Copy signbit for -nan

Is this useful? Doubtful. But doesn't hurt either, and it is useful for
toml-test.

Also write the signbit back, and don't write explicit +inf – just inf is
enough.

decode_test.go

@@ -423,6 +423,74 @@ func TestDecodeFloatOverflow(t *testing.T) {
 	}
 }
 
+func TestDecodeSignbit(t *testing.T) {
+	var m struct {
+		N1, N2   float64
+		I1, I2   float64
+		Z1, Z2   float64
+		ZF1, ZF2 float64
+	}
+	_, err := Decode(`
+n1 = nan
+n2 = -nan
+i1 = inf
+i2 = -inf
+z1 = 0
+z2 = -0
+zf1 = 0.0
+zf2 = -0.0
+`, &m)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if h := fmt.Sprintf("%v %v", m.N1, math.Signbit(m.N1)); h != "NaN false" {
+		t.Error("N1:", h)
+	}
+	if h := fmt.Sprintf("%v %v", m.I1, math.Signbit(m.I1)); h != "+Inf false" {
+		t.Error("I1:", h)
+	}
+	if h := fmt.Sprintf("%v %v", m.Z1, math.Signbit(m.Z1)); h != "0 false" {
+		t.Error("Z1:", h)
+	}
+	if h := fmt.Sprintf("%v %v", m.ZF1, math.Signbit(m.ZF1)); h != "0 false" {
+		t.Error("ZF1:", h)
+	}
+
+	if h := fmt.Sprintf("%v %v", m.N2, math.Signbit(m.N2)); h != "NaN true" {
+		t.Error("N2:", h)
+	}
+	if h := fmt.Sprintf("%v %v", m.I2, math.Signbit(m.I2)); h != "-Inf true" {
+		t.Error("I2:", h)
+	}
+	if h := fmt.Sprintf("%v %v", m.Z2, math.Signbit(m.Z2)); h != "0 false" { // Correct: -0 is same as 0
+		t.Error("Z2:", h)
+	}
+	if h := fmt.Sprintf("%v %v", m.ZF2, math.Signbit(m.ZF2)); h != "-0 true" {
+		t.Error("ZF2:", h)
+	}
+
+	buf := new(bytes.Buffer)
+	err = NewEncoder(buf).Encode(m)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	want := strings.ReplaceAll(`
+		N1 = nan
+		N2 = -nan
+		I1 = inf
+		I2 = -inf
+		Z1 = 0.0
+		Z2 = 0.0
+		ZF1 = 0.0
+		ZF2 = -0.0
+	`, "\t", "")[1:]
+	if buf.String() != want {
+		t.Errorf("\nwant:\n%s\nhave:\n%s", want, buf.String())
+	}
+}
+
 func TestDecodeSizedInts(t *testing.T) {
 	type table struct {
 		U8  uint8

encode.go

@@ -280,18 +280,30 @@ func (enc *Encoder) eElement(rv reflect.Value) {
 	case reflect.Float32:
 		f := rv.Float()
 		if math.IsNaN(f) {
+			if math.Signbit(f) {
+				enc.wf("-")
+			}
 			enc.wf("nan")
 		} else if math.IsInf(f, 0) {
-			enc.wf("%cinf", map[bool]byte{true: '-', false: '+'}[math.Signbit(f)])
+			if math.Signbit(f) {
+				enc.wf("-")
+			}
+			enc.wf("inf")
 		} else {
 			enc.wf(floatAddDecimal(strconv.FormatFloat(f, 'f', -1, 32)))
 		}
 	case reflect.Float64:
 		f := rv.Float()
 		if math.IsNaN(f) {
+			if math.Signbit(f) {
+				enc.wf("-")
+			}
 			enc.wf("nan")
 		} else if math.IsInf(f, 0) {
-			enc.wf("%cinf", map[bool]byte{true: '-', false: '+'}[math.Signbit(f)])
+			if math.Signbit(f) {
+				enc.wf("-")
+			}
+			enc.wf("inf")
 		} else {
 			enc.wf(floatAddDecimal(strconv.FormatFloat(f, 'f', -1, 64)))
 		}

encode_test.go

@@ -470,7 +470,7 @@ func TestEncodeNaN(t *testing.T) {
 		Nan float32 `toml:"nan"`
 		Inf float32 `toml:"inf"`
 	}{float32(math.NaN()), float32(math.Inf(-1))}
-	encodeExpected(t, "", s1, "nan = nan\ninf = +inf\n", nil)
+	encodeExpected(t, "", s1, "nan = nan\ninf = inf\n", nil)
 	encodeExpected(t, "", s2, "nan = nan\ninf = -inf\n", nil)
 }
 

parse.go

@@ -2,6 +2,7 @@ package toml
 
 import (
 	"fmt"
+	"math"
 	"os"
 	"strconv"
 	"strings"
@@ -323,7 +324,9 @@ func (p *parser) valueFloat(it item) (any, tomlType) {
 		p.panicItemf(it, "Invalid float %q: '.' must be followed by one or more digits", it.val)
 	}
 	val := strings.Replace(it.val, "_", "", -1)
-	if val == "+nan" || val == "-nan" { // Go doesn't support this, but TOML spec does.
+	signbit := false
+	if val == "+nan" || val == "-nan" {
+		signbit = val == "-nan"
 		val = "nan"
 	}
 	num, err := strconv.ParseFloat(val, 64)
@@ -334,6 +337,9 @@ func (p *parser) valueFloat(it item) (any, tomlType) {
 			p.panicItemf(it, "Invalid float value: %q", it.val)
 		}
 	}
+	if signbit {
+		num = math.Copysign(num, -1)
+	}
 	return num, p.typeOfPrimitive(it)
 }