Number Skeletons

Contents

1. Overview
2. Syntax
3. Examples
4. Skeleton Stems and Options
   1. Notation
      1. Scientific and Engineering Notation: Long Form
      2. Scientific and Engineering Notation: Concise Form
   2. Unit
   3. Per Unit
   4. Unit Width
   5. Precision
      1. Fraction Precision
      2. Significant Digits Precision
      3. Trailing Zero Display
      4. Wildcard Character
   6. Rounding Mode
   7. Integer Width
   8. Scale
   9. Grouping
   10. Symbols
   11. Sign Display
   12. Decimal Separator Display

Overview

Number skeletons are a locale-agnostic way to configure a NumberFormatter in ICU. Number skeletons work in MessageFormat.

Number skeletons consist of case-sensitive tokens that correspond to settings in ICU NumberFormatter. For example, to format a currency in compact notation with the sign always shown, you could use this skeleton:

sign-always compact-short currency/GBP

Since ICU 67, you can also use more concise syntax:

+! K currency/GBP

To use a skeleton in MessageFormat, use the “number” type and prefix the skeleton with ::

{0, number, :: +! K currency/GBP}

The ICU toSkeleton() API outputs the long-form skeletons, but all parts of ICU that read user-specified number skeletons accept both long-form and concise skeletons.

Syntax

A token consists of a stem and zero or more options. The stem is what occurs before the first "/" character in a token, and the options are each of the subsequent "/"-delimited strings. For example, "compact-short" and “currency” are stems, and "GBP" is an option.

Tokens are space-separated, with exceptions for concise skeletons listed at the end of this document.

Stems might also be dynamic strings (not a fixed list); these are called blueprint stems. For example, to format a number with 2-3 significant digits, you could use the following stem:

@@#

A few examples of number skeletons are shown below. The list of available stems and options can be found below in Skeleton Stems and Options.

Examples

Skeleton Stems and Options

The full set of features supported by number skeletons is listed by category below.

Notation

Use one of the following stems to select compact or simple notation:

• compact-short or K (concise)
• compact-long or KK (concise)
• notation-simple (or omit since this is default)

There are two ways to select scientific or engineering notation: using long-form syntax or concise syntax.

Scientific and Engineering Notation: Long Form

Start with the stem scientific or engineering. Those stems take the following optional options:

• /sign-xxx sets the sign display option for the exponent; see Sign.
• /*ee sets exponent digits to “at least 2”; use /*eee for at least 3 digits, etc.
  • Prior to ICU 67, use /+ee instead of /*ee.

For example, all the following skeletons are valid:

• scientific
• scientific/sign-always
• scientific/*ee
• scientific/*ee/sign-always

Scientific and Engineering Notation: Concise Form

The following are examples of concise form:

More precisely:

1. Start with E for scientific or EE for engineering.
2. Allow either +! or +? as a concise sign display option.
3. Expect one or more 0s. If more than one, set minimum integer digits.

Unit

The supported types of units are percent, currency, and measurement units. The following skeleton tokens are accepted:

• percent or % (concise)
• Special: %x100 to scale the number by 100 and then format with percent
• permille
• base-unit
• currency/XXX
• measure-unit/aaaa-bbbb or unit/bbb (concise)

The percent, permille, and base-unit stems do not take any options.

The currency stem takes one required option: the three-letter ISO code of the currency to be formatted.

The measure-unit stem takes one required option: the unit identifier of the unit to be formatted. The full unit identifier is required: both the type and the subtype (for example, length-meter).

The unit stem is an alternative to measure-unit that accepts a core unit identifier with the subtype but not the type (for example, meter instead of length-meter). It also supports variations allowed by UTS 35, including the per unit with the -per- infix (for example, unit/furlong-per-second).

Per Unit

To specify a unit to put in the denominator, use the following skeleton token. As with the measure-unit stem, pass the unit identifier as the option:

• per-measure-unit/aaaa-bbbb

Note that if the unit stem is used, the denominator can be placed in the same token as the numerator.

Unit Width

The unit width can be specified by the following stems:

• unit-width-narrow
• unit-width-short
• unit-width-full-name
• unit-width-iso-code
• unit-width-hidden

For more details, see UNumberUnitWidth.

Precision

The precision category has more blueprint stems than most other categories; they are documented in detail below. The following non-blueprint stems are accepted:

• precision-integer (round to the nearest integer) — accepts fraction-precision options
• precision-unlimited (do not perform rounding; display all digits)
• precision-increment/dddd (round to dddd, a decimal number) — see below
• precision-currency-standard
• precision-currency-cash

To round to the nearest nickel, for example, use the skeleton precision-increment/0.05. For more information on the decimal number syntax, see Scale.

precision-increment/dddd implies minimum and maximum fraction digits equal to the precision of the decimal. There is not currently a way in skeleton syntax to specify an arbitrary min/max fraction digits when a rounding increment is used.

