@sap/cds-compiler 2.11.4 → 2.12.0

package/lib/json/from-csn.js

@@ -705,7 +705,7 @@ let csnFilename = '';
 let virtualLine = 1;
 /** @type {CSN.Location[]} */
 let dollarLocations = [];
-let arrayLvlCnt = 0;
+let arrayLevelCount = 0;
 
 /**
  * @param {Object.<string, SchemaSpec>} specs
@@ -1164,6 +1164,12 @@ function symbol( id, spec, xsn ) { // for CSN property '#'
   xsn.sym = { id, location: location() };
 }
 
+// returns: false = no "...", true = "..." without UP TO, 'upTo' = "..." with UP TO
+function isEllipsis( val ) {
+  return val && typeof val === 'object' && '...' in val && Object.keys(val).length === 1 &&
+    (val['...'] === true || 'upTo');
+}
+
 function annoValue( val, spec ) {
   if (val == null)              // TODO: reject undefined
     return { val, literal: 'null', location: location() };
@@ -1171,22 +1177,39 @@ function annoValue( val, spec ) {
   if (lit !== 'object')
     return { val, literal: lit, location: location() };
   if (Array.isArray( val )) {
-    const ec = val.reduce((c, v) => ((v && v['...'] && Object.keys(v).length === 1) ? ++c : c), 0);
-    if (arrayLvlCnt === 0 && ec > 1) {
-      error( 'syntax-csn-duplicate-ellipsis', location(true), { code: '...' },
-             'Expected no more than one $(CODE)' );
+    let seenEllipsis = false;
+    if (arrayLevelCount > 0) {  // TODO: also inside structure (possible in CSN!)
+      if (val.some( isEllipsis )) {
+        error( 'syntax-csn-unexpected-ellipsis', location(true), { code: '...' },
+               'Unexpected $(CODE) in nested array' );
+      }
     }
-    if (arrayLvlCnt > 0 && ec > 0) {
-      error( 'syntax-csn-unexpected-ellipsis', location(true), { code: '...' },
-             'Unexpected $(CODE) in nested array' );
+    else {
+      for (const item of val) {
+        if (seenEllipsis !== true) {
+          seenEllipsis = isEllipsis( item ) || seenEllipsis;
+        }
+        else if (isEllipsis( item )) { // with or without UP TO
+          // error position at the beginning of the array, but that is fine
+          error( 'syntax-csn-duplicate-ellipsis', location(true), { code: '...' },
+                 'Expected no more than one $(CODE)' );
+          break;
+        }
+      }
     }
-    arrayLvlCnt++;
+    arrayLevelCount++;
     const retval = {
       location: location(),
       val: arrayOf( annoValue )( val, spec ),
       literal: 'array',
     };
-    arrayLvlCnt--;
+    arrayLevelCount--;
+    if (seenEllipsis === 'upTo') {
+      error( 'syntax-csn-expecting-ellipsis', location(true), // at closing bracket
+             { code: '... up to', newcode: '...' },
+             // TODO: should we be more CSN specific in the message?
+             'Expecting an array item $(NEWCODE) after an item with $(CODE)' );
+    }
     return retval;
   }
   if (typeof val['#'] === 'string') {
@@ -1205,12 +1228,19 @@ function annoValue( val, spec ) {
       return refSplit( val['='], '=' );
     }
   }
-  else if (val['...'] && Object.keys(val).length === 1) {
-    return {
+  else if (val['...'] && Object.keys( val ).length === 1) {
+    // TODO: only if not nested - see error above
+    ++virtualLine;
+    const ell = val['...'];
+    const r = {
       val: '...',
       literal: 'token',
       location: location(),
     };
+    if (ell !== true)
+      r.upTo = annoValue( ell, schema['@'] );
+    ++virtualLine;
+    return r;
   }
   const struct = Object.create(null);
   ++virtualLine;
@@ -1354,12 +1384,11 @@ function exprArgs( cond, spec, xsn, csn ) {
 
 function condition( cond, spec ) {
   const loc = location();
-  const x = {
+  return {
     op: { val: 'xpr', location: loc },
     args: exprArgs( cond, spec ),
     location: loc,
   };
-  return x;
 }
 
 function vZeroValue( obj, spec, xsn ) {
@@ -1666,8 +1695,8 @@ function pushLocation( obj ) {
     error( 'syntax-csn-expected-object', location(true), { prop: '$location' } );
   }
   // hidden feature: string $location
-  const m = /:(\d+)(?::(\d+)(?:-[0-9-]+)?)?$/.exec( loc ); // extra - at end for .refloc
-  if (!m) {
+  const m = /:(\d+)(?::(\d+))?$/.exec( loc ); // extra '^'s at end deliberately left out
+  if (!m) {                     // without location or with '^'s: do not use
     dollarLocations.push( null );
   }
   else {
@@ -1710,8 +1739,10 @@ function toXsn( csn, filename, options, messageFunctions ) {
   csnFilename = filename;
   virtualLine = 1;
   dollarLocations = [];
+  arrayLevelCount = 0;
   inExtensions = null;
   vocabInDefinitions = null;
+
   const xsn = { $frontend: 'json' };
 
