[clang-format] Add more support for C# 8 nullables

[exv]

Summary

This adds support for the null-coalescing assignment and null-forgiving
operators.

https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/operators/null-coalescing-operator

https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/operators/null-forgiving

[MyDeveloperDay]

Thank you for the patch

clang/lib/Format/TokenAnnotator.cpp

3170

should this honour SpaceBeforeAssignmentOperators ?

[exv]

Remove special whitespace handling for ??=

clang/lib/Format/TokenAnnotator.cpp

3170

Good point. Looking more closely, since I have called setKind(tok::equal) on the operator, it seems that the rest of clang-format reliably understands it as an assignment operator and formats it correctly if I don't intervene. I have updated the test case to test this behavior.

[exv]

Fix incorrect arc usage

[exv]

Fixing arc, hopefully?

[exv]

Hey all, I'm really sorry for the noise, I screwed up my arc command to revise the submission.

[HazardyKnusperkeks]

clang/unittests/Format/FormatTestCSharp.cpp

365

Why do you escape ?? I've never seen that.

[exv]

clang/unittests/Format/FormatTestCSharp.cpp

365

Apparently, ??= is a trigraph for "#" in C, so it must be escaped:

https://en.wikipedia.org/wiki/Digraphs_and_trigraphs#C

To be honest, I'd never heard of this bizarre language "feature" until I wrote this patch.

[curdeius]

Looks very good already. Just some questions and test requests.

Requesting changes so it shows up nicely in the revision queue.

clang/lib/Format/TokenAnnotator.cpp

1578

Does adding false and true here may negatively affect JS in any way? (Just thinking out loud)

3168

Should it be renamed to include C#?

clang/unittests/Format/FormatTestCSharp.cpp

365

Not only in C but in C++ too. Fortunately removed in C++17.
Nobody misses trigraphs...

380

Please test also true! or !true!.

381

Could you test that ! is correctly kept together with the corresponding identifier around the line breaks, please?
You might adjust ColumnLimit for that.

[exv]

• [clang-format] Add more support for C# 8 nullables
• Fix null-coalescing assignment whitespace behavior
• Incorporate feedback from curdeius
• Combine JS and CS types/parsing

clang/lib/Format/TokenAnnotator.cpp

1578

Nothing I can see: https://www.typescriptlang.org/docs/handbook/release-notes/typescript-2-0.html#non-null-assertion-operator

3168

Now that you've mentioned it, I do see that JS also defines an almost duplicate set of null-related operators, including null-propagating, null-coalescing and null-coalescing assignment operators, and clang-format handles them with a different code path than C# 😅. They do have slightly different names, but the way they should be parsed and interpreted is the same, and I think it makes sense to use the same logic and token definitions for both of these languages. This resulted in a lot of special-handling for C# tokens that wasn't necessary being removed too.

However, renaming them would introduce a lot of noise, so I think I would prefer to do that in a separate commit.

[curdeius]

LGTM. That's great. Thanks for working on this and for the cleanup.

clang/lib/Format/FormatTokenLexer.cpp

102

Could you explain in what case the operator precedence may be wrong here?

clang/lib/Format/TokenAnnotator.cpp

3168

Fair enough.

[exv]

• Remove obsolete comment

[exv]

According to the wiki, it seems like someone who has commit access must now submit this change.

clang/lib/Format/FormatTokenLexer.cpp

102

It's unlikely this comment is still relevant for this section. I will delete it.

[curdeius]

In D101702#2736886, @exv wrote:

According to the wiki, it seems like someone who has commit access must now submit this change.

Yes, I'll land it. Please provide "Name <mail@address>" for commit attribution if it differs from the info in your Phabricator account.

[exv]

Eliza Velasquez <[e-mail]>

[exv]

I have completed two other formatting fixes related to C# 8 nullables since my last commit. Should I create a new review request for those or update this one?

[curdeius]

Preferably a new one unless it's something really minor.

[curdeius]

Could you please rebase to re-run the CI (that failed for sum unrelated reason)?

