Allow module aliases to use escaped identifiers as the alias name.

allevato · allevato · commit 2b0f9aa76500 · 2025-03-11T17:18:43.000-04:00
The original module names themselves must still be valid unescaped identifiers; most of the serialization logic in the compiler depends on the name of a module matching its name on the file system, and it would be very complex to turn escaped identifiers into file-safe names.
diff --git a/include/swift/Parse/Lexer.h b/include/swift/Parse/Lexer.h
@@ -400,6 +400,10 @@ class Lexer {
   // its name into runtime metadata.
   static bool identifierMustAlwaysBeEscaped(StringRef str);
 
+  /// Determines if the given string is a valid non-operator
+  /// identifier if it were surrounded by backticks.
+  static bool isValidAsEscapedIdentifier(StringRef identifier);
+
   /// Determine the token kind of the string, given that it is a valid
   /// non-operator identifier. Return tok::identifier if the string is not a
   /// reserved word.
diff --git a/lib/Frontend/ArgsToFrontendOptionsConverter.cpp b/lib/Frontend/ArgsToFrontendOptionsConverter.cpp
@@ -936,9 +936,12 @@ bool ModuleAliasesConverter::computeModuleAliases(std::vector<std::string> args,
     // it should be called only once
     options.ModuleAliasMap.clear();
 
-    auto validate = [&options, &diags](StringRef value, bool allowModuleName) -> bool
-    {
-      if (!allowModuleName) {
+    // validatingModuleName should be true if validating the alias target (an
+    // actual module name), or true if validating the alias name (which can be
+    // an escaped identifier).
+    auto validate = [&options, &diags](StringRef value,
+                                       bool validatingModuleName) -> bool {
+      if (!validatingModuleName) {
         if (value == options.ModuleName ||
             value == options.ModuleABIName ||
             value == options.ModuleLinkName ||
@@ -947,7 +950,8 @@ bool ModuleAliasesConverter::computeModuleAliases(std::vector<std::string> args,
           return false;
         }
       }
-      if (!Lexer::isIdentifier(value)) {
+      if ((validatingModuleName && !Lexer::isIdentifier(value)) ||
+          !Lexer::isValidAsEscapedIdentifier(value)) {
         diags.diagnose(SourceLoc(), diag::error_bad_module_name, value, false);
         return false;
       }
diff --git a/lib/Parse/Lexer.cpp b/lib/Parse/Lexer.cpp
@@ -581,26 +581,40 @@ static bool isValidIdentifierStartCodePoint(uint32_t c) {
   return true;
 }
 
+static bool isForbiddenRawIdentifierWhitespace(uint32_t c) {
+  if ((c >= 0x0009 && c <= 0x000D) ||
+      c == 0x0085 ||
+      c == 0x00A0 ||
+      c == 0x1680 ||
+      (c >= 0x2000 && c <= 0x200A) ||
+      (c >= 0x2028 && c <= 0x2029) ||
+      c == 0x202F ||
+      c == 0x205F ||
+      c == 0x3000)
+    return true;
+
+  return false;
+}
+
+static bool isPermittedRawIdentifierWhitespace(uint32_t c) {
+  return c == 0x0020 || c == 0x200E || c == 0x200F;
+}
+
 static bool isValidIdentifierEscapedCodePoint(uint32_t c) {
   // An escaped identifier is terminated by a backtick, and the backslash is
   // reserved for possible future escaping.
   if (c == '`' || c == '\\')
     return false;
 
+  if ((c >= 0x0000 && c <= 0x001F) || c == 0x007F)
+    return false;
+
   // This is the set of code points satisfying the `White_Space` property,
   // excluding the set satisfying the `Pattern_White_Space` property, and
   // excluding any other ASCII non-printables and Unicode separators. In
   // other words, the only whitespace code points allowed in a raw
   // identifier are U+0020, and U+200E/200F (LTR/RTL marks).
-  if ((c >= 0x0009 && c <= 0x000D) ||
-      c == 0x0085 ||
-      c == 0x00A0 ||
-      c == 0x1680 ||
-      (c >= 0x2000 && c <= 0x200A) ||
-      (c >= 0x2028 && c <= 0x2029) ||
-      c == 0x202F ||
-      c == 0x205F ||
-      c == 0x3000)
+  if (isForbiddenRawIdentifierWhitespace(c))
     return false;
 
   return true;
@@ -644,6 +658,17 @@ static bool advanceIfValidContinuationOfOperator(char const *&ptr,
   return advanceIf(ptr, end, Identifier::isOperatorContinuationCodePoint);
 }
 
