[PATCH] Better support for multiline string literals (including C++11 raw string literals).

[Alexander Kornienko]

Hi djasper,

Calculate characters in the first and the last line correctly so that
we only break before the literal when needed.

http://llvm-reviews.chandlerc.com/D1544

Files:
  lib/Format/ContinuationIndenter.cpp
  lib/Format/ContinuationIndenter.h
  unittests/Format/FormatTest.cpp

Index: lib/Format/ContinuationIndenter.cpp
===================================================================
--- lib/Format/ContinuationIndenter.cpp
+++ lib/Format/ContinuationIndenter.cpp
@@ -579,35 +579,60 @@
   return Penalty;
 }
 
+unsigned
+ContinuationIndenter::addMultilineStringLiteral(const FormatToken &Current,
+                                                LineState &State) {
+  StringRef Text = Current.TokenText;
+  size_t FirstLineBreak = Text.find('\n');
+  size_t LastLineBreak = Text.find_last_of('\n');
+  assert(FirstLineBreak != StringRef::npos);
+  unsigned StartColumn = State.Column - Current.CodePointCount;
+  State.Column =
+      encoding::getCodePointCount(Text.substr(LastLineBreak + 1), Encoding);
+
+  // Break before further function parameters on all levels.
+  for (unsigned i = 0, e = State.Stack.size(); i != e; ++i)
+    State.Stack[i].BreakBeforeParameter = true;
+
+  unsigned ColumnsUsed =
+      StartColumn +
+      encoding::getCodePointCount(Text.substr(0, FirstLineBreak), Encoding);
+  if (ColumnsUsed > getColumnLimit())
+    return Style.PenaltyExcessCharacter * (ColumnsUsed - getColumnLimit());
+  return 0;
+}
+
 unsigned ContinuationIndenter::breakProtrudingToken(const FormatToken &Current,
                                                     LineState &State,
                                                     bool DryRun) {
   if (!Current.isOneOf(tok::string_literal, tok::comment))
     return 0;
 
   llvm::OwningPtr<BreakableToken> Token;
   unsigned StartColumn = State.Column - Current.CodePointCount;
-  unsigned OriginalStartColumn =
-      SourceMgr.getSpellingColumnNumber(Current.getStartOfNonWhitespace()) - 1;
 
   if (Current.is(tok::string_literal) &&
       Current.Type != TT_ImplicitStringLiteral) {
+    // Don't break string literals with (in case of non-raw strings, escaped)
+    // newlines. As clang-format must not change the string's content, it is
+    // unlikely that we'll end up with a better format.
+    if (Current.TokenText.find('\n') != StringRef::npos)
+      return addMultilineStringLiteral(Current, State);
+
     // Only break up default narrow strings.
     if (!Current.TokenText.startswith("\""))
       return 0;
-    // Don't break string literals with escaped newlines. As clang-format must
-    // not change the string's content, it is unlikely that we'll end up with
-    // a better format.
-    if (Current.TokenText.find("\\\n") != StringRef::npos)
-      return 0;
     // Exempts unterminated string literals from line breaking. The user will
     // likely want to terminate the string before any line breaking is done.
     if (Current.IsUnterminatedLiteral)
       return 0;
 
     Token.reset(new BreakableStringLiteral(Current, StartColumn,
                                            Line.InPPDirective, Encoding));
   } else if (Current.Type == TT_BlockComment && Current.isTrailingComment()) {
+    unsigned OriginalStartColumn =
+        SourceMgr.getSpellingColumnNumber(Current.getStartOfNonWhitespace()) -
+        1;
     Token.reset(new BreakableBlockComment(
         Style, Current, StartColumn, OriginalStartColumn, !Current.Previous,
         Line.InPPDirective, Encoding));
@@ -707,6 +732,10 @@
   const FormatToken &Current = *State.NextToken;
   if (!Current.is(tok::string_literal))
     return false;
+  // We never consider raw string literals "multiline" for the purpose of
+  // AlwaysBreakBeforeMultilineStrings implementation.
+  if (Current.TokenText.startswith("R\""))
+    return false;
   if (Current.getNextNonComment() &&
       Current.getNextNonComment()->is(tok::string_literal))
     return true; // Implicit concatenation.
Index: lib/Format/ContinuationIndenter.h
===================================================================
--- lib/Format/ContinuationIndenter.h
+++ lib/Format/ContinuationIndenter.h
@@ -84,6 +84,9 @@
   unsigned breakProtrudingToken(const FormatToken &Current, LineState &State,
                                 bool DryRun);
 