[krasimir]

clang/lib/Format/TokenAnnotator.cpp

1570

nit: s/isIdentifier/IsIdentifier/ (LLVM style uses CamelCase for local variables)

clang/unittests/Format/FormatTestCSharp.cpp

365

nit: you could use raw string literals for the ??= test cases (e.g., R"(test ??= ABC;)").

[krasimir]

clang/lib/Format/FormatTokenLexer.cpp

135

nit: use spaces for indentation, no tabs (Phabricator makes it look this line starts with a tab character)

[jbcoe]

Thanks for this patch!

[HazardyKnusperkeks]

clang/lib/Format/FormatTokenLexer.cpp

135

That's not a tab, I fell for that one too once. It just visualizes that there was only whitespace added.

[exv]

• Incorporate krasimir's feedback

[exv]

Rebased and comments addressed!

[exv]

Rebase again in an attempt to fix CI

clang/lib/Format/FormatToken.h

#define LIST_TOKEN_TYPES                                                       \
   TYPE(TrailingReturnArrow)                                                    \
   TYPE(TrailingUnaryOperator)                                                  \
   TYPE(TypeDeclarationParen)                                                   \
   TYPE(TypenameMacro)                                                          \
   TYPE(UnaryOperator)                                                          \
   TYPE(UntouchableMacroFunc)                                                   \
   TYPE(CSharpStringLiteral)                                                    \
   TYPE(CSharpNamedArgumentColon)                                               \
-  TYPE(CSharpNullable)                                                         \
-  TYPE(CSharpNullCoalescing)                                                   \
-  TYPE(CSharpNullConditional)                                                  \
   TYPE(CSharpNullConditionalLSquare)                                           \
   TYPE(CSharpGenericTypeConstraint)                                            \
   TYPE(CSharpGenericTypeConstraintColon)                                       \
   TYPE(CSharpGenericTypeConstraintComma)                                       \
   TYPE(Unknown)
 
 /// Determines the semantic type of a syntactic token, e.g. whether "<" is a
 /// template opener or binary operator.

clang/lib/Format/FormatTokenLexer.h

 private:
   void tryMergePreviousTokens();
 
   bool tryMergeLessLess();
   bool tryMergeNSStringLiteral();
   bool tryMergeJSPrivateIdentifier();
   bool tryMergeCSharpStringLiteral();
   bool tryMergeCSharpKeywordVariables();
-  bool tryMergeCSharpDoubleQuestion();
-  bool tryMergeCSharpNullConditional();
+  bool tryMergeNullishCoalescingEqual();
   bool tryTransformCSharpForEach();
   bool tryMergeForEach();
   bool tryTransformTryUsageForC();
 
   bool tryMergeTokens(ArrayRef<tok::TokenKind> Kinds, TokenType NewType);
 
   // Returns \c true if \p Tok can only be followed by an operand in JavaScript.
   bool precedesOperand(FormatToken *Tok);

clang/lib/Format/FormatTokenLexer.cpp

if (tryMergeConflictMarkers())
     return;
   if (tryMergeLessLess())
     return;
   if (tryMergeForEach())
     return;
   if (Style.isCpp() && tryTransformTryUsageForC())
     return;
 
+  if (Style.Language == FormatStyle::LK_JavaScript || Style.isCSharp()) {
+    static const tok::TokenKind NullishCoalescingOperator[] = {tok::question,
+                                                               tok::question};
+    static const tok::TokenKind NullPropagatingOperator[] = {tok::question,
+                                                             tok::period};
+    static const tok::TokenKind FatArrow[] = {tok::equal, tok::greater};
+
+    if (tryMergeTokens(FatArrow, TT_JsFatArrow))
+      return;
+    if (tryMergeTokens(NullishCoalescingOperator,
+                       TT_JsNullishCoalescingOperator)) {
+      // Treat like the "||" operator (as opposed to the ternary ?).
+      Tokens.back()->Tok.setKind(tok::pipepipe);
+      return;
+    }
+    if (tryMergeTokens(NullPropagatingOperator, TT_JsNullPropagatingOperator)) {
+      // Treat like a regular "." access.
+      Tokens.back()->Tok.setKind(tok::period);
+      return;
+    }
+    if (tryMergeNullishCoalescingEqual()) {
+      return;
+    }
+  }
+
   if (Style.isCSharp()) {
+    static const tok::TokenKind CSharpNullConditionalLSquare[] = {
+        tok::question, tok::l_square};
+
     if (tryMergeCSharpKeywordVariables())
       return;
     if (tryMergeCSharpStringLiteral())
       return;
-    if (tryMergeCSharpDoubleQuestion())
-      return;
-    if (tryMergeCSharpNullConditional())
-      return;
     if (tryTransformCSharpForEach())
       return;
-    static const tok::TokenKind JSRightArrow[] = {tok::equal, tok::greater};
-    if (tryMergeTokens(JSRightArrow, TT_JsFatArrow))
+    if (tryMergeTokens(CSharpNullConditionalLSquare,
+                       TT_CSharpNullConditionalLSquare)) {
+      // Treat like a regular "[" operator.
+      Tokens.back()->Tok.setKind(tok::l_square);
       return;
     }
+  }
 
