Details
-
Type:
Tests
-
Status: Reviewing (View workflow)
-
Priority:
major
-
Resolution: Unresolved
-
Affects Version/s: None
-
Fix Version/s: 64.1
-
Component/s: format_number
-
Labels:
-
Time Needed:Minutes
-
tracCc:pedberg
-
tracOwner:shane
-
tracProject:all
-
tracReporter:shane
-
tracStatus:accepted
Description
This behavior difference was discovered between ICU4C and ICU4J.
Sample code, ICU4C:
#include "common.h"
int main() {
ErrorCode status;
{
LocalPointer<DecimalFormat> df(dynamic_cast<DecimalFormat*>(
NumberFormat::createInstance("en", UNUM_CURRENCY, status)));
UnicodeString result;
df->format(1.23, result, status);
std::cout << "symbol: " << result << std::endl;
}
{
LocalPointer<DecimalFormat> df(dynamic_cast<DecimalFormat*>(
NumberFormat::createInstance("en", UNUM_CURRENCY_ISO, status)));
UnicodeString result;
df->format(1.23, result, status);
std::cout << "iso_code: " << result << std::endl;
}
{
LocalPointer<DecimalFormat> df(dynamic_cast<DecimalFormat*>(
NumberFormat::createInstance("en", UNUM_CURRENCY_PLURAL, status)));
UnicodeString result;
df->format(1.23, result, status);
std::cout << "plural: " << result << std::endl;
}
{
LocalPointer<DecimalFormat> df(dynamic_cast<DecimalFormat*>(
NumberFormat::createInstance("en", UNUM_CURRENCY, status)));
UnicodeString currencyCode = UnicodeString(df->getCurrency());
std::cout << "currency: \"" << currencyCode << "\"" << std::endl;
}
std::cout << u_errorName(status) << std::endl;
u_cleanup();
return 0;
}
ICU4J:
import java.text.ParsePosition;
import com.ibm.icu.text.*;
import com.ibm.icu.util.*;
public class NoCurrencyBehavior {
public static void main(String[] args) {
DecimalFormatSymbols EN = DecimalFormatSymbols.getInstance(ULocale.ENGLISH);
{
DecimalFormat df = (DecimalFormat) NumberFormat.getInstance(ULocale.ENGLISH, NumberFormat.CURRENCYSTYLE);
System.out.println("symbol: " + df.format(1.23));
}
{
DecimalFormat df = (DecimalFormat) NumberFormat.getInstance(ULocale.ENGLISH, NumberFormat.ISOCURRENCYSTYLE);
System.out.println("iso_code: " + df.format(1.23));
}
{
DecimalFormat df = (DecimalFormat) NumberFormat.getInstance(ULocale.ENGLISH, NumberFormat.PLURALCURRENCYSTYLE);
System.out.println("plural: " + df.format(1.23));
}
{
DecimalFormat df = (DecimalFormat) NumberFormat.getInstance(ULocale.ENGLISH, NumberFormat.CURRENCYSTYLE);
System.out.print("currency: " + df.getCurrency());
}
}
}
Consensus opinion: we should use the unknown currency "XXX", like ICU4J and JDK are already doing in the ISO code format, with the following output:
XXX 1.23 XXX 1.23 1.23 (unknown currency) currency: XXX
The plural long-name string, "unknown currency", comes from locale data.
Follow up with CLDR to add the currency symbol.
—
In addition, one more behavior difference was discovered.
ICU4C sample code:
#include "common.h"
int main() {
ErrorCode status;
const UChar pattern[] = {0xA4, u'0', u'.', u'#', u'#', 0};
const UChar XXX[] = {u'X', u'X', u'X', 0};
{
DecimalFormat df(pattern, {"en", status}, status);
UnicodeString result;
df.format(1.1, result, status);
std::cout << result << std::endl;
}
{
DecimalFormat df(pattern, {"en", status}, status);
df.setCurrency(XXX, status);
UnicodeString result;
df.format(1.1, result, status);
std::cout << result << std::endl;
}
{
LocalPointer<DecimalFormat> df(dynamic_cast<DecimalFormat*>(
NumberFormat::createCurrencyInstance("en", status)));
UnicodeString result;
df->format(1.1, result, status);
std::cout << result << std::endl;
}
std::cout << u_errorName(status) << std::endl;
u_cleanup();
return 0;
}
ICU4J:
import com.ibm.icu.text.*;
import com.ibm.icu.util.*;
public class RoundingNoCurrency {
public static void main(String[] args) {
{
DecimalFormat df = new DecimalFormat("\u00A40.##", DecimalFormatSymbols.getInstance(ULocale.ENGLISH));
System.out.println(df.format(1.1));
}
{
DecimalFormat df = new DecimalFormat("\u00A40.##", DecimalFormatSymbols.getInstance(ULocale.ENGLISH));
df.setCurrency(Currency.getInstance("XXX"));
System.out.println(df.format(1.1));
}
{
DecimalFormat df = (DecimalFormat) NumberFormat.getCurrencyInstance(ULocale.ENGLISH);
System.out.println(df.format(1.1));
}
}
}
Consensus opinion: Since we should apply the default currency when no currency is set, as stated in row 10, we should also apply XXX's rounding rules, which are a fixed two fraction digits.
Desired output:
XXX 1.10 XXX 1.10 1.10 (unknown currency)
