Links

• Formatting Messages | ICU User Guide
• Pattern_Syntax and Pattern_Whitespace
  • Unicode pattern syntax. argNameOrNumber in MessageFormat (the first argument) is composed of either digits or anything that is NOT from Pattern_Syntax and Pattern_Whitespace (so excludes all punctuation, spaces, etc.)
• JS Lib: SlexAxton/messageformat.js

Patterns and Their Interpretation

Ref: icu-project.org/…/MessageFormat.html, Pattern_Syntax

MessageFormat uses patterns of the following form:

message = messageText (argument messageText)*
argument = noneArg | simpleArg | pluralArg | selectArg | selectordinalArg

noneArg          =  '{'  argNameOrNumber  '}'
simpleArg        =  '{'  argNameOrNumber  ','  argType            [',' argStyle]  '}'
pluralArg        =  '{'  argNameOrNumber  ','  "plural"        ','   pluralStyle  '}'
selectArg        =  '{'  argNameOrNumber  ','  "select"        ','   selectStyle  '}'
selectordinalArg =  '{'  argNameOrNumber  ','  "selectordinal" ','   pluralStyle  '}'

argNameOrNumber  =  argName | argNumber
argName          =  [^[[:Pattern_Syntax:][:Pattern_White_Space:]]]+
argNumber        =  '0' | ('1'..'9' ('0'..'9')*)

argType          =  "number" | "date" | "time" | "spellout" | "ordinal" | "duration"
argStyle         =  "short" | "medium" | "long" | "full" | "integer" | "currency" | "percent" | argStyleText

pluralStyle      =  [offsetValue] ( explicitValue|pluralKeyword '{' message '}')+  // the "other" pluralKeyword is required.
offsetValue      =  "offset:" number
explicitValue    =  '=' number  // adjacent, no white space in between
pluralKeyword    =  'zero' | 'one' | 'two' | 'few' | 'many' | 'other' | keyword

selectStyle      =  (selectKeyword '{' message '}')+  // the "other" selectKeyword is required.
selectKeyword    =  'other' | keyword

keyword          =  [^[[:Pattern_Syntax:][:Pattern_White_Space:]]]+

Pattern_White_Space between syntax elements is ignored, except:

• between the {curly braces} and their sub-message
• between the '=' and the number in explicitValue (i.e. there must be no space between them.  e.g. "=1")

Plurals

Predefined keyword's:  'zero', 'one', 'two', 'few', 'many' and 'other'.

You must always define message text for the other case (it's the fallback.)

Matching Priority / Algorithm

• Exact Matches
  • Match the input number against the explicitValue clauses.  If found, use that messageText and return.
• Keyword Matches
  • set keyword = PluralRules(input_number - offset) (offset defaults to 0)
  • Use clause corresponding to this keyword and return (if found).
• Fallback
  • Use the messageText corresponding to the other clause.

Formatting

• Format number-minus-offset using a NumberFormat for the PluralFormat's locale.
  • If you need special number formatting, you have to use a MessageFormat and explicitly specify a NumberFormat argument. (Note that argument is formatting without subtracting the offset! If you need a custom format and have a non-zero offset, then you need to pass the number-minus-offset value as a separate parameter.)
• Replace an unquoted pound sign (#) in the selected sub-message by the formatted number-minus-offset value from the previous step.

Gender and "select"

The main use case for the select format (selectArg) is gender based inflection.

When names or nouns are inserted into sentences, their gender can affect pronouns, verb forms, articles, and adjectives. Special care needs to be taken for the case where the gender cannot be determined. The impact varies between languages:

• English has three genders, and unknown gender is handled as a special case. Names use the gender of the named person (if known), nouns referring to people use natural gender, and inanimate objects are usually neutral. The gender only affects pronouns: "he", "she", "it", "they".
• German differs from English in that the gender of nouns is rather arbitrary, even for nouns referring to people ("Mädchen", girl, is neutral). The gender affects pronouns ("er", "sie", "es"), articles ("der", "die", "das"), and adjective forms ("guter Mann", "gute Frau", "gutes Mädchen").
• French has only two genders; as in German the gender of nouns is rather arbitrary - for sun and moon, the genders are the opposite of those in German. The gender affects pronouns ("il", "elle"), articles ("le", "la"), adjective forms ("bon", "bonne"), and sometimes verb forms ("allé", "allée").
• Polish distinguishes five genders (or noun classes), human masculine, animate non-human masculine, inanimate masculine, feminine, and neuter.
• Noun clauses:  Some other languages have noun classes that are not related to gender, but similar in grammatical use. Some African languages have around 20 noun classes.

The fallback keyword is "other" (just like with pluralization.)  Some common keywords are: "male", "female", "mixed" (for groups of people) and "Unknown".

Quoting

• messageText can contain quoted literal strings including syntax characters.
  • A quoted literal string begins with an ASCII apostrophe and a syntax character (usually a curly brace/{}) and continues until the next single apostrophe.
• A double ASCII apostrohpe inside or outside of a quoted string represents one literal apostrophe.
• Quotable syntax characters are the curly braces ("{", "}") in all messageText parts, plus the "#" sign in a messageText immediately inside a pluralStyle.
• See also MessagePattern.ApostropheMode
• In argStyleText, every single ASCII apostrophe begins and ends quoted literal text, and unquoted {curly braces} must occur in matched pairs.

Recommendation: Use the real apostrophe character, «’» (U+2019), for human-readable text, and use the ASCII apostrophe, «'» (U+0027), only in program syntax, like quoting in MessageFormat. See the annotations for U+0027 Apostrophe in The Unicode Standard.