All rounding occurs after the number is scaled, such as in scientific notation or the scale skeleton.

Fraction Precision

The following are examples of fraction-precision stems:

More precisely, the fraction precision stem starts with ., then contains zero or more 0 symbols, which implies the minimum fraction digits. Then it contains either a *, for unlimited maximum fraction digits, or zero or more # symbols, which implies the minimum fraction digits when added to the 0 symbols.

Note that the stem . is considered valid and is equivalent to precision-integer.

Fraction-precision stems accept a single optional option: a number of significant digits. The options here correspond to the API functions on FractionPrecision. Some options require specifying r or s for relaxed mode or strict mode. For more information, see the API docs for UNumberRoundingPriority.

Precisely, the option follows the syntax of the significant digits stem (see below), but one of the following must be true:

• Option has one or more @s followed by the wildcard character (withMinDigits)
• Option has exactly one @ followed by zero or more #s (withMaxDigits)
• Option has one or more @s followed by zero or more #s and ends in s or r (withSignificantDigits)

Significant Digits Precision

The following are examples of stems for significant figures:

The precise syntax is very similar to fraction precision. The blueprint stem starts with one or more @ symbols, which implies the minimum significant digits. Then it contains either a *, for unlimited maximum significant digits, or zero or more # symbols, which implies the minimum significant digits when added to the @ symbols.

Trailing Zero Display

Starting with ICU 69, a new option called trailingZeroDisplay was added. To enable this in an ICU number skeleton, append /w to any precision token:

Wildcard Character

Prior to ICU 67, the symbol + was used for unlimited precision, instead of * (for example, .00+). For backwards compatibility, either + or * is accepted. This applies for both fraction digits and significant digits.

Rounding Mode

The rounding mode can be specified by the following stems:

• rounding-mode-ceiling
• rounding-mode-floor
• rounding-mode-down
• rounding-mode-up
• rounding-mode-half-even
• rounding-mode-half-down
• rounding-mode-half-up
• rounding-mode-unnecessary

For more details, see Rounding Modes.

Integer Width

The following examples show how to specify integer width (minimum or maximum integer digits):

The long-form option starts with either a single * symbol, signaling no limit on the number of integer digits (no truncateAt), or zero or more # symbols. It should then be followed by zero or more 0 symbols, indicating the minimum integer digits (the argument to zeroFillTo). If there is no * symbol, the maximum integer digits (the argument to truncateAt) is the number of # symbols plus the number of 0 symbols.

The concise skeleton is simply one or more 0 characters. This supports minimum integer digits but not maximum integer digits.

The special stem integer-width-trunc covers the case when both truncateAt and zeroFillTo are zero.

Prior to ICU 67, use the symbol + instead of *.

Scale

To specify the scale, use the following stem and option:

• scale/dddd

where dddd is a decimal number. For example, the following are valid skeletons:

• scale/100 (multiply by 100)
• scale/1E2 (same as above)
• scale/0.5 (multiply by 0.5)

The decimal number should conform to a standard decimal number syntax. In C++, it is parsed using the decimal number library described in LocalizedNumberFormatter::formatDecimal. In Java, it is parsed using BigDecimal. For maximum compatibility, it is highly recommended that your decimal number is able to be parsed by both engines.

Grouping

The grouping strategy can be specified by the following stems:

• group-off or ,_ (concise)
• group-min2 or ,? (concise)
• group-auto (or omit since this is the default)
• group-on-aligned or ,! (concise)
• group-thousands (no concise equivalent)

For more details, see UNumberGroupingStrategy.

Symbols

The following stems are allowed for specifying the number symbols:

• latin (use Latin-script digits)
• numbering-system/nnnn (use the nnnn numbering system)

A custom NDecimalFormatSymbols instance is not supported at this time.

Sign Display

The following stems specify sign display:

• sign-auto (or omit since this is the default)
• sign-always or +! (concise)
• sign-never or +_ (concise)
• sign-accounting or () (concise)
• sign-accounting-always or ()! (concise)
• sign-except-zero or +? (concise)
• sign-accounting-except-zero or ()? (concise)
• sign-negative or +- (concise)
• sign-accounting-negative or ()- (concise)

For more details, see UNumberSignDisplay.

Decimal Separator Display

The following stems specify decimal separator display:

• decimal-auto
• decimal-always

For more details, see UNumberDecimalSeparatorDisplay.