Fix #14 - format negative zero as zero.

Negative zero is a construct of IEEE 754 floating point, not a mathematical
concept that (most) humans should care about. And since you may see tick values
that are very small negative numbers due to limited precision, such as -8e-17,
it’s helpful to format these as zero (“0.00”), not negative zero (“-0.00”).

src/locale.js

@@ -64,6 +64,21 @@ export default function(locale) {
         // Perform the initial formatting.
         value = formatType(value, precision);
 
+        // If the original value was negative, it may be rounded to zero during
+        // formatting; treat this as (positive) zero.
+        if (valueNegative) {
+          var i = -1, n = value.length, c;
+          valueNegative = false;
+          while (++i < n) {
+            if (c = value.charCodeAt(i), (48 < c && c < 58)
+                || (type === "x" && 96 < c && c < 103)
+                || (type === "X" && 64 < c && c < 71)) {
+              valueNegative = true;
+              break;
+            }
+          }
+        }
+
         // Compute the prefix and suffix.
         valuePrefix = (valueNegative ? (sign === "(" ? sign : "-") : sign === "-" || sign === "(" ? "" : sign) + valuePrefix;
         valueSuffix = valueSuffix + (type === "s" ? prefixes[8 + prefixExponent / 3] : "") + (valueNegative && sign === "(" ? ")" : "");

test/format-type-d-test.js

@@ -225,7 +225,8 @@ tape("format(\"-,d\") explicitly only use a sign for negative numbers", function
   test.end();
 });
 
-tape("format(\"d\") can format negative zero", function(test) {
-  test.equal(format.format("1d")(-0), "-0");
+tape("format(\"d\") can format negative zero as zero", function(test) {
+  test.equal(format.format("1d")(-0), "0");
+  test.equal(format.format("1d")(-1e-12), "0");
   test.end();
 });

test/format-type-e-test.js

@@ -16,6 +16,12 @@ tape("format(\"e\") can output exponent notation", function(test) {
   test.end();
 });
 
+tape("format(\"e\") can format negative zero as zero", function(test) {
+  test.equal(format.format("1e")(-0), "0.000000e+0");
+  test.equal(format.format("1e")(-1e-12), "-1.000000e-12");
+  test.end();
+});
+
 tape("format(\",e\") does not group Infinity", function(test) {
   test.equal(format.format(",e")(Infinity), "Infinity");
   test.end();

test/format-type-f-test.js

@@ -40,8 +40,9 @@ tape("format(\"f\") can display integers in fixed-point notation", function(test
   test.end();
 });
 
-tape("format(\"f\") can format negative zero", function(test) {
-  test.equal(format.format("1f")(-0), "-0.000000");
+tape("format(\"f\") can format negative zero as zero", function(test) {
+  test.equal(format.format("1f")(-0), "0.000000");
+  test.equal(format.format("1f")(-1e-12), "0.000000");
   test.end();
 });
 

test/format-type-x-test.js

@@ -50,6 +50,11 @@ tape("format(\"x\") rounds non-integers", function(test) {
   test.end();
 });
 
+tape("format(\"x\") does not consider -0xeee to be positive", function(test) {
+  test.equal(format.format("x")(-0xeee), "-eee");
+  test.end();
+});
+
 tape("format(\"X\") returns the expected hexadecimal (uppercase) string", function(test) {
   test.equal(format.format("X")(0xdeadbeef), "DEADBEEF");
   test.end();
@@ -60,6 +65,11 @@ tape("format(\"#X\") returns the expected hexadecimal (uppercase) string with pr
   test.end();
 });
 
+tape("format(\"X\") does not consider -0xeee to be positive", function(test) {
+  test.equal(format.format("X")(-0xeee), "-EEE");
+  test.end();
+});
+
 tape("format(\"#[width]x\") considers the prefix", function(test) {
   test.equal(format.format("20x")(0xdeadbeef),   "            deadbeef");
   test.equal(format.format("#20x")(0xdeadbeef),  "          0xdeadbeef");