   // eslint-disable-next-line object-curly-newline

package/lib/json/to-csn.js

@@ -100,7 +100,7 @@ const transformers = {
   offset: expression,
   on: onCondition,
   // definitions, extensions, members ----------------------------------------
-  returns: definition,          // storing the return type of actions
+  returns,                      // storing the return type of actions
   notNull: value,
   default: expression,
   // targetElement: ignore,     // special display of foreign key, renameTo: select
@@ -556,7 +556,7 @@ function sources( srcDict, csn ) {
   }
 }
 
-function attachAnnotations( annotate, prop, dict, inferred, returns = false ) {
+function attachAnnotations( annotate, prop, dict, inferred, insideReturns = false ) {
   const annoDict = Object.create( dictionaryPrototype );
   for (const name in dict) {
     const elem = dict[name];
@@ -574,7 +574,7 @@ function attachAnnotations( annotate, prop, dict, inferred, returns = false ) {
       annoDict[name] = sub;
   }
   if (Object.keys( annoDict ).length) {
-    if (returns)
+    if (insideReturns)
       annotate.returns = { elements: annoDict };
     else
       annotate[prop] = annoDict;
@@ -674,7 +674,8 @@ function enumDict( dict, csn, node ) {
     // no 'elements' with SELECT or inferred elements with gensrc;
     // hidden or visible 'elements' will be set in query()
     return undefined;
-  if (universalCsn && node.type && !node.type.$inferred && node.$expand === 'annotate')
+  if (universalCsn && node.type && !node.type.$inferred && node.$expand === 'annotate' &&
+      node.type._artifact && !node.type._artifact.builtin)
     // derived type of enum type with individual annotations: also set $origin
     csn.$origin = originRef( node.type._artifact );
   return insertOrderDict( dict );
@@ -834,6 +835,14 @@ function foreignKeys( dict, csn, node ) {
   csn.keys = keys;
 }
 
+function returns( art, csn, _node, prop ) {
+  // TODO: currently, the `returns` structure might just have been created by the propagator
+  // if that is the case, there should be no reason to store it in universal CSN
+  if (universalCsn && art.$inferred === 'proxy')
+    return undefined;
+  return definition( art, csn, _node, prop );
+}
+
 function definition( art, _csn, _node, prop ) {
   if (!art || typeof art !== 'object')
     return undefined;           // TODO: complain with strict
@@ -1163,7 +1172,7 @@ function value( node ) {
   if (node.literal === 'array')
     return node.val.map( value );
   if (node.literal === 'token' && node.val === '...')
-    return extra( { '...': true } );
+    return extra( { '...': !node.upTo || value( node.upTo ) } );
   if (node.literal !== 'struct')
     // no val (undefined) as true only for annotation values (and struct elem values)
     return node.name && !('val' in node) || node.val;
@@ -1516,6 +1525,9 @@ function compactExpr( e ) {     // TODO: options
   return e && expression( e, true );
 }
 
+/**
+ * @param {CSN.Options} options
+ */
 function initModuleVars( options = { csnFlavor: 'gensrc' } ) {
   gensrcFlavor = options.parseCdl || options.csnFlavor === 'gensrc' ||
                  options.toCsn && options.toCsn.flavor === 'gensrc';

package/lib/language/antlrParser.js

@@ -112,7 +112,7 @@ const rules = {
   expr: { func: 'conditionEOF', returns: 'cond' }, // yes, condition
 };
 
-function parse( source, filename = '<undefined>.cds', options = {}, messageFunctions, rule = 'cdl' ) {
+function parse( source, filename = '<undefined>.cds', options = {}, messageFunctions = null, rule = 'cdl' ) {
   const lexer = new Lexer( new antlr4.InputStream(source) );
   const tokenStream = new RewriteTypeTokenStream(lexer);
   /** @type {object} */
@@ -166,8 +166,8 @@ function parse( source, filename = '<undefined>.cds', options = {}, messageFunct
   // Do not warn if docComments are explicitly disabled.
   if (options.docComment !== false) {
     for (const token of tokenStream.tokens) {
-      if (token.channel === antlr4.Token.HIDDEN_CHANNEL && token.type === parser.constructor.DocComment && !token.isUsed) {
-        messageFunctions.info('syntax-ignoring-doc-comment', parser.multiLineTokenLocation(token), {},
+      if (token.type === parser.constructor.DocComment && !token.isUsed) {
+        messageFunctions.info('syntax-ignoring-doc-comment', parser.tokenLocation(token), {},
           "Ignoring doc-comment as it does not belong to any artifact");
       }
     }

package/lib/language/docCommentParser.js

@@ -69,7 +69,7 @@ function parseDocComment(comment) {
  * @param {string} content
  */
 function isWhiteSpaceOnly(content) {
-  return content.trim().length === 0;
+  return /^\s*$/.test(content);
 }
 
