$state.reload instead of broadcastUpdate?

[vbraun]

Background: I'm having some interop issues with angular-material, e.g. #214 or angular/material#1742. In both cases the initial render is ok, but subsequent changes of the language via setCurrentLanguage(...) don't produce the desired effect.

My current workaround is to just flip the language variable and reload the current state:

    gettextCatalog.currentLanguage = ...
    $state.reload()

I realize that this uses ui-router and loses content of <input> fields, on the other hand it is more robust since you only have to be able to render once. Another advantage is that you can use one-time bindings {{::"Text"|translate}}. Is there any other issue that I haven't thought about? Would there be interest to "bless" this workaround by adding a special setter setCurrentLanguageAndReload, say?

[rubenv]

I'm not sure why changes to angular-gettext are needed for this?

Can't you just add this to your application:

$rootScope.$on("gettextLanguageChanged", function () {
    $state.reload();
});

?

[rubenv]

Also note that there isn't currently a dependency on ui-router in angular-gettext (and we won't add one, not every project needs ui-router).

[vbraun]

Good idea, I didn't think about overwriting the gettextLanguageChanged signal!

I'm still having some issues with

 <md-button translate>Text</md-button>

in angular-material. This is first hit by the angular-material compiler, which rewrites it to

 <md-button translate><span>Text</span></md-button>

and then angular-gettext can't extract the msgid. It should run first so it can access the untransformed "Text". And then it must not terminate so that angular-material can do its thing. I'm currently playing around with changing

     restrict: 'AE',
-    terminal: true,
-    terminal: false,
+    priority: 10000,
     compile: function compile(element, attrs) {

in the translate directive. Should this be configurable? I read #147 but I still don't know why it sets terminal=true right now.

[rubenv]

@vbraun That's a different discussion, which we should probably have in #147.

We could probably tweak both values, but I'm hesitant to do so because I'm not entirely sure what the effects might be (and I don't want to break existing code out there). If we can prove that it's safe: sure.

[rubenv]

As for the state switching, that seems solved, so I'll be closing this issue (feel free to reopen if needed).