The render function called by ScriptTemplateView has currently 3 parameters provided:
String template: the content of the template resource
Map<String, Object>: the model to use to render the view
String url: the url of the 2 view
To achieve i18n support for messages and nested template loading, we need to provide these additional informations:
a ResourceBundleMessageSource instance (or the ApplicationContext that allows to retrieve it)
the view Locale
a Function<String, String> that allows the render function to call ScriptTemplateView#getTemplate(String)
I see mainly 2 ways to support that:
We could be possible leverage setExposeContextBeansAsAttributes() or setExposedContextBeanNames() to access to context beans and expose them via model attributes.
We could transform the 3rd parameter passed to the script function (currently String url) to RenderingContext that would contains String url, Locale locale, ResourceBundleMessageSource messageSource and Function<String, String> templateLoader properties. This would be a breaking change for people using url but ScriptTemplateView is a rather feature, and url is not widely used, so I consider this as an option in order to be consistent and provide such flexible mechanism for further needs + it provides these properties in a type-safe way which would be valuable for Kotlin JSR-223 support.
Not sure if the following stuff is appropriate here but I didn't want to open a new issue without discussion. So, the situation is following:
Is there a way to use two scripting engines at the same time or this feature is not supported currently?
This change would be breaking (on the script side, not on java side) for people using the third url parameter that we introduced as part of #18033 in 4.2.2, but ScriptTemplateView is a rather young functionality, and this variant with an url parameter is likely to not being used very widely, so I would take the opportunity of our major 5.0 release to define a consistent and future proof render function SPI.
As you can see here the migration path is trivial.
I preferred this option over using setExposeContextBeansAsAttributes() or setExposedContextBeanNames() because:
It allows to be consistent about the various parameters we pass to the rendering function (url versus others)
It is type-safe which matters for Kotlin Script and more discoverable than something based on model attributes
It avoids to mix model attributes designed to be rendered in the template, and those informations designed to be used in the rendering function
It will allow us to add new rendering properties without breaking the script API
I also choose to provide the ApplicationContext rather than directly the ResourceBundleMessageSource bean because that avoids to make assumption about how to retrieve it (by type, by name) and allows a wide range of use cases in addition to i18n. See it in action in this Kotlin Script example.