LibJS: Store Intl mathematical values as strings when appropriate

The IntlMV is meant to be arbitrarily precise. If the user provides a
string value to be formatted, we lose precision by converting extremely
large values to a double. We were never able to address this, as support
for arbitrary precision was a big FIXME. But ICU can handle it by just
passing the raw string on through.

Userland/Libraries/LibJS/Runtime/Intl/MathematicalValue.cpp

@@ -19,20 +19,20 @@ double MathematicalValue::as_number() const
     return m_value.get<double>();
 }
 
-bool MathematicalValue::is_bigint() const
+bool MathematicalValue::is_string() const
 {
-    return m_value.has<Crypto::SignedBigInteger>();
+    return m_value.has<String>();
 }
 
-Crypto::SignedBigInteger const& MathematicalValue::as_bigint() const
+String const& MathematicalValue::as_string() const
 {
-    VERIFY(is_bigint());
+    VERIFY(is_string());
-    return m_value.get<Crypto::SignedBigInteger>();
+    return m_value.get<String>();
 }
 
 bool MathematicalValue::is_mathematical_value() const
 {
-    return is_number() || is_bigint();
+    return is_number() || is_string();
 }
 
 bool MathematicalValue::is_positive_infinity() const
@@ -69,8 +69,8 @@ bool MathematicalValue::is_nan() const
         [](double value) -> ::Locale::NumberFormat::Value {
             return value;
         },
-        [](Crypto::SignedBigInteger const& value) -> ::Locale::NumberFormat::Value {
+        [](String const& value) -> ::Locale::NumberFormat::Value {
-            return MUST(value.to_base(10));
+            return value;
         },
         [](auto symbol) -> ::Locale::NumberFormat::Value {
             switch (symbol) {

Userland/Libraries/LibJS/Runtime/Intl/MathematicalValue.h

@@ -6,6 +6,7 @@
 
 #pragma once
 
+#include <AK/String.h>
 #include <AK/Variant.h>
 #include <LibCrypto/BigInt/SignedBigInteger.h>
 #include <LibJS/Runtime/BigInt.h>
@@ -31,7 +32,7 @@ public:
     {
     }
 
-    explicit MathematicalValue(Crypto::SignedBigInteger value)
+    explicit MathematicalValue(String value)
         : m_value(move(value))
     {
     }
@@ -44,15 +45,15 @@ public:
     MathematicalValue(Value value)
         : m_value(value.is_number()
                   ? value_from_number(value.as_double())
-                  : ValueType(value.as_bigint().big_integer()))
+                  : ValueType(MUST(value.as_bigint().big_integer().to_base(10))))
     {
     }
 
     bool is_number() const;
     double as_number() const;
 
-    bool is_bigint() const;
+    bool is_string() const;
-    Crypto::SignedBigInteger const& as_bigint() const;
+    String const& as_string() const;
 
     bool is_mathematical_value() const;
     bool is_positive_infinity() const;
@@ -63,7 +64,7 @@ public:
     ::Locale::NumberFormat::Value to_value() const;
 
 private:
-    using ValueType = Variant<double, Crypto::SignedBigInteger, Symbol>;
+    using ValueType = Variant<double, String, Symbol>;
 
     static ValueType value_from_number(double number);
 

Userland/Libraries/LibJS/Runtime/Intl/NumberFormat.cpp

@@ -192,7 +192,7 @@ ThrowCompletionOr<MathematicalValue> to_intl_mathematical_value(VM& vm, Value va
 
     // 2. If Type(primValue) is BigInt, return the mathematical value of primValue.
     if (primitive_value.is_bigint())
-        return primitive_value.as_bigint().big_integer();
+        return MUST(value.as_bigint().big_integer().to_base(10));
 
     // FIXME: The remaining steps are being refactored into a new Runtime Semantic, StringIntlMV.
     //        We short-circuit some of these steps to avoid known pitfalls.
@@ -212,6 +212,9 @@ ThrowCompletionOr<MathematicalValue> to_intl_mathematical_value(VM& vm, Value va
     // 6. Let mv be the MV, a mathematical value, of ? ToNumber(str), as described in 7.1.4.1.1.
     auto mathematical_value = TRY(primitive_value.to_number(vm)).as_double();
 
+    if (Value(mathematical_value).is_nan())
+        return MathematicalValue::Symbol::NotANumber;
+
     // 7. If mv is 0 and the first non white space code point in str is -, return negative-zero.
     if (mathematical_value == 0.0 && string.bytes_as_string_view().trim_whitespace(TrimMode::Left).starts_with('-'))
         return MathematicalValue::Symbol::NegativeZero;
@@ -225,7 +228,7 @@ ThrowCompletionOr<MathematicalValue> to_intl_mathematical_value(VM& vm, Value va
         return MathematicalValue::Symbol::NegativeInfinity;
 
     // 10. Return mv.
-    return mathematical_value;
+    return string;
 }
 
 // 15.5.19 PartitionNumberRangePattern ( numberFormat, x, y ), https://tc39.es/ecma402/#sec-partitionnumberrangepattern

Userland/Libraries/LibJS/Tests/builtins/Intl/NumberFormat/NumberFormat.prototype.format.js

@@ -46,11 +46,15 @@ describe("style=decimal", () => {
         expect(en.format(1)).toBe("1");
         expect(en.format(12)).toBe("12");
         expect(en.format(123)).toBe("123");
+        expect(en.format("987654321987654321")).toBe("987,654,321,987,654,321");
 
         const ar = new Intl.NumberFormat("ar");
         expect(ar.format(1)).toBe("\u0661");
         expect(ar.format(12)).toBe("\u0661\u0662");
         expect(ar.format(123)).toBe("\u0661\u0662\u0663");
+        expect(ar.format("987654321987654321")).toBe(
+            "\u0669\u0668\u0667\u066c\u0666\u0665\u0664\u066c\u0663\u0662\u0661\u066c\u0669\u0668\u0667\u066c\u0666\u0665\u0664\u066c\u0663\u0662\u0661"
+        );
     });
 
     test("integer digits", () => {

Userland/Libraries/LibLocale/NumberFormat.cpp

@@ -753,10 +753,8 @@ private:
 
         while (static_cast<bool>(formatted->nextPosition(position, status)) && icu_success(status)) {
             if (position.getCategory() == UFIELD_CATEGORY_NUMBER_RANGE_SPAN) {
-                if (position.getField() == 0)
+                auto& range = position.getField() == 0 ? start_range : end_range;
-                    start_range.emplace(position.getField(), position.getStart(), position.getLimit());
+                range = Range { position.getField(), position.getStart(), position.getLimit() };
-                else
-                    end_range.emplace(position.getField(), position.getStart(), position.getLimit());
             } else {
                 ranges.empend(position.getField(), position.getStart(), position.getLimit());
             }