   if (tryMergeNSStringLiteral())
     return;
 
   if (Style.Language == FormatStyle::LK_JavaScript) {
     static const tok::TokenKind JSIdentity[] = {tok::equalequal, tok::equal};
     static const tok::TokenKind JSNotIdentity[] = {tok::exclaimequal,
                                                    tok::equal};
     static const tok::TokenKind JSShiftEqual[] = {tok::greater, tok::greater,
                                                   tok::greaterequal};
-    static const tok::TokenKind JSRightArrow[] = {tok::equal, tok::greater};
     static const tok::TokenKind JSExponentiation[] = {tok::star, tok::star};
     static const tok::TokenKind JSExponentiationEqual[] = {tok::star,
                                                            tok::starequal};
-    static const tok::TokenKind JSNullPropagatingOperator[] = {tok::question,
-                                                               tok::period};
-    static const tok::TokenKind JSNullishOperator[] = {tok::question,
-                                                       tok::question};
-    static const tok::TokenKind JSNullishEqual[] = {tok::question,
-                                                    tok::question, tok::equal};
     static const tok::TokenKind JSPipePipeEqual[] = {tok::pipepipe, tok::equal};
     static const tok::TokenKind JSAndAndEqual[] = {tok::ampamp, tok::equal};
 
     // FIXME: Investigate what token type gives the correct operator priority.
     if (tryMergeTokens(JSIdentity, TT_BinaryOperator))
       return;
     if (tryMergeTokens(JSNotIdentity, TT_BinaryOperator))
       return;
     if (tryMergeTokens(JSShiftEqual, TT_BinaryOperator))
       return;
-    if (tryMergeTokens(JSRightArrow, TT_JsFatArrow))
-      return;
     if (tryMergeTokens(JSExponentiation, TT_JsExponentiation))
       return;
     if (tryMergeTokens(JSExponentiationEqual, TT_JsExponentiationEqual)) {
       Tokens.back()->Tok.setKind(tok::starequal);
       return;
     }
-    if (tryMergeTokens(JSNullishOperator, TT_JsNullishCoalescingOperator)) {
-      // Treat like the "||" operator (as opposed to the ternary ?).
-      Tokens.back()->Tok.setKind(tok::pipepipe);
-      return;
-    }
-    if (tryMergeTokens(JSNullPropagatingOperator,
-                       TT_JsNullPropagatingOperator)) {
-      // Treat like a regular "." access.
-      Tokens.back()->Tok.setKind(tok::period);
-      return;
-    }
     if (tryMergeTokens(JSAndAndEqual, TT_JsAndAndEqual) ||
-        tryMergeTokens(JSPipePipeEqual, TT_JsPipePipeEqual) ||
-        tryMergeTokens(JSNullishEqual, TT_JsNullishCoalescingEqual)) {
+        tryMergeTokens(JSPipePipeEqual, TT_JsPipePipeEqual)) {
       // Treat like the "=" assignment operator.
       Tokens.back()->Tok.setKind(tok::equal);
       return;
     }
     if (tryMergeJSPrivateIdentifier())
       return;
   }
 

