clang-format: Add column layout formatting for braced lists

With this patch, braced lists (with more than 3 elements are formatted in a
column layout if possible). E.g.:

  static const uint16_t CallerSavedRegs64Bit[] = {
    X86::RAX, X86::RDX, X86::RCX, X86::RSI, X86::RDI,
    X86::R8,  X86::R9,  X86::R10, X86::R11, 0
  };

Required other changes:
- FormatTokens can now have a special role that contains extra data and can do
  special formattings. A comma separated list is currently the only
  implementation.
- Move penalty calculation entirely into ContinuationIndenter (there was a last
  piece still in UnwrappedLineFormatter).

Review: http://llvm-reviews.chandlerc.com/D1457

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@189018 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 184 insertions, 0 deletions

diff --git a/lib/Format/FormatToken.cpp b/lib/Format/FormatToken.cpp
new file mode 100644
index 0000000000..4e232afda5
--- /dev/null
+++ b/lib/Format/FormatToken.cpp
@@ -0,0 +1,184 @@
+//===--- FormatToken.cpp - Format C++ code --------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+/// \brief This file implements specific functions of \c FormatTokens and their
+/// roles.
+///
+//===----------------------------------------------------------------------===//
+
+#include "FormatToken.h"
+#include "ContinuationIndenter.h"
+#include "clang/Format/Format.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/Support/Debug.h"
+
+namespace clang {
+namespace format {
+
+TokenRole::~TokenRole() {}
+
+void TokenRole::precomputeFormattingInfos(const FormatToken *Token) {}
+
+unsigned CommaSeparatedList::format(LineState &State,
+                                    ContinuationIndenter *Indenter,
+                                    bool DryRun) {
+  if (!State.NextToken->Previous || !State.NextToken->Previous->Previous ||
+      Commas.size() <= 2)
+    return 0;
+
+  // Ensure that we start on the opening brace.
+  const FormatToken *LBrace = State.NextToken->Previous->Previous;
+  if (LBrace->isNot(tok::l_brace) ||
+      LBrace->Next->Type == TT_DesignatedInitializerPeriod)
+    return 0;
+
+  // Find the best ColumnFormat, i.e. the best number of columns to use.
+  unsigned RemainingCharacters = Style.ColumnLimit - State.Stack.back().Indent;
+  const ColumnFormat *Format = getColumnFormat(RemainingCharacters);
+  if (!Format)
+    return 0;
+
+  // Format the entire list.
+  unsigned Penalty = 0;
+  unsigned Column = 0;
+  unsigned Item = 0;
+  while (State.NextToken != LBrace->MatchingParen) {
+    bool NewLine = false;
+    unsigned ExtraSpaces = 0;
+
+    // If the previous token was one of our commas, we are now on the next item.
+    if (Item < Commas.size() && State.NextToken->Previous == Commas[Item]) {
+      if (!State.NextToken->isTrailingComment()) {
+        ExtraSpaces += Format->ColumnSizes[Column] - ItemLengths[Item];
+        ++Column;
+      }
+      ++Item;
+    }
+
+    if (Column == Format->Columns || State.NextToken->MustBreakBefore) {
+      Column = 0;
+      NewLine = true;
+    }
+
+    // Place token using the continuation indenter and store the penalty.
+    Penalty += Indenter->addTokenToState(State, NewLine, DryRun, ExtraSpaces);
+  }
+  return Penalty;
+}
+
+// Returns the lengths in code points between Begin and End (both included),
+// assuming that the entire sequence is put on a single line.
+static unsigned CodePointsBetween(const FormatToken *Begin,
+                                  const FormatToken *End) {
+  return End->TotalLength - Begin->TotalLength + Begin->CodePointCount;
+}
+
+void CommaSeparatedList::precomputeFormattingInfos(const FormatToken *Token) {
+  if (!Token->MatchingParen)
+    return;
+
+  FormatToken *ItemBegin = Token->Next;
