Fix float precision (#649)

The issue is that numbers like
2.01 or 3.01 can not be precisely represented with binary floating point
numbers.

This replaces all occurrences of 'std::numeric_limits<T>::digits10 + 1' with
'std::numeric_limits<T>::max_digits10'.

Background:
Using 'std::numeric_limits<T>::digits10 + 1' is not precise enough.
Converting a 'float' into a 'string' and back to a 'float' will not always
produce the original 'float' value. To guarantee that the 'string'
representation has sufficient precision the value
'std::numeric_limits<T>::max_digits10' has to be used.

include/yaml-cpp/node/convert.h

@@ -93,7 +93,7 @@ struct convert<_Null> {
   struct convert<type> {                                                 \
     static Node encode(const type& rhs) {                                \
       std::stringstream stream;                                          \
-      stream.precision(std::numeric_limits<type>::digits10 + 1);         \
+      stream.precision(std::numeric_limits<type>::max_digits10);         \
       stream << rhs;                                                     \
       return Node(stream.str());                                         \
     }                                                                    \

src/emitterstate.cpp

@@ -24,8 +24,8 @@ EmitterState::EmitterState()
   m_seqFmt.set(Block);
   m_mapFmt.set(Block);
   m_mapKeyFmt.set(Auto);
-  m_floatPrecision.set(std::numeric_limits<float>::digits10 + 1);
+  m_floatPrecision.set(std::numeric_limits<float>::max_digits10);
-  m_doublePrecision.set(std::numeric_limits<double>::digits10 + 1);
+  m_doublePrecision.set(std::numeric_limits<double>::max_digits10);
 }
 
 EmitterState::~EmitterState() {}
@@ -349,7 +349,7 @@ bool EmitterState::SetMapKeyFormat(EMITTER_MANIP value, FmtScope::value scope) {
 }
 
 bool EmitterState::SetFloatPrecision(std::size_t value, FmtScope::value scope) {
-  if (value > std::numeric_limits<float>::digits10 + 1)
+  if (value > std::numeric_limits<float>::max_digits10)
     return false;
   _Set(m_floatPrecision, value, scope);
   return true;
@@ -357,7 +357,7 @@ bool EmitterState::SetFloatPrecision(std::size_t value, FmtScope::value scope) {
 
 bool EmitterState::SetDoublePrecision(std::size_t value,
                                       FmtScope::value scope) {
-  if (value > std::numeric_limits<double>::digits10 + 1)
+  if (value > std::numeric_limits<double>::max_digits10)
     return false;
   _Set(m_doublePrecision, value, scope);
   return true;

test/integration/emitter_test.cpp

@@ -901,18 +901,18 @@ TEST_F(EmitterTest, SingleChar) {
 
 TEST_F(EmitterTest, DefaultPrecision) {
   out << BeginSeq;
-  out << 1.234f;
+  out << 1.3125f;
-  out << 3.14159265358979;
+  out << 1.23455810546875;
   out << EndSeq;
-  ExpectEmit("- 1.234\n- 3.14159265358979");
+  ExpectEmit("- 1.3125\n- 1.23455810546875");
 }
 
 TEST_F(EmitterTest, SetPrecision) {
   out << BeginSeq;
-  out << FloatPrecision(3) << 1.234f;
+  out << FloatPrecision(3) << 1.3125f;
-  out << DoublePrecision(6) << 3.14159265358979;
+  out << DoublePrecision(6) << 1.23455810546875;
   out << EndSeq;
-  ExpectEmit("- 1.23\n- 3.14159");
+  ExpectEmit("- 1.31\n- 1.23456");
 }
 
 TEST_F(EmitterTest, DashInBlockContext) {

test/node/node_test.cpp

@@ -391,7 +391,13 @@ TEST(NodeTest, AutoBoolConversion) {
   EXPECT_TRUE(!!node["foo"]);
 }
 
-TEST(NodeTest, FloatingPrecision) {
+TEST(NodeTest, FloatingPrecisionFloat) {
+  const float x = 0.123456789;
+  Node node = Node(x);
+  EXPECT_EQ(x, node.as<float>());
+}
+
+TEST(NodeTest, FloatingPrecisionDouble) {
   const double x = 0.123456789;
   Node node = Node(x);
   EXPECT_EQ(x, node.as<double>());
@@ -452,108 +458,108 @@ class NodeEmitterTest : public ::testing::Test {
 TEST_F(NodeEmitterTest, SimpleFlowSeqNode) {
   Node node;
   node.SetStyle(EmitterStyle::Flow);
-  node.push_back(1.01);
+  node.push_back(1.5);
-  node.push_back(2.01);
+  node.push_back(2.25);
-  node.push_back(3.01);
+  node.push_back(3.125);
 
-  ExpectOutput("[1.01, 2.01, 3.01]", node);
+  ExpectOutput("[1.5, 2.25, 3.125]", node);
 }
 
 TEST_F(NodeEmitterTest, NestFlowSeqNode) {
   Node node, cell0, cell1;
 
