[clang-format] Fix indent of trailing raw string param after newline

[krasimir]

Summary

Currently clang-format uses ContinuationIndent to indent the contents of a raw
string literal that is the last parameter of the function call. This is to
achieve formatting similar to trailing:

f(1, 2, R"pb(
    x: y)pb");

However this had the unfortunate consequence of producing format like this:

fffffff(1, 2,
        R"pb(
    a: b
        )pb");

This patch makes clang-format consider indenting a trailing raw string param
after a newline based off the start of the format delimiter, producing:

fffffff(1, 2,
        R"pb(
          a: b
        )pb");

[djasper]

Looks good. Thank you.

lib/Format/ContinuationIndenter.h

private:
   void moveStateToNewBlock(LineState &State);
 
   /// Reformats a raw string literal.
   ///
   /// \returns An extra penalty induced by reformatting the token.
   unsigned reformatRawStringLiteral(const FormatToken &Current,
                                     LineState &State,
                                     const FormatStyle &RawStringStyle,
-                                    bool DryRun);
+                                    bool DryRun, bool Newline);
 
   /// If the current token is at the end of the current line, handle
   /// the transition to the next line.
   unsigned handleEndOfLine(const FormatToken &Current, LineState &State,
-                           bool DryRun, bool AllowBreak);
+                           bool DryRun, bool AllowBreak, bool Newline);
 
   /// If \p Current is a raw string that is configured to be reformatted,
   /// return the style to be used.
   llvm::Optional<FormatStyle> getRawStringStyle(const FormatToken &Current,
                                                 const LineState &State);
 
   /// If the current token sticks out over the end of the line, break
   /// it if possible.

lib/Format/ContinuationIndenter.cpp

else if (Current.isStringLiteral() && State.StartOfStringLiteral == 0)
     State.StartOfStringLiteral = State.Column;
   else if (!Current.isOneOf(tok::comment, tok::identifier, tok::hash) &&
            !Current.isStringLiteral())
     State.StartOfStringLiteral = 0;
 
   State.Column += Current.ColumnWidth;
   State.NextToken = State.NextToken->Next;
 
-  unsigned Penalty = handleEndOfLine(Current, State, DryRun, AllowBreak);
+  unsigned Penalty =
+      handleEndOfLine(Current, State, DryRun, AllowBreak, Newline);
 
   if (Current.Role)
     Current.Role->formatFromToken(State, this, DryRun);
   // If the previous has a special role, let it consume tokens as appropriate.
   // It is necessary to start at the previous token for the only implemented
   // role (comma separated list). That way, the decision whether or not to break
   // after the "{" is already done and both options are tried and evaluated.
   // FIXME: This is ugly, find a better way.
static unsigned getLastLineEndColumn(StringRef Text, unsigned StartColumn,
   } else {
     return encoding::columnWidthWithTabs(Text.substr(LastNewlinePos),
                                          /*StartColumn=*/0, TabWidth, Encoding);
   }
 }
 
 unsigned ContinuationIndenter::reformatRawStringLiteral(
     const FormatToken &Current, LineState &State,
-    const FormatStyle &RawStringStyle, bool DryRun) {
+    const FormatStyle &RawStringStyle, bool DryRun, bool Newline) {
   unsigned StartColumn = State.Column - Current.ColumnWidth;
   StringRef OldDelimiter = *getRawStringDelimiter(Current.TokenText);
   StringRef NewDelimiter =
       getCanonicalRawStringDelimiter(Style, RawStringStyle.Language);
   if (NewDelimiter.empty() || OldDelimiter.empty())
     NewDelimiter = OldDelimiter;
   // The text of a raw string is between the leading 'R"delimiter(' and the
   // trailing 'delimiter)"'.
unsigned ContinuationIndenter::reformatRawStringLiteral(
   //     indent, and
   //   - if the content does not start on a newline, it is the first start
   //     column.
   // These rules have the advantage that the formatted content both does not
   // violate the rectangle rule and visually flows within the surrounding
   // source.
   bool ContentStartsOnNewline = Current.TokenText[OldPrefixSize] == '\n';
   // If this token is the last parameter (checked by looking if it's followed by
-  // `)`, the base the indent off the line's nested block indent. Otherwise,
-  // base the indent off the arguments indent, so we can achieve:
+  // `)` and is not on a newline, the base the indent off the line's nested
+  // block indent. Otherwise, base the indent off the arguments indent, so we
+  // can achieve:
+  //
   // fffffffffff(1, 2, 3, R"pb(
   //     key1: 1  #
   //     key2: 2)pb");
   //
   // fffffffffff(1, 2, 3,
   //             R"pb(
   //               key1: 1  #
   //               key2: 2
+  //             )pb");
+  //
+  // fffffffffff(1, 2, 3,
+  //             R"pb(
+  //               key1: 1  #
+  //               key2: 2
   //             )pb",
   //             5);
-  unsigned CurrentIndent = (Current.Next && Current.Next->is(tok::r_paren))
-                               ? State.Stack.back().NestedBlockIndent
-                               : State.Stack.back().Indent;
+  unsigned CurrentIndent =
+      (!Newline && Current.Next && Current.Next->is(tok::r_paren))
   unsigned NextStartColumn = ContentStartsOnNewline
                                  ? CurrentIndent + Style.IndentWidth
                                  : FirstStartColumn;
 
   // The last start column is the column the raw string suffix starts if it is
   // put on a newline.
   // The last start column is the intended indentation of the raw string postfix
   // if it is put on a newline. It is determined by the following rules:
unsigned ContinuationIndenter::addMultilineToken(const FormatToken &Current,
 
   if (ColumnsUsed > getColumnLimit(State))
     return Style.PenaltyExcessCharacter * (ColumnsUsed - getColumnLimit(State));
   return 0;
 }
 
 unsigned ContinuationIndenter::handleEndOfLine(const FormatToken &Current,
                                                LineState &State, bool DryRun,
-                                               bool AllowBreak) {
+                                               bool AllowBreak, bool Newline) {
   unsigned Penalty = 0;
   // Compute the raw string style to use in case this is a raw string literal
   // that can be reformatted.
   auto RawStringStyle = getRawStringStyle(Current, State);
   if (RawStringStyle && !Current.Finalized) {
-    Penalty = reformatRawStringLiteral(Current, State, *RawStringStyle, DryRun);
+    Penalty = reformatRawStringLiteral(Current, State, *RawStringStyle, DryRun,
+                                       Newline);
   } else if (Current.IsMultiline && Current.isNot(TT_BlockComment)) {
     // Don't break multi-line tokens other than block comments and raw string
     // literals. Instead, just update the state.
     Penalty = addMultilineToken(Current, State);
   } else if (State.Line->Type != LT_ImportStatement) {
     // We generally don't break import statements.
     LineState OriginalState = State;
 

unittests/Format/FormatTestRawStrings.cpp

 })test",
             format(R"test(
 int f() {
   int a = g(R"pb(key:
                  1)pb", 2);
 })test", Style));
 }
 
+TEST_F(FormatTestRawStrings, IndentsLastParamAfterNewline) {
+  FormatStyle Style = getRawStringPbStyleWithColumns(60);
+  expect_eq(R"test(
+fffffffffffffffffffff("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",
+                      R"pb(
+                        b: c
+                      )pb");)test",
+            format(R"test(
+fffffffffffffffffffff("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",
+                      R"pb(
+                      b: c
+                      )pb");)test",
+                   Style));
+}
 } // end namespace
 } // end namespace format
 } // end namespace clang