+  SmallVector<bool, 8> MustBreakBeforeItem;
+
+  // The lengths of an item if it is put at the end of the line. This includes
+  // trailing comments which are otherwise ignored for column alignment.
+  SmallVector<unsigned, 8> EndOfLineItemLength;
+
+  for (unsigned i = 0, e = Commas.size() + 1; i != e; ++i) {
+    // Skip comments on their own line.
+    while (ItemBegin->HasUnescapedNewline && ItemBegin->isTrailingComment())
+      ItemBegin = ItemBegin->Next;
+
+    MustBreakBeforeItem.push_back(ItemBegin->MustBreakBefore);
+    const FormatToken *ItemEnd = NULL;
+    if (i == Commas.size()) {
+      ItemEnd = Token->MatchingParen;
+      const FormatToken *NonCommentEnd = ItemEnd->getPreviousNonComment();
+      ItemLengths.push_back(CodePointsBetween(ItemBegin, NonCommentEnd));
+      if (Style.Cpp11BracedListStyle) {
+        // In Cpp11 braced list style, the } and possibly other subsequent
+        // tokens will need to stay on a line with the last element.
+        while (ItemEnd->Next && !ItemEnd->Next->CanBreakBefore)
+          ItemEnd = ItemEnd->Next;
+      } else {
+        // In other braced lists styles, the "}" can be wrapped to the new line.
+        ItemEnd = Token->MatchingParen->Previous;
+      }
+    } else {
+      ItemEnd = Commas[i];
+      // The comma is counted as part of the item when calculating the length.
+      ItemLengths.push_back(ItemEnd->TotalLength - ItemBegin->TotalLength +
+                            ItemBegin->CodePointCount);
+      // Consume trailing comments so the are included in EndOfLineItemLength.
+      if (ItemEnd->Next && !ItemEnd->Next->HasUnescapedNewline &&
+          ItemEnd->Next->isTrailingComment())
+        ItemEnd = ItemEnd->Next;
+    }
+    EndOfLineItemLength.push_back(CodePointsBetween(ItemBegin, ItemEnd));
+    ItemBegin = ItemEnd->Next;
+  }
+
+  // We can never place more than ColumnLimit / 3 items in a row (because of the
+  // spaces and the comma).
+  for (unsigned Columns = 1; Columns <= Style.ColumnLimit / 3; ++Columns) {
+    ColumnFormat Format;
+    Format.Columns = Columns;
+    Format.ColumnSizes.resize(Columns);
+    Format.LineCount = 0;
+    bool HasRowWithSufficientColumns = false;
+    unsigned Column = 0;
+    for (unsigned i = 0, e = ItemLengths.size(); i != e; ++i) {
+      if (MustBreakBeforeItem[i] || Column == Columns) {
+        ++Format.LineCount;
+        Column = 0;
+      }
+      if (Column == Columns - 1)
+        HasRowWithSufficientColumns = true;
+      unsigned length =
+          (Column == Columns - 1) ? EndOfLineItemLength[i] : ItemLengths[i];
+      Format.ColumnSizes[Column] =
+          std::max(Format.ColumnSizes[Column], length);
+      ++Column;
+    }
+    // If all rows are terminated early (e.g. by trailing comments), we don't
+    // need to look further.
+    if (!HasRowWithSufficientColumns)
+      break;
+    Format.TotalWidth = Columns - 1; // Width of the N-1 spaces.
+    for (unsigned i = 0; i < Columns; ++i) {
+      Format.TotalWidth += Format.ColumnSizes[i];
+    }
+
+    // Ignore layouts that are bound to violate the column limit.
+    if (Format.TotalWidth > Style.ColumnLimit)
+      continue;
+
+    Formats.push_back(Format);
+  }
+}
+
+const CommaSeparatedList::ColumnFormat *
+CommaSeparatedList::getColumnFormat(unsigned RemainingCharacters) const {
+  const ColumnFormat *BestFormat = NULL;
+  for (SmallVector<ColumnFormat, 4>::const_reverse_iterator
+           I = Formats.rbegin(),
+           E = Formats.rend();
+       I != E; ++I) {
+    if (I->TotalWidth <= RemainingCharacters) {
+      if (BestFormat && I->LineCount > BestFormat->LineCount)
+        break;
+      BestFormat = &*I;
+    }
+  }
+  return BestFormat;
+}
+
+} // namespace format
+} // namespace clang