 /**

package/lib/language/genericAntlrParser.js

@@ -11,14 +11,17 @@ const { ATNState } = require('antlr4/atn/ATNState');
 const { dictAdd, dictAddArray } = require('../base/dictionaries');
 const locUtils = require('../base/location');
 const { parseDocComment } = require('./docCommentParser');
+const { parseMultiLineStringLiteral } = require('./multiLineStringParser');
 const { functionsWithoutParens, specialFunctions } = require('../compiler/builtins');
 
 
 // Push message `msg` with location `loc` to array of errors:
 function _message( parser, severity, id, loc, ...args ) {
   const msg = parser.$messageFunctions[severity]; // set in antlrParser.js
-  return msg( id,
-              (loc instanceof antlr4.CommonToken) ? parser.tokenLocation(loc) : loc, ...args );
+  if (loc instanceof antlr4.CommonToken) {
+    loc = parser.tokenLocation(loc);
+  }
+  return msg( id, loc, ...args );
 }
 
 // Class which is to be used as grammar option with
@@ -49,7 +52,7 @@ GenericAntlrParser.prototype = Object.assign(
     attachLocation,
     startLocation,
     tokenLocation,
-    multiLineTokenLocation,
+    valueWithTokenLocation,
     previousTokenAtLocation,
     combinedLocation,
     surroundByParens,
@@ -84,6 +87,7 @@ GenericAntlrParser.prototype = Object.assign(
     isStraightBefore,
     meltKeywordToIdentifier,
     prepareGenericKeywords,
+    parseMultiLineStringLiteral,
     constructor: GenericAntlrParser, // keep this last
   }
 );
@@ -290,62 +294,66 @@ function startLocation( token = this._ctx.start ) {
  *
  * @param {object} token
  * @param {object} endToken
- * @param {any} val
+ * @return {CSN.Location}
  */
-function tokenLocation( token, endToken, val ) {
+function tokenLocation( token, endToken = null ) {
   if (!token)
     return undefined;
   if (!endToken)                // including null
     endToken = token;
+
   /** @type {CSN.Location} */
-  const r = {
+  const loc = {
     file: this.filename,
     line: token.line,
     col: token.column + 1,
-    // we only have single-line tokens, except for docComments
+    // Default for single line tokens
     endLine: endToken.line,
     endCol: endToken.stop - endToken.start + endToken.column + 2, // after the last char (special for EOF?)
-  };
-  if (val !== undefined)
-    return { location: r, val };
-  return r;
+  }
+
+  // This check is done for performance reason. No need to access a token's
+  // data if we know that it spans only one single line.
+  const isMultiLineToken = (
+    endToken.type === this.constructor.DocComment ||
+    endToken.type === this.constructor.String ||
+    endToken.type === this.constructor.UnterminatedLiteral
+  );
+  if (isMultiLineToken) {
+    // Count the number of newlines in the token.
+    const source = endToken.source[1].data;
+    let newLineCount = 0;
+    let lastNewlineIndex = endToken.start;
+    for (let i = endToken.start; i < endToken.stop; i++) {
+      // Note: We do NOT check for CR, LS, and PS (/[\r\u2028\u2029]/)
+      //       because ANTLR only uses LF for line break detection.
+      if (source[i] === 10) { // code point of '\n'
+        newLineCount++;
+        lastNewlineIndex = i;
+      }
+    }
+    if (newLineCount > 0) {
+      loc.endLine = endToken.line + newLineCount;
+      loc.endCol = endToken.stop - lastNewlineIndex + 1;
+    }
+  }
+
+  return loc;
 }
 
