L20n

Axel Hecht

Mozilla Corporation

Overview

• Goals
• Design decisions
• File format
• Getting a string
• Examples
• Diskussion

First, the name …

• Localization 2.0
• Code name for architecture
• Maybe going to get renamed
• … but what are the 20 chars?
  • L12345678901234567890n

Prerequisites

• targeted at Mozilla
• applicable elsewhere
• independent of platform/language
  • installed locally or web-based
  • JavaScript, C++, python, …
  • Language bindings map l20n concepts to programming environment
• attractive for coders and localizers

Goals

• provide plain localized strings, of course
• Toolkit friendly
• (named) parameter or variable substitution
• Plurals
• Declensions
• VCS-based workflow
• Complex only where required
• Fallback from one locale to another

Design decisions

• localized string lookup
  • English strings are no keys
  • lookup by semantic symbolic names
• Toolkit friendly
  • Rich composited strings
• Plurals
  • Language-dependent plural forms
  • More than one plural per language
  • Arrays as values
  • Real localizer-defined macros

Design decisions, cont.

• Declensions
  • values depend on gender, case
  • … or attributes
  • strings are really objects
  • Hashes as values
• parameter substitution
  • complex scenarios
  • independent of caller language!
  • → handle parameter substitution inside the library
• Locale fallback
  • flexible context setup

VCS-based workflow

• files only contain localized information
  • yes, mono-lingual files
• VCSes much better at tracking process
• support real changesets, blame, and much more
• mail is yesterday
• … and Mozilla needs it

File format

• Suitable for humans and software editors
• Simple grammar
• Well-specified editing behaviour (to be done)
  • Don't break vi
• simlified, Entities with JSON values
• Localizable Object Lists, with comments

Example

/**
 * Doxygen style comments
 * @rules yet to be determined
 */
<simple: "string values">
<arrays: ["with", "more", "values"]>
<mouse[plural(n)]: ["one mouse", "${n}i mice"]>
<tree: "Baum"
  gender: "male"
  Genitiv: "Baums">
<sentence[tree.gender]:
  {male: "Des ${tree.Genetiv} bunte Kleider",
   female: "Der ${tree.Genitiv} bunte Kleider"}>
        

Getting a string

• Start with an expression, mouse
• Create a context chain
  • Several lol files
  • locale fallback implented through context chain
  • programmatic key-value-mappings (n → 3)
• evaluate expression in top context
• convert result to string

Evaluate mouse

• resolve mouse entity

  ["one mouse", "${n}i mice"]

• for each index item
  • Evaluate expression, here plural(n), in the top context

    1

  • resolve the index item

    "${n}i mice"

• evaluate parameter expressions, here ${n}i, and replace them

  3 mice

Complexity

• parameter substitution inside the library and locale allows different parameters depending on locale
• calling application just provides context in a language independent way
• simple languages have simple expressions
• complex languages have complex expressions
• no complexity burdened on software developer

Examples

• Application and Toolkit
• Accesskeys and Labels
• Locale fallback
• demo
• Declensions

Diskussion