+  unsigned addMultilineStringLiteral(const FormatToken &Current,
+                                     LineState &State);
+
   /// \brief Returns \c true if the next token starts a multiline string
   /// literal.
   ///
Index: unittests/Format/FormatTest.cpp
===================================================================
--- unittests/Format/FormatTest.cpp
+++ unittests/Format/FormatTest.cpp
@@ -5340,13 +5340,53 @@
 }
 
 TEST_F(FormatTest, DontSplitStringLiteralsWithEscapedNewlines) {
-  EXPECT_EQ("aaaaaaaaaaa =\n"
-            "    \"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\\\n"
-            "  aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\\\n"
-            "  aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\";",
-            format("aaaaaaaaaaa = \"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\\\n"
-                   "  aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\\\n"
-                   "  aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\";"));
+  EXPECT_EQ(
+      "aaaaaaaaaaa = \"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\\\n"
+      "  aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\\\n"
+      "  aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\";",
+      format("aaaaaaaaaaa  =  \"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\\\n"
+             "  aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\\\n"
+             "  aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\";"));
+}
+
+TEST_F(FormatTest, CountsCharactersInMultilineRawStringLiterals) {
+  EXPECT_EQ("f(g(R\"x(raw literal)x\", a), b);",
+            format("f(g(R\"x(raw literal)x\", a), b);", getGoogleStyle()));
+  EXPECT_EQ("fffffffffff(g(R\"x(\n"
+            "multiline raw string literal xxxxxxxxxxxxxx\n"
+            ")x\",\n"
+            "              a),\n"
+            "            b);",
+            format("fffffffffff(g(R\"x(\n"
+                   "multiline raw string literal xxxxxxxxxxxxxx\n"
+                   ")x\", a), b);",
+                   getGoogleStyleWithColumns(20)));
+  EXPECT_EQ("fffffffffff(\n"
+            "    g(R\"x(qqq\n"
+            "multiline raw string literal xxxxxxxxxxxxxx\n"
+            ")x\",\n"
+            "      a),\n"
+            "    b);",
+            format("fffffffffff(g(R\"x(qqq\n"
+                   "multiline raw string literal xxxxxxxxxxxxxx\n"
+                   ")x\", a), b);",
+                   getGoogleStyleWithColumns(20)));
+
+  EXPECT_EQ("fffffffffff(R\"x(\n"
+            "multiline raw string literal xxxxxxxxxxxxxx\n"
+            ")x\");",
+            format("fffffffffff(R\"x(\n"
+                   "multiline raw string literal xxxxxxxxxxxxxx\n"
+                   ")x\");",
+                   getGoogleStyleWithColumns(20)));
+  EXPECT_EQ("fffffffffff(R\"x(\n"
+            "multiline raw string literal xxxxxxxxxxxxxx\n"
+            ")x\" +\n"
+            "            bbbbbb);",
+            format("fffffffffff(R\"x(\n"
+                   "multiline raw string literal xxxxxxxxxxxxxx\n"
+                   ")x\" + bbbbbb);",
+                   getGoogleStyleWithColumns(20)));
 }
 
 TEST_F(FormatTest, SkipsUnknownStringLiterals) {
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D1544.1.patch
Type: text/x-patch
Size: 7064 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20130829/9fa28ee9/attachment.bin>