(obj: any, prefix?: string[]): any {
   return result;
 }
 
+function isListToken(x: any) {
+    return typeof(x) === 'string' && TOKEN_MAP.createListTokenString(x).test();
+}
+
+function containsListToken(xs: any[]) {
+  return xs.some(isListToken);
+}
+
 /**
  * Central place where we keep a mapping from Tokens to their String representation
  *
@@ -230,7 +270,7 @@ export function resolve(obj: any, prefix?: string[]): any {
  * All instances of TokenStringMap share the same storage, so that this process
  * works even when different copies of the library are loaded.
  */
-class TokenStringMap {
+class TokenMap {
   private readonly tokenMap: {[key: string]: Token};
 
   constructor() {
@@ -239,7 +279,7 @@ class TokenStringMap {
   }
 
   /**
-   * Generating a unique string for this Token, returning a key
+   * Generate a unique string for this Token, returning a key
    *
    * Every call for the same Token will produce a new unique string, no
    * attempt is made to deduplicate. Token objects should cache the
@@ -249,35 +289,56 @@ class TokenStringMap {
    * hint. This may be used to produce aesthetically pleasing and
    * recognizable token representations for humans.
    */
-  public register(token: Token, representationHint?: string): string {
-    const counter = Object.keys(this.tokenMap).length;
-    const representation = representationHint || `TOKEN`;
+  public registerString(token: Token, representationHint?: string): string {
+    const key = this.register(token, representationHint);
+    return `${BEGIN_STRING_TOKEN_MARKER}${key}${END_TOKEN_MARKER}`;
+  }
 
-    const key = `${representation}.${counter}`;
-    if (new RegExp(`[^${VALID_KEY_CHARS}]`).exec(key)) {
-      throw new Error(`Invalid characters in token representation: ${key}`);
-    }
+  /**
+   * Generate a unique string for this Token, returning a key
+   */
+  public registerList(token: Token, representationHint?: string): string[] {
+    const key = this.register(token, representationHint);
+    return [`${BEGIN_LIST_TOKEN_MARKER}${key}${END_TOKEN_MARKER}`];
+  }
 
-    this.tokenMap[key] = token;
-    return `${BEGIN_TOKEN_MARKER}${key}${END_TOKEN_MARKER}`;
+  /**
+   * Returns a `TokenString` for this string.
+   */
+  public createStringTokenString(s: string) {
+    return new TokenString(s, BEGIN_STRING_TOKEN_MARKER, `[${VALID_KEY_CHARS}]+`, END_TOKEN_MARKER);
   }
 
   /**
    * Returns a `TokenString` for this string.
    */
-  public createTokenString(s: string) {
-    return new TokenString(s, BEGIN_TOKEN_MARKER, `[${VALID_KEY_CHARS}]+`, END_TOKEN_MARKER);
+  public createListTokenString(s: string) {
+    return new TokenString(s, BEGIN_LIST_TOKEN_MARKER, `[${VALID_KEY_CHARS}]+`, END_TOKEN_MARKER);
   }
 
   /**
    * Replace any Token markers in this string with their resolved values
    */
-  public resolveMarkers(s: string): any {
-    const str = this.createTokenString(s);
+  public resolveStringTokens(s: string): any {
+    const str = this.createStringTokenString(s);
     const fragments = str.split(this.lookupToken.bind(this));
     return fragments.join();
   }
 
+  public resolveListTokens(xs: string[]): any {
+    // Must be a singleton list token, because concatenation is not allowed.
+    if (xs.length !== 1) {
+      throw new Error(`Cannot add elements to list token, got: ${xs}`);
+    }
+
+    const str = this.createListTokenString(xs[0]);
+    const fragments = str.split(this.lookupToken.bind(this));
+    if (fragments.length !== 1) {
+      throw new Error(`Cannot concatenate strings in a tokenized string array, got: ${xs[0]}`);
+    }
+    return fragments.values()[0];
+  }
+
   /**
    * Find a Token by key
    */
@@ -288,16 +349,30 @@ class TokenStringMap {
 
     return this.tokenMap[key];
   }
+
+  private register(token: Token, representationHint?: string): string {
+    const counter = Object.keys(this.tokenMap).length;
+    const representation = representationHint || `TOKEN`;
+
+    const key = `${representation}.${counter}`;
+    if (new RegExp(`[^${VALID_KEY_CHARS}]`).exec(key)) {
+      throw new Error(`Invalid characters in token representation: ${key}`);
+    }
+
+    this.tokenMap[key] = token;
+    return key;
+  }
 }
 
-const BEGIN_TOKEN_MARKER = '${Token[';
+const BEGIN_STRING_TOKEN_MARKER = '${Token[';
+const BEGIN_LIST_TOKEN_MARKER = '#{Token[';
 const END_TOKEN_MARKER = ']}';
 const VALID_KEY_CHARS = 'a-zA-Z0-9:._-';
 
 /**
  * Singleton instance of the token string map
  */
-const TOKEN_STRING_MAP = new TokenStringMap();
+const TOKEN_MAP = new TokenMap();
 
 /**
  * Interface that Token joiners implement
@@ -382,6 +457,10 @@ type Fragment =  StringFragment | TokenFragment;
 class TokenStringFragments {
   private readonly fragments = new Array<Fragment>();
 
+  public get length() {
+    return this.fragments.length;
+  }
+
   public values(): any[] {
     return this.fragments.map(f => f.type === 'token' ? resolve(f.token) : f.str);
   }