 
 // Valid C# attribute targets:
 // https://docs.microsoft.com/en-us/dotnet/csharp/programming-guide/concepts/attributes/#attribute-targets
 const llvm::StringSet<> FormatTokenLexer::CSharpAttributeTargets = {
     "assembly", "module",   "field",  "event", "method",
     "param",    "property", "return", "type",
 };
 
-bool FormatTokenLexer::tryMergeCSharpDoubleQuestion() {
-  if (Tokens.size() < 2)
-    return false;
-  auto &FirstQuestion = *(Tokens.end() - 2);
-  auto &SecondQuestion = *(Tokens.end() - 1);
-  if (!FirstQuestion->is(tok::question) || !SecondQuestion->is(tok::question))
-    return false;
-  FirstQuestion->Tok.setKind(tok::question); // no '??' in clang tokens.
-  FirstQuestion->TokenText = StringRef(FirstQuestion->TokenText.begin(),
-                                       SecondQuestion->TokenText.end() -
-                                           FirstQuestion->TokenText.begin());
-  FirstQuestion->ColumnWidth += SecondQuestion->ColumnWidth;
-  FirstQuestion->setType(TT_CSharpNullCoalescing);
-  Tokens.erase(Tokens.end() - 1);
-  return true;
-}
-
-// Merge '?[' and '?.' pairs into single tokens.
-bool FormatTokenLexer::tryMergeCSharpNullConditional() {
+bool FormatTokenLexer::tryMergeNullishCoalescingEqual() {
   if (Tokens.size() < 2)
     return false;
-  auto &Question = *(Tokens.end() - 2);
-  auto &PeriodOrLSquare = *(Tokens.end() - 1);
-  if (!Question->is(tok::question) ||
-      !PeriodOrLSquare->isOneOf(tok::l_square, tok::period))
-    return false;
-  Question->TokenText =
-      StringRef(Question->TokenText.begin(),
-                PeriodOrLSquare->TokenText.end() - Question->TokenText.begin());
-  Question->ColumnWidth += PeriodOrLSquare->ColumnWidth;
-
-  if (PeriodOrLSquare->is(tok::l_square)) {
-    Question->Tok.setKind(tok::question); // no '?[' in clang tokens.
-    Question->setType(TT_CSharpNullConditionalLSquare);
-  } else {
-    Question->Tok.setKind(tok::question); // no '?.' in clang tokens.
-    Question->setType(TT_CSharpNullConditional);
-  }
-
+  auto &NullishCoalescing = *(Tokens.end() - 2);
+  auto &Equal = *(Tokens.end() - 1);
+  if (NullishCoalescing->getType() != TT_JsNullishCoalescingOperator ||
+      !Equal->is(tok::equal))
+    return false;
+  NullishCoalescing->Tok.setKind(tok::equal); // no '??=' in clang tokens.
+  NullishCoalescing->TokenText =
+      StringRef(NullishCoalescing->TokenText.begin(),
+                Equal->TokenText.end() - NullishCoalescing->TokenText.begin());
+  NullishCoalescing->ColumnWidth += Equal->ColumnWidth;
+  NullishCoalescing->setType(TT_JsNullishCoalescingEqual);
   Tokens.erase(Tokens.end() - 1);
   return true;
 }
 
 bool FormatTokenLexer::tryMergeCSharpKeywordVariables() {
   if (Tokens.size() < 2)
     return false;
   auto &At = *(Tokens.end() - 2);