+/// Returns true if the given string is entirely whitespace (considering only
+/// those whitespace code points permitted in raw identifiers).
+static bool isEntirelyWhitespace(StringRef string) {
+  if (string.empty()) return false;
+  char const *p = string.data(), *end = string.end();
+  if (!advanceIf(p, end, isPermittedRawIdentifierWhitespace))
+    return false;
+  while (p < end && advanceIf(p, end, isPermittedRawIdentifierWhitespace));
+  return p == end;
+}
+
 bool Lexer::isIdentifier(StringRef string) {
   if (string.empty()) return false;
   char const *p = string.data(), *end = string.end();
@@ -665,6 +690,19 @@ bool Lexer::identifierMustAlwaysBeEscaped(StringRef str) {
   return mustEscape;
 }
 
+bool Lexer::isValidAsEscapedIdentifier(StringRef string) {
+  if (string.empty())
+    return false;
+  char const *p = string.data(), *end = string.end();
+  if (!advanceIfValidEscapedIdentifier(p, end))
+    return false;
+  while (p < end && advanceIfValidEscapedIdentifier(p, end))
+    ;
+  if (p != end)
+    return false;
+  return !isEntirelyWhitespace(string);
+}
+
 /// Determines if the given string is a valid operator identifier,
 /// without escaping characters.
 bool Lexer::isOperator(StringRef string) {
@@ -2271,10 +2309,10 @@ void Lexer::lexEscapedIdentifier() {
   while (advanceIfValidEscapedIdentifier(CurPtr, BufferEnd))
     ;
 