-  cell0.push_back(1.01);
+  cell0.push_back(1.5);
-  cell0.push_back(2.01);
+  cell0.push_back(2.25);
-  cell0.push_back(3.01);
+  cell0.push_back(3.125);
 
-  cell1.push_back(4.01);
+  cell1.push_back(4.5);
-  cell1.push_back(5.01);
+  cell1.push_back(5.25);
-  cell1.push_back(6.01);
+  cell1.push_back(6.125);
 
   node.SetStyle(EmitterStyle::Flow);
   node.push_back(cell0);
   node.push_back(cell1);
 
-  ExpectOutput("[[1.01, 2.01, 3.01], [4.01, 5.01, 6.01]]", node);
+  ExpectOutput("[[1.5, 2.25, 3.125], [4.5, 5.25, 6.125]]", node);
 }
 
 TEST_F(NodeEmitterTest, MixBlockFlowSeqNode) {
   Node node, cell0, cell1;
 
   cell0.SetStyle(EmitterStyle::Flow);
-  cell0.push_back(1.01);
+  cell0.push_back(1.5);
-  cell0.push_back(2.01);
+  cell0.push_back(2.25);
-  cell0.push_back(3.01);
+  cell0.push_back(3.125);
 
-  cell1.push_back(4.01);
+  cell1.push_back(4.5);
-  cell1.push_back(5.01);
+  cell1.push_back(5.25);
-  cell1.push_back(6.01);
+  cell1.push_back(6.125);
 
   node.SetStyle(EmitterStyle::Block);
   node.push_back(cell0);
   node.push_back(cell1);
 
-  ExpectOutput("- [1.01, 2.01, 3.01]\n-\n  - 4.01\n  - 5.01\n  - 6.01", node);
+  ExpectOutput("- [1.5, 2.25, 3.125]\n-\n  - 4.5\n  - 5.25\n  - 6.125", node);
 }
 
 TEST_F(NodeEmitterTest, NestBlockFlowMapListNode) {
   Node node, mapNode, blockNode;
 
-  node.push_back(1.01);
+  node.push_back(1.5);
-  node.push_back(2.01);
+  node.push_back(2.25);
-  node.push_back(3.01);
+  node.push_back(3.125);
 
   mapNode.SetStyle(EmitterStyle::Flow);
   mapNode["position"] = node;
 
-  blockNode.push_back(1.01);
+  blockNode.push_back(1.0625);
   blockNode.push_back(mapNode);
 
-  ExpectOutput("- 1.01\n- {position: [1.01, 2.01, 3.01]}", blockNode);
+  ExpectOutput("- 1.0625\n- {position: [1.5, 2.25, 3.125]}", blockNode);
 }
 
 TEST_F(NodeEmitterTest, NestBlockMixMapListNode) {
   Node node, mapNode, blockNode;
 
-  node.push_back(1.01);
+  node.push_back(1.5);
-  node.push_back(2.01);
+  node.push_back(2.25);
-  node.push_back(3.01);
+  node.push_back(3.125);
 
   mapNode.SetStyle(EmitterStyle::Flow);
   mapNode["position"] = node;
 
-  blockNode["scalar"] = 1.01;
+  blockNode["scalar"] = 1.0625;
   blockNode["object"] = mapNode;
 
   ExpectAnyOutput(blockNode,
-                  "scalar: 1.01\nobject: {position: [1.01, 2.01, 3.01]}",
+                  "scalar: 1.0625\nobject: {position: [1.5, 2.25, 3.125]}",
-                  "object: {position: [1.01, 2.01, 3.01]}\nscalar: 1.01");
+                  "object: {position: [1.5, 2.25, 3.125]}\nscalar: 1.5");
 }
 
 TEST_F(NodeEmitterTest, NestBlockMapListNode) {
   Node node, mapNode;
 
-  node.push_back(1.01);
+  node.push_back(1.5);
-  node.push_back(2.01);
+  node.push_back(2.25);
-  node.push_back(3.01);
+  node.push_back(3.125);
 
   mapNode.SetStyle(EmitterStyle::Block);
   mapNode["position"] = node;
 
-  ExpectOutput("position:\n  - 1.01\n  - 2.01\n  - 3.01", mapNode);
+  ExpectOutput("position:\n  - 1.5\n  - 2.25\n  - 3.125", mapNode);
 }
 
 TEST_F(NodeEmitterTest, NestFlowMapListNode) {
   Node node, mapNode;
 
-  node.push_back(1.01);
+  node.push_back(1.5);
-  node.push_back(2.01);
+  node.push_back(2.25);
-  node.push_back(3.01);
+  node.push_back(3.125);
 
   mapNode.SetStyle(EmitterStyle::Flow);
   mapNode["position"] = node;
 
-  ExpectOutput("{position: [1.01, 2.01, 3.01]}", mapNode);
+  ExpectOutput("{position: [1.5, 2.25, 3.125]}", mapNode);
 }
 }
 }