clang/lib/Format/TokenAnnotator.cpp

case tok::kw_operator:
           CurrentToken->Previous->setType(TT_OverloadedOperator);
       }
       if (CurrentToken && CurrentToken->is(tok::l_paren))
         CurrentToken->setType(TT_OverloadedOperatorLParen);
       if (CurrentToken && CurrentToken->Previous->is(TT_BinaryOperator))
         CurrentToken->Previous->setType(TT_OverloadedOperator);
       break;
     case tok::question:
-      if (Tok->is(TT_CSharpNullConditionalLSquare)) {
-        if (!parseSquare())
-          return false;
-        break;
-      }
-      if (Tok->isOneOf(TT_CSharpNullConditional, TT_CSharpNullCoalescing))
-        break;
       if (Style.Language == FormatStyle::LK_JavaScript && Tok->Next &&
           Tok->Next->isOneOf(tok::semi, tok::comma, tok::colon, tok::r_paren,
                              tok::r_brace)) {
         // Question marks before semicolons, colons, etc. indicate optional
         // types (fields, parameters), e.g.
         //   function(x?: string, y?) {...}
         //   class X { y?; }
         Tok->setType(TT_JsTypeOptionalQuestion);
         break;
       }
       // Declarations cannot be conditional expressions, this can only be part
       // of a type declaration.
       if (Line.MustBeDeclaration && !Contexts.back().IsExpression &&
           Style.Language == FormatStyle::LK_JavaScript)
         break;
       if (Style.isCSharp()) {
         // `Type?)`, `Type?>`, `Type? name;` and `Type? name =` can only be
         // nullable types.
         // Line.MustBeDeclaration will be true for `Type? name;`.
         if ((!Contexts.back().IsExpression && Line.MustBeDeclaration) ||
             (Tok->Next && Tok->Next->isOneOf(tok::r_paren, tok::greater)) ||
             (Tok->Next && Tok->Next->is(tok::identifier) && Tok->Next->Next &&
              Tok->Next->Next->is(tok::equal))) {
-          Tok->setType(TT_CSharpNullable);
+          Tok->setType(TT_JsTypeOptionalQuestion);
           break;
         }
       }
       parseConditional();
       break;
     case tok::kw_template:
       parseTemplateDeclaration();
       break;
static bool isDeductionGuide(FormatToken &Current) {
     return false;
   }
 
   void determineTokenType(FormatToken &Current) {
     if (!Current.is(TT_Unknown))
       // The token type is already known.
       return;
 
-    if (Style.isCSharp() && CurrentToken->is(tok::question)) {
-      if (CurrentToken->TokenText == "??") {
-        Current.setType(TT_CSharpNullCoalescing);
-        return;
-      }
-      if (CurrentToken->TokenText == "?.") {
-        Current.setType(TT_CSharpNullConditional);
-        return;
-      }
-      if (CurrentToken->TokenText == "?[") {
-        Current.setType(TT_CSharpNullConditionalLSquare);
-        return;
-      }
-    }
-
-    if (Style.Language == FormatStyle::LK_JavaScript) {
-      if (Current.is(tok::exclaim)) {
-        if (Current.Previous &&
-            (Keywords.IsJavaScriptIdentifier(
-                 *Current.Previous, /* AcceptIdentifierName= */ true) ||
-             Current.Previous->isOneOf(
-                 tok::kw_namespace, tok::r_paren, tok::r_square, tok::r_brace,
-                 Keywords.kw_type, Keywords.kw_get, Keywords.kw_set) ||
-             Current.Previous->Tok.isLiteral())) {
+    if ((Style.Language == FormatStyle::LK_JavaScript || Style.isCSharp()) &&
+        Current.is(tok::exclaim)) {
+      if (Current.Previous) {
+        bool IsIdentifier =
+            Style.Language == FormatStyle::LK_JavaScript
+                ? Keywords.IsJavaScriptIdentifier(
+                      *Current.Previous, /* AcceptIdentifierName= */ true)
+                : Current.Previous->is(tok::identifier);
+        if (IsIdentifier ||
+                tok::kw_false, tok::kw_true, Keywords.kw_type, Keywords.kw_get,
+                Keywords.kw_set) ||
+            Current.Previous->Tok.isLiteral()) {
           Current.setType(TT_JsNonNullAssertion);
           return;
         }
-        if (Current.Next &&
-            Current.Next->isOneOf(TT_BinaryOperator, Keywords.kw_as)) {
-          Current.setType(TT_JsNonNullAssertion);
-          return;
-        }
-      }
-    }
+      }
 
     // Line.MightBeFunctionDecl can only be true after the parentheses of a
     // function declaration have been found. In this case, 'Current' is a
     // trailing token of this declaration and thus cannot be a name.
     if (Current.is(Keywords.kw_instanceof)) {
       Current.setType(TT_BinaryOperator);
     } else if (isStartOfName(Current) &&
                (!Line.MightBeFunctionDecl || Current.NestingLevel != 0)) {
if (Style.isCpp()) {
     if (Left.is(TT_TemplateCloser) && Right.is(TT_StartOfName))
       return true;
 
