A related class is the ChoiceFormat (§) class described in the Formatting Messages chapter. It maps ranges of numeric values to strings.
NumberFormat is the abstract base class for all number formats. It provides an interface for formatting and parsing numbers. It also provides methods to determine which locales have number formats, and what their names are. NumberFormat helps format and parse numbers for any locale. Your program can be written to be completely independent of the locale conventions for decimal points or thousands-separators. It can also be written to be independent of the particular decimal digits used or whether the number format is a decimal. A normal decimal number can also be displayed as a currency or as a percentage.
format a number for the current Locale, use one of the static factory
methods to create a format, then call a format method to format it. To
format a number for a different Locale, specify the Locale in the call
to createInstance(). You can control the numbering system to be used for number formatting by creating a Locale that uses the @numbers keyword defined. For example, by default, the Thai locale "th" uses the western digits 0-9. To create a number format that uses the native Thai digits instead, first create a locale with "@numbers=thai" defined. See the description on Locales for details.
The following methods are used for instantiating NumberFormat objects:
To create a format for spelled-out numbers, use a constructor on RuleBasedNumberFormat (§).
Currency formatting, i.e., the formatting of monetary values, combines a number with a suitable display symbol or name for a currency. By default, the currency is set from the locale data from when the currency format instance is created, based on the country code in the locale ID. However, for all but trivial uses, this is fragile because countries change currencies over time, and the locale data for a particular country may not be available.
For proper currency formatting, both the number and the currency must be specified. Aside from achieving reliably correct results, this also allows to format monetary values in any currency with the format of any locale, like in exchange rate lists. If the locale data does not contain display symbols or names for a currency, then the 3-letter ISO code itself is displayed.
The locale ID and the currency code are effectively independent: The locale ID defines the general format for the numbers, and whether the currency symbol or name is displayed before or after the number, while the currency code selects the actual currency with its symbol, name, number of digits, and rounding mode.
In ICU and Java, the currency is specified in the form of a 3-letter ISO 4217 code. For example, the code "USD" represents the US Dollar and "EUR" represents the Euro currency.
In terms of APIs, the currency code is set as an attribute on a number format object (on a currency instance), while the number value is passed into each format() call or returned from parse() as usual.
The functionality of Currency and setCurrency() is more advanced in ICU than in the base JDK. When using ICU, setting the currency automatically adjusts the number format object appropriately, i.e., it sets not only the currency symbol and display name, but also the correct number of fraction digits and the correct rounding mode. This is not the case with the base JDK. See the API references for more details.
There is ICU4C sample code at icu/source/samples/numfmt/main.cpp which illustrates the use of NumberFormat.setCurrency().
You can also control the display of numbers with methods such as getMinimumFractionDigits. If you want even more control over the format or parsing, or want to give your users more control, cast the NumberFormat returned from the factory methods to a DecimalNumberFormat. This works for the vast majority of countries.
You can also use forms of the parse and format methods with ParsePosition and UFieldPosition to enable you to:
For example, you can align numbers in two ways:
NumberFormat can produce many of the same formats as printf.
DecimalFormat is a NumberFormat that converts numbers into strings using the decimal numbering system. This is the formatter that provides standard number formatting and parsing services for most usage scenarios in most locales. In order to access features of DecimalFormat not exposed in the NumberFormat API, you may need to cast your NumberFormat object to a DecimalFormat. You may also construct a DecimalFormat directly, but this is not recommended because it can hinder proper localization.
For a complete description of DecimalFormat, including the pattern syntax, formatting and parsing behavior, and available API, see the ICU4J DecimalFormat API or ICU4C DecimalFormat API documentation.
DecimalFormatSymbols specifies the exact characters a DecimalFormat uses for various parts of a number (such as the characters to use for the digits, the character to use as the decimal point, or the character to use as the minus sign).
This class represents the set of symbols needed by DecimalFormat to format numbers. DecimalFormat creates its own instance of DecimalFormatSymbols from its locale data. The DecimalFormatSymbols can be adopted by a DecimalFormat instance, or it can be specified when a DecimalFormat is created. If you need to change any of these symbols, can get the DecimalFormatSymbols object from your DecimalFormat and then modify it.
RuleBasedNumberFormat can format and parse numbers in spelled-out format, e.g. "one hundred and thirty-four". For example:
RuleBasedNumberFormat is based on rules describing how to format a number. The rule syntax is designed primarily for formatting and parsing numbers as spelled-out text, though other kinds of formatting are possible. As a convenience, custom API is provided to allow selection from three predefined rule definitions, when available: SPELLOUT, ORDINAL, and DURATION. Users can request formatters either by providing a locale and one of these predefined rule selectors, or by specifying the rule definitions directly.
Unlike the other standard number formats, there is no corresponding factory method on NumberFormat. Instead, RuleBasedNumberFormat objects are instantiated via constructors. Constructors come in two flavors, ones that take rule text, and ones that take one of the predefined selectors. Constructors that do not take a Locale parameter use the current default locale.
The following constructors are available:
RuleBasedNumberFormat can be used like other NumberFormats. For example, in Java:
Rule descriptions can provide multiple named rule sets, for example, the rules for en_US spellout provides a '%simplified' rule set that displays text without commas or the word 'and'. Rule sets can be queried and set on a RuleBasedNumberFormat. This lets you customize a RuleBasedNumberFormat for use through its inherited NumberFormat API. For example, in Java:
You can also format a number specifying the ruleset directly, using an additional overload of format provided by RuleBasedNumberFormat. For example, in Java:
The following example provides a quick look at the RuleBasedNumberFormat rule syntax.
These rules format a number using standard decimal place-value notation, but using words instead of digits, e.g. 123.4 formats as 'one two three point four':
Rulesets are invoked by first applying negative and fractional rules, and then using a recursive process. It starts by finding the rule whose range includes the current value and applying that rule. If the rule so directs, it emits text, including text obtained by recursing on new values as directed by the rule. As you can see, the rules are designed to accomodate recursive processing of numbers, and so are best suited for formatting numbers in ways that are inherently recursive.
C/C++: See icu/source/samples/numfmt/ in the ICU source distribution for code samples showing the use of ICU number formatting.