 /**
- * Return location of `token`.  In contrast to `tokenLocation()`, this function
- * can handle multiline tokens.
+ * Return `val` with the location of `token`.  If `endToken` is provided, use its end
+ * location as end location in the result.
+ *
+ * @param {object} startToken
+ * @param {object} endToken
+ * @param {any} val
  */
-function multiLineTokenLocation(token, val) {
-  if (!token)
+function valueWithTokenLocation( val, startToken, endToken = null ) {
+  if (!startToken)
     return undefined;
-
-  // Count the number of newlines in the token.
-  const source = token.source[1].data;
-  let newLineCount = 0;
-  let lastNewlineIndex = token.start;
-  for (let i = token.start; i < token.stop; i++) {
-    // Note: We do NOT check for CR, LS, and PS (/[\r\u2028\u2029]/)
-    //       because ANTLR only uses LF for line break detection.
-    if (source[i] === 10) { // ASCII code for '\n'
-      newLineCount++;
-      lastNewlineIndex = i;
-    }
-  }
-  if (newLineCount === 0)
-    // endCol calculation below requires at least one newLine.
-    return this.tokenLocation(token, token, val);
-
-  /** @type {CSN.Location} */
-  const r = {
-    file: this.filename,
-    line: token.line,
-    col: token.column + 1,
-    endLine: token.line + newLineCount,
-    endCol: token.stop - lastNewlineIndex + 1, // after the last char (special for EOF?)
-  };
-  if (val !== undefined)
-    return { location: r, val };
-  return r;
+  const loc = this.tokenLocation( startToken, endToken );
+  return { location: loc, val };
 }
 
 function previousTokenAtLocation( location ) {
@@ -406,7 +414,7 @@ function docComment( node ) {
     this.warning( 'syntax-duplicate-doc-comment', token, {},
                   'Repeated doc comment - previous doc is replaced' );
   }
-  node.doc = this.multiLineTokenLocation( token, parseDocComment( token.text ) );
+  node.doc = this.valueWithTokenLocation( parseDocComment( token.text ), token );
 }
 
 // Classify token (identifier category) for implicit names,
@@ -507,7 +515,7 @@ function valuePathAst( ref ) {
 // If a '-' is directly before an unsigned number, consider it part of the number;
 // otherwise (including for '+'), represent it as extra unary prefix operator.
 function signedExpression( signToken, expr ) {
-  const sign = this.tokenLocation( signToken, undefined, signToken.text );
+  const sign = this.valueWithTokenLocation( signToken.text, signToken );
   const nval =
         (signToken.text === '-' &&
         expr && // expr may be null if `-` rule can't be parsed
@@ -557,8 +565,16 @@ function numberLiteral( token, sign, text = token.text ) {
 function quotedLiteral( token, literal ) {
   /** @type {CSN.Location} */
   const location = this.tokenLocation( token );
-  const pos = token.text.search( '\'' ) + 1; // pos of char after quote
-  const val = token.text.slice( pos, -1 ).replace( /''/g, '\'' );
+  let pos;
+  let val;
+
+  if (token.text.startsWith('`')) {
+    val = this.parseMultiLineStringLiteral(token);
+    literal = 'string';
+  } else {
+    pos = token.text.search( '\'' ) + 1; // pos of char after quote
+    val = token.text.slice( pos, -1 ).replace( /''/g, '\'' );
+  }
 
   if (!literal)
     literal = token.text.slice( 0, pos - 1 ).toLowerCase();
@@ -588,6 +604,7 @@ function quotedLiteral( token, literal ) {
   };
 
   function atChar(i) {
+    // Is only used with single-line strings.
     return location.col + pos + i;
   }
 }
@@ -730,7 +747,7 @@ function assignProps( target, annos = [], props = null, location = null) {
   for (const key in props) {
     let val = props[key];
     if (val instanceof antlr4.CommonToken)
-      val = this.tokenLocation( val, undefined, true);
+      val = this.valueWithTokenLocation( true, val);
     // only copy properties which are not undefined, null, {} or []
     if (val != null &&
         (typeof val !== 'object' ||
@@ -744,15 +761,15 @@ function assignProps( target, annos = [], props = null, location = null) {
 
 // Create AST node for prefix operator `op` and arguments `args`
 function createPrefixOp( token, args ) {
-  const op = this.tokenLocation( token, undefined, token.text.toLowerCase() );
+  const op = this.valueWithTokenLocation( token.text.toLowerCase(), token );
   return { op, args, location: this.combinedLocation( op, args[args.length - 1] ) };
 }
 
 // Create AST node for binary operator `op` and arguments `args`
 function leftAssocBinaryOp( left, opToken, eToken, right, extraProp = 'quantifier' ) {
-  const op = this.tokenLocation( opToken, undefined, opToken.text.toLowerCase() );
+  const op = this.valueWithTokenLocation( opToken.text.toLowerCase() , opToken);
   const extra = eToken
-        ? this.tokenLocation( eToken, undefined, eToken.text.toLowerCase() )
+        ? this.valueWithTokenLocation( eToken.text.toLowerCase(), eToken )
         : undefined;
   if (!left.$parens &&
       (left.op && left.op.val) === (op && op.val) &&