clang-format: Add support for formatting lambdas with explicit template parameters.

[thakis]

Summary

Ports r359967 to clang-format.

[krasimir]

I'll need some more time to investigate the implications of this with respect to Objective-C disambiguation stuff.
Specifically, this may interact with funny ways with the heuristic outlined in (old) UnwrappedLineParser.cpp line 1453:

// In a C++ lambda a template type can only occur after an arrow. We use
// this as an heuristic to distinguish between Objective-C expressions
// followed by an `a->b` expression, such as:
// ([obj func:arg] + a->b)

At least we'll have to update this comment accordingly, otherwise it will be confusing.

clang/lib/Format/TokenAnnotator.cpp

44

nit: add a [ in the second case, like [...]<...>(. Also I'd suggest extending this sentence with: ", where the [ opens a lambda capture list", because we explicitly check the [ for that and to disambiguate between the case this is interested in and Objective-C method calls followed by function-invocation style call, like in [obj meth]().

50

The first Left->Previous check is unnecessary here, following the previous if.

55

nit: consider replacing with

Check for `[...]`

[thakis]

Thanks for the thorough review! Indeed, this still gets []<bool b = true && false>(A &&a){}(); wrong, for the reason you mention.

clang/lib/Format/TokenAnnotator.cpp

50

I removed the previous if, it makes the function more symmetric.

[thakis]

fix nits

[thakis]

Do you think this should land with the comment FIXME for now? It improves formatting of this language feature when that heuristic is not used, and changing the heuristic is independent of the rest of this patch.

[krasimir]

This looks good with the FIXME.

clang/lib/Format/TokenAnnotator.cpp

 /// For Objective-C and Objective-C++, both identifiers and keywords
 /// are valid inside @selector(...) (or a macro which
 /// invokes @selector(...)). So, we allow treat any identifier or
 /// keyword as a potential Objective-C selector component.
 static bool canBeObjCSelectorComponent(const FormatToken &Tok) {
   return Tok.Tok.getIdentifierInfo() != nullptr;
 }
 
+/// With `Left` being '(', check if we're at either `[...](` or
+/// `[...]<...>(`, where the [ opens a lambda capture list.
+static bool isLambdaParameterList(const FormatToken *Left) {
+  // Skip <...> if present.
+  if (Left->Previous && Left->Previous->is(tok::greater) &&
+      Left->Previous->MatchingParen &&
+      Left->Previous->MatchingParen->is(TT_TemplateOpener))
+    Left = Left->Previous->MatchingParen;
+
+  // Check for `[...]`.
+  return Left->Previous && Left->Previous->is(tok::r_square) &&
+         Left->Previous->MatchingParen &&
+         Left->Previous->MatchingParen->is(TT_LambdaLSquare);
+}
+
 /// A parser that gathers additional information about tokens.
 ///
 /// The \c TokenAnnotator tries to match parenthesis and square brakets and
 /// store a parenthesis levels. It also tries to resolve matching "<" and ">"
 /// into template parameter lists.
 class AnnotatingParser {
 public:
   AnnotatingParser(const FormatStyle &Style, AnnotatedLine &Line,
if (Left->is(TT_OverloadedOperatorLParen)) {
                 (Left->Previous->endsSequence(tok::identifier,
                                               Keywords.kw_function)))) {
       // function(...) or function f(...)
       Contexts.back().IsExpression = false;
     } else if (Style.Language == FormatStyle::LK_JavaScript && Left->Previous &&
                Left->Previous->is(TT_JsTypeColon)) {
       // let x: (SomeType);
       Contexts.back().IsExpression = false;
-    } else if (Left->Previous && Left->Previous->is(tok::r_square) &&
-               Left->Previous->MatchingParen &&
-               Left->Previous->MatchingParen->is(TT_LambdaLSquare)) {
+    } else if (isLambdaParameterList(Left)) {
       // This is a parameter list of a lambda expression.
       Contexts.back().IsExpression = false;
     } else if (Line.InPPDirective &&
                (!Left->Previous || !Left->Previous->is(tok::identifier))) {
       Contexts.back().IsExpression = true;
     } else if (Contexts[Contexts.size() - 2].CaretFound) {
       // This is the parameter list of an ObjC block.
       Contexts.back().IsExpression = false;

clang/lib/Format/UnwrappedLineParser.cpp

while (FormatTok->isNot(tok::l_brace)) {
     case tok::star:
     case tok::kw_const:
     case tok::comma:
     case tok::less:
     case tok::greater:
     case tok::identifier:
     case tok::numeric_constant:
     case tok::coloncolon:
+    case tok::kw_class:
     case tok::kw_mutable:
     case tok::kw_noexcept:
+    case tok::kw_template:
+    case tok::kw_typename:
       nextToken();
       break;
     // Specialization of a template with an integer parameter can contain
     // arithmetic, logical, comparison and ternary operators.
     //
     // FIXME: This also accepts sequences of operators that are not in the scope
     // of a template argument list.
     //
     // In a C++ lambda a template type can only occur after an arrow. We use
     // this as an heuristic to distinguish between Objective-C expressions
     // followed by an `a->b` expression, such as:
     // ([obj func:arg] + a->b)
     // Otherwise the code below would parse as a lambda.
+    //
+    // FIXME: This heuristic is incorrect for C++20 generic lambdas with
+    // explicit template lists: []<bool b = true && false>(U &&u){}
     case tok::plus:
     case tok::minus:
     case tok::exclaim:
     case tok::tilde:
     case tok::slash:
     case tok::percent:
     case tok::lessless:
     case tok::pipe:

clang/unittests/Format/FormatTest.cpp

verifyFormat(
       "      return aaaaaaaaaaaaaaaaa;\n"
       "    });",
       getLLVMStyleWithColumns(70));
   verifyFormat("[]() //\n"
                "    -> int {\n"
                "  return 1; //\n"
                "};");
 
+  // Lambdas with explicit template argument lists.
+  verifyFormat(
+      "auto L = []<template <typename> class T, class U>(T<U> &&a) {};\n");
+
   // Multiple lambdas in the same parentheses change indentation rules. These
   // lambdas are forced to start on new lines.
   verifyFormat("SomeFunction(\n"
                "    []() {\n"
                "      //\n"
                "    },\n"
                "    []() {\n"
                "      //\n"
                "    });");
 
   // A lambda passed as arg0 is always pushed to the next line.
   verifyFormat("SomeFunction(\n"
                "    [this] {\n"
                "      //\n"
                "    },\n"
                "    1);\n");
 
-  // A multi-line lambda passed as arg1 forces arg0 to be pushed out, just like the arg0
-  // case above.
+  // A multi-line lambda passed as arg1 forces arg0 to be pushed out, just like
+  // the arg0 case above.
   auto Style = getGoogleStyle();
   Style.BinPackArguments = false;
   verifyFormat("SomeFunction(\n"
                "    a,\n"
                "    [this] {\n"
                "      //\n"
                "    },\n"
                "    b);\n",