Memory and glossaries

The memory

The bundle keeps one translation memory for the whole application, in the table phil_translation_memory: one unit per pair of languages and source fingerprint, the latest translation winning. A regional variant reads the units of its parent after its own. Every target that becomes usable — translated, reviewed, confirmed, or imported as such — is learnt, its origin the document. The memory then proposes it:

• to the new segments of any document, as suggested, when the source is the same, white space aside (prefill);
• on the screen, for any segment not yet usable, exact or close — down to min_score, at most max_suggestions.

With a policy, both proposals keep only the units learnt from a document the current user may translate (canTranslate() on the unit's origin and the target language), and those learnt from no document — filled by the application with learn() without an origin: one document's translation never shows on the screen of a user who may not see that document. An application that gives other origins to the units it learns answers for them in its policy too.

Phil\TranslationMemory\Memory\TranslationMemory is a public service: an application can look up the memory itself, or fill it from an existing translation memory with learn(), or with its store, the alias MemoryStore.

Glossaries

Phil\TranslationBundle\Application\Glossaries keeps one glossary per pair of languages, in the table phil_translation_term:

$glossaries->put(LanguageTag::of('en'), LanguageTag::of('fr'), new Term('hit points', ['points de vie', 'PV'], forbidden: ['vie'], note: 'Always plural.'));
$glossaries->put(LanguageTag::of('en'), LanguageTag::of('fr'), new Term('Dungeon Master'));   // kept as it is
$glossaries->remove(LanguageTag::of('en'), LanguageTag::of('fr'), 'hit points');

put() adds a term, or gives a term the glossary has its new translations, case, note and forms; remove() throws NotFound for a term it does not have; export() and import() exchange the glossary as TBX. A regional variant has glossaries of its own — en to fr-CA —, which the glossary pages edit like any other (Glossary pages). terms() lists the rows of a pair of languages in alphabetical order, and glossary() gives the library's Glossary, which the screen uses to show each segment's terms and its target's issues. The glossary documentation of phil/translation-memory says how terms match.