[ICU4C] NumberFormat/DecimalFormat treats certain floating-point values as longs instead of doubles

Question

NumberFormat/DecimalFormat doesn't seem to parse strings with the "#.0" format (where # is any number) as a double. The following code illustrates this:

#include <cstdio>
#include <iostream>
#include <unicode/decimfmt.h>
#include <unicode/numfmt.h>
#include <unicode/unistr.h>
#include <unicode/ustream.h>

int main() {
    UErrorCode status = U_ZERO_ERROR;
    // DecimalFormat doesn't work either
    NumberFormat* f = NumberFormat::createInstance(status);
    f->setGroupingUsed(false);
    f->setParseIntegerOnly(false);

    UnicodeString str("2.0"); // Change to "2.#" for it to work, where # is any non-zero number
    Formattable formattable;
    f->parse(str, formattable, status);
    if (U_FAILURE(status)) {
        printf("ERROR: %s\n", u_errorName(status));
        delete f;
        return 1;
    } else {
        if (formattable.getType() == Formattable::kDouble) {
            printf("kDouble: %f\n", formattable.getDouble());
        } else if ((formattable.getType() == Formattable::kLong)) {
            printf("kLong: %d\n", formattable.getLong());
        } else {
            printf("ERROR: unexpected type: %d\n", formattable.getType());
        }
    }

    str.remove(); // Clear the string
    f->format(2.0f, str);
    std::cout << "formatted: \"" << str << '\"' << std::endl; // Outputs "2"
    delete f;

    return 0;
}

When parsing "2.0", the Formattable's type is 2 (Formattable::Type::kLong). When parsing "2.1" the Formattable's type is 1 (Formattable:Type::kDouble) - as it should be for both strings.

Issues also arise when you try to format a float as a UnicodeString (e.g. a float 2.0 gets formatted to "2").

So: How can I parse/format any double number without it being interpreted as an integer in ICU?

Answer 1

You can call formattable.getDouble(status) - the only reason getType returns long is that the particular value will fit in a long.

As to format, if you call f->setMinimumFractionDigits(1); before format, you will get "2.0" with your code, setting the min digits to 2 gives "2.00" etc.

hth