Brief: internationalization

Localizing strings and captions

Localization is reached for when one logic runs in several languages. A user-visible string — the caption of a class, a property, an action, a form, the text of a message — is localized by a string data identifier in curly braces inside a string literal. When the value is sent to the client, the platform looks each identifier up in the project files whose name ends with ResourceBundle, in the required locale, and substitutes the translation found; if there is no translation, the identifier itself is left without the braces.

CLASS Book '{use.case.i18n.book}';
name '{use.case.i18n.book.name}' = DATA STRING[40] (Book);

The current locale — language, country, timezone — is taken from the Authentication.language[CustomUser] property and the like, and for actions started by the system, from the server locale; user data is not translated. The mechanism is described in internationalization.

Analogy: ResourceBundle keys written straight into the caption text instead of a call to a translation function.

Reverse translation

Reverse translation removes the need to place identifiers by hand: captions are written in the code as plain text in one language, and it is the platform that finds the entry whose value is that text — the entry keeps its own identifier as the key, and the text is its value. It is turned on by the logics.lsfStrLiteralsLanguage launch parameter, which sets the language of the string literals in lsf code: at server start a value -> identifier dictionary is built from all ResourceBundle files of the project in that locale, and a plain literal matching an entry value is replaced at code parse time with that entry's identifier and then behaves as a localizable one. Leading and trailing spaces are excluded from the match and kept around the substitution, and a literal that is empty or all spaces is not replaced at all.

The replacement applies to any plain literal, not only to captions, so technical strings — JSON keys, addresses, formats, external identifiers — are written in the raw form of a string literal, which takes part neither in localization nor in reverse translation:

'content'    // plain: localized, and replaced by reverse translation on a match
r'rawContent'  // raw: neither