-  // If we have the terminating "`", it's an escaped identifier, unless it
-  // contained only operator characters.
-  if (*CurPtr == '`' &&
-      !isOperator(StringRef(IdentifierStart, CurPtr - IdentifierStart))) {
+  // If we have the terminating "`", it's an escaped/raw identifier, unless it
+  // contained only operator characters or was entirely whitespace.
+  StringRef IdStr(IdentifierStart, CurPtr - IdentifierStart);
+  if (*CurPtr == '`' && !isOperator(IdStr) && !isEntirelyWhitespace(IdStr)) {
     ++CurPtr;
     formEscapedIdentifierToken(Quote);
     return;
diff --git a/test/ClangImporter/module-alias-escaped-identifier.swift b/test/ClangImporter/module-alias-escaped-identifier.swift
@@ -0,0 +1,11 @@
+// RUN: %empty-directory(%t)
+// RUN: %target-swift-frontend(mock-sdk: %clang-importer-sdk) -module-alias "//my/project:uncommon_name=CommonName" -typecheck -I %S/Inputs/custom-modules %s -Rmodule-loading 2> %t/load-result.output
+
+// RUN: %FileCheck %s -input-file %t/load-result.output -check-prefix CHECK-FOO
+// CHECK-FOO: import `//my/project:uncommon_name`
+// CHECK-FOO-NEXT: remark: loaded module 'CommonName'
+
+import `//my/project:uncommon_name`
+
+_ = MyStruct()
+_ = `//my/project:uncommon_name`.MyStruct()
diff --git a/test/ClangImporter/module-alias.swift b/test/ClangImporter/module-alias.swift
@@ -0,0 +1,11 @@
+// RUN: %empty-directory(%t)
+// RUN: %target-swift-frontend(mock-sdk: %clang-importer-sdk) -module-alias UncommonName=CommonName -typecheck -I %S/Inputs/custom-modules %s -Rmodule-loading 2> %t/load-result.output
+
+// RUN: %FileCheck %s -input-file %t/load-result.output -check-prefix CHECK-FOO
+// CHECK-FOO: import UncommonName
+// CHECK-FOO-NEXT: remark: loaded module 'CommonName'
+
+import UncommonName
+
+_ = MyStruct()
+_ = UncommonName.MyStruct()
diff --git a/test/Frontend/module-alias-escaped-identifier.swift b/test/Frontend/module-alias-escaped-identifier.swift
@@ -0,0 +1,40 @@
+/// Test the -module-alias flag with an escaped identifier alias.
+
+// RUN: %empty-directory(%t)
+// RUN: %{python} %utils/split_file.py -o %t %s
+
+/// Create a module Bar
+// RUN: %target-swift-frontend -module-name Bar %t/FileBar.swift -emit-module -emit-module-path %t/Bar.swiftmodule
+
+/// Check Bar.swiftmodule is created
+// RUN: test -f %t/Bar.swiftmodule
+
+/// Create a module Foo that imports `//my/project:cat` with -module-alias "//my/project:cat=Bar" with a serialized module loader
+// RUN: %target-swift-frontend -module-name Foo %t/FileFoo.swift -module-alias "//my/project:cat=Bar" -I %t -emit-module -emit-module-path %t/Foo.swiftmodule -Rmodule-loading 2> %t/load-result-foo.output
+
+/// Check Foo.swiftmodule is created and Bar.swiftmodule is loaded
+// RUN: test -f %t/Foo.swiftmodule
+// RUN: test -f %t/Bar.swiftmodule
+// RUN: not test -f %t/*cat.swiftmodule
+
+// RUN: %FileCheck %s -input-file %t/load-result-foo.output -check-prefix CHECK-FOO
+// CHECK-FOO: remark: loaded module {{.*}}Bar.swiftmodule
+
+/// Create a module Zoo that imports `//my/project:cat` with -module-alias "//my/project:cat=Bar" with a source loader
+// RUN: %target-swift-frontend -module-name Zoo %t/FileFoo.swift -module-alias "//my/project:cat=Bar" -I %t -emit-module -emit-module-path %t/Zoo.swiftmodule -enable-source-import -Rmodule-loading 2> %t/load-result-zoo.output
+
+// RUN: test -f %t/Zoo.swiftmodule
+// RUN: test -f %t/Bar.swiftmodule
+// RUN: not test -f %t/*cat.swiftmodule
+
+// RUN: %FileCheck %s -input-file %t/load-result-zoo.output -check-prefix CHECK-ZOO
+// CHECK-ZOO: remark: loaded module {{.*}}Bar.swiftmodule
+
+
+// BEGIN FileBar.swift
+public func bar() {}
+
+// BEGIN FileFoo.swift
+import `//my/project:cat`
+
+`//my/project:cat`.bar()
diff --git a/test/Frontend/module-alias-explicit-build-escaped-identifier.swift b/test/Frontend/module-alias-explicit-build-escaped-identifier.swift
@@ -0,0 +1,65 @@
+/// Test the -module-alias flag with an escaped identifier and the explicit module loader.
+// UNSUPPORTED: OS=windows-msvc
+// RUN: %empty-directory(%t)
+// RUN: mkdir -p %t/inputs
+// RUN: mkdir -p %t/outputs
+
+/// Create a module Bar
+// RUN: echo 'public func bar() {}' > %t/inputs/FileBar.swift
+// RUN: %target-swift-frontend -module-name Bar %t/inputs/FileBar.swift -emit-module -emit-module-path %t/inputs/Bar.swiftmodule
+// RUN: %target-swift-emit-pcm -module-name SwiftShims %swift-lib-dir/swift/shims/module.modulemap -o %t/inputs/SwiftShims.pcm
+// RUN: %target-swift-emit-pcm -module-name _SwiftConcurrencyShims %swift-lib-dir/swift/shims/module.modulemap -o %t/inputs/_SwiftConcurrencyShims.pcm
+
+/// Check Bar.swiftmodule is created
+// RUN: test -f %t/inputs/Bar.swiftmodule
+
+/// Next create an explicit module dependency map to build module Foo
+// RUN: echo 'import `//my/project:cat`' > %t/inputs/FileFoo.swift
+
+// RUN: echo "[{" > %/t/inputs/map.json
+// RUN: echo "\"moduleName\": \"Bar\"," >> %/t/inputs/map.json
+// RUN: echo "\"modulePath\": \"%/t/inputs/Bar.swiftmodule\"," >> %/t/inputs/map.json
+// RUN: echo "\"isFramework\": false" >> %/t/inputs/map.json
+// RUN: echo "}," >> %/t/inputs/map.json
+// RUN: echo "{" >> %/t/inputs/map.json
+// RUN: echo "\"moduleName\": \"Swift\"," >> %/t/inputs/map.json
+// RUN: echo "\"modulePath\": \"%/stdlib_module\"," >> %/t/inputs/map.json
+// RUN: echo "\"isFramework\": false" >> %/t/inputs/map.json
+// RUN: echo "}," >> %/t/inputs/map.json
+// RUN: echo "{" >> %/t/inputs/map.json
+// RUN: echo "\"moduleName\": \"SwiftOnoneSupport\"," >> %/t/inputs/map.json
+// RUN: echo "\"modulePath\": \"%/ononesupport_module\"," >> %/t/inputs/map.json
+// RUN: echo "\"isFramework\": false" >> %/t/inputs/map.json
+// RUN: echo "}," >> %/t/inputs/map.json
+// RUN: echo "{" >> %/t/inputs/map.json
+// RUN: echo "\"moduleName\": \"_Concurrency\"," >> %/t/inputs/map.json
+// RUN: echo "\"modulePath\": \"%/concurrency_module\"," >> %/t/inputs/map.json
+// RUN: echo "\"isFramework\": false" >> %/t/inputs/map.json
+// RUN: echo "}," >> %/t/inputs/map.json
+// RUN: echo "{" >> %/t/inputs/map.json
+// RUN: echo "\"moduleName\": \"SwiftShims\"," >> %/t/inputs/map.json
+// RUN: echo "\"isFramework\": false," >> %/t/inputs/map.json
+// RUN: echo "\"clangModuleMapPath\": \"%swift-lib-dir/swift/shims/module.modulemap\"," >> %/t/inputs/map.json
+// RUN: echo "\"clangModulePath\": \"%t/inputs/SwiftShims.pcm\"" >> %/t/inputs/map.json
+// RUN: echo "}," >> %/t/inputs/map.json
+// RUN: echo "{" >> %/t/inputs/map.json
+// RUN: echo "\"moduleName\": \"_SwiftConcurrencyShims\"," >> %/t/inputs/map.json
+// RUN: echo "\"isFramework\": false," >> %/t/inputs/map.json
+// RUN: echo "\"clangModuleMapPath\": \"%swift-lib-dir/swift/shims/module.modulemap\"," >> %/t/inputs/map.json
+// RUN: echo "\"clangModulePath\": \"%t/inputs/_SwiftConcurrencyShims.pcm\"" >> %/t/inputs/map.json
+// RUN: echo "}," >> %/t/inputs/map.json
+// RUN: echo "{" >> %/t/inputs/map.json
+// RUN: echo "\"moduleName\": \"_StringProcessing\"," >> %/t/inputs/map.json
+// RUN: echo "\"modulePath\": \"%/string_processing_module\"," >> %/t/inputs/map.json
+// RUN: echo "\"isFramework\": false" >> %/t/inputs/map.json
+// RUN: echo "}]" >> %/t/inputs/map.json
+
+/// Create a module Foo that imports `//my/project:cat` with -module-alias "//my/project:cat=Bar" with an explicit module loader
+// RUN: %target-swift-frontend -module-name Foo %t/inputs/FileFoo.swift -module-alias "//my/project:cat=Bar" -I %t/inputs -emit-module -emit-module-path %t/outputs/Foo.swiftmodule -disable-implicit-swift-modules -explicit-swift-module-map-file %t/inputs/map.json -Rmodule-loading 2> %t/outputs/load-result.output
+
+// RUN: test -f %t/outputs/Foo.swiftmodule
+// RUN: test -f %t/inputs/Bar.swiftmodule
+// RUN: not test -f %t/inputs/*cat.swiftmodule
+
+// RUN: %FileCheck %s -input-file %t/outputs/load-result.output -check-prefix CHECK
+// CHECK: remark: loaded module {{.*}}Bar.swiftmodule
