On 21.12.2017 05:02, Nathan Harvey wrote:
That's true, "use" does offer control over the scope. Why is it more difficult to implement it in a dynamic context versus a static one?
"use" activates categories, but to activate them, you need to know them. In dynamic Groovy that is the runtime, in static Groovy it is the compiler. