Several directives (t-key, t-log, t-translation, …) are represented by
wrapper ASTs that contain another AST or null.
Because these wrappers don't share the same type as their children,
various AST type checks were broken.
This commit addresses those issues by:
- Commuting Translation / TranslationContext and Multi ASTs so that
Multi children are spread as expected (see parseChildren).
- Parsing the t-key directive before t-esc / t-out (in line with https://github.com/odoo/owl/pull/1685).
- Ensuring wrappers around TSet ASTs are recognized as having no
representation, so compileMulti properly discards children equivalent
to TSet.
A new directive t-tanslation-context and a new family of directives (of the form
t-translation-context-...) are introduced to allow to translate terms in contexts.
Set t-translation-context="fr" on a node makes every call of the translation
function to be done with "fr" as seconde parameter when translating an
attribute/content within that node or its children (if no closer
t-translation-context directive is found).
A directive t-translation-context-attr="fr" can be used on a node to target its
attribute "attr". For example, if a div has an attribute title="a title",
use t-translation-context-title="pt" will make "a title" to be translated
in the context "pt". Note that this takes precedence over any other directive
t-translation-context found on a parent (or the div itself).
The translation function is in charge of the interpretation of the context: OWL
does not associate any meaning with a translation context.
Before this commit, the translate function was not applied to text
content inside the body of a t-set (unless that content was itself
inside some html). This is not clearly not intended.
To fix it, we just need to call the translate function at the
appropriate moment.
closes#1434
Before this commit, the parser would remove all consecutive white spaces
for text nodes. After that, the code generator would call the translate
function with the resulting string, which then is different than what we
would expect.
With this commit, we make sure we apply the translation before removing
the additional whitespace. To do that, we have to move the code
processing the string from the parser into the code generator, which
actually makes sense, as the parser should only collect all useful
information without applying too much logic.
closes#1351
Mathieu Duckerts-Antoine