     // spaces inside square brackets.
     if (Left.is(tok::l_square) || Right.is(tok::r_square))
       return Style.SpacesInSquareBrackets;
 
     // No space before ? in nullable types.
-    if (Right.is(TT_CSharpNullable))
-      return false;
-
-    // Require space after ? in nullable types except in generics and casts.
-    if (Left.is(TT_CSharpNullable))
-      return !Right.isOneOf(TT_TemplateCloser, tok::r_paren);
-
-    // No space before or after '?.'.
-    if (Left.is(TT_CSharpNullConditional) || Right.is(TT_CSharpNullConditional))
+    if (Right.is(TT_JsTypeOptionalQuestion))
       return false;
 
-    // Space before and after '??'.
-    if (Left.is(TT_CSharpNullCoalescing) || Right.is(TT_CSharpNullCoalescing))
-      return true;
-
-    // No space before '?['.
-    if (Right.is(TT_CSharpNullConditionalLSquare))
+    // No space before null forgiving '!'.
+    if (Right.is(TT_JsNonNullAssertion))
       return false;
 
     // No space between consecutive commas '[,,]'.
     if (Left.is(tok::comma) && Right.is(tok::comma))
       return false;
 
-    // Possible space inside `?[ 0 ]`.
-    if (Left.is(TT_CSharpNullConditionalLSquare))
-      return Style.SpacesInSquareBrackets;
-
     // space after var in `var (key, value)`
     if (Left.is(Keywords.kw_var) && Right.is(tok::l_paren))
       return true;
 
     // space between keywords and paren e.g. "using ("
     if (Right.is(tok::l_paren))
       if (Left.isOneOf(tok::kw_using, Keywords.kw_async, Keywords.kw_when,
                        Keywords.kw_lock))

clang/unittests/Format/FormatTestCSharp.cpp

 }
 
 TEST_F(FormatTestCSharp, CSharpNullCoalescing) {
   verifyFormat("var test = ABC ?? DEF");
   verifyFormat("string myname = name ?? \"ABC\";");
   verifyFormat("return _name ?? \"DEF\";");
 }
 
+TEST_F(FormatTestCSharp, CSharpNullCoalescingAssignment) {
+  FormatStyle Style = getGoogleStyle(FormatStyle::LK_CSharp);
+  Style.SpaceBeforeAssignmentOperators = true;
+
+  verifyFormat(R"(test ??= ABC;)", Style);
+  verifyFormat(R"(test ??= true;)", Style);
+
+  Style.SpaceBeforeAssignmentOperators = false;
+
+  verifyFormat(R"(test??= ABC;)", Style);
+  verifyFormat(R"(test??= true;)", Style);
+}
+
+TEST_F(FormatTestCSharp, CSharpNullForgiving) {
+  FormatStyle Style = getGoogleStyle(FormatStyle::LK_CSharp);
+
+  verifyFormat("var test = null!;", Style);
+  verifyFormat("string test = someFunctionCall()! + \"ABC\"!", Style);
+  verifyFormat("int test = (1! + 2 + bar! + foo())!", Style);
+  verifyFormat(R"(test ??= !foo!;)", Style);
+  verifyFormat("test = !bar! ?? !foo!;", Style);
+  verifyFormat("bool test = !(!true && !true! || !null && !null! || !false && "
+               "!false! && !bar()! + (!foo()))!",
+               Style);
+
+  // Check that line break keeps identifier with the bang.
+  Style.ColumnLimit = 14;
+
+  verifyFormat("var test =\n"
+               "    foo!;",
+               Style);
+}
+
 TEST_F(FormatTestCSharp, AttributesIndentation) {
   FormatStyle Style = getMicrosoftStyle(FormatStyle::LK_CSharp);
   Style.AlwaysBreakAfterReturnType = FormatStyle::RTBS_None;
 
   verifyFormat("[STAThread]\n"
                "static void Main(string[] args)\n"
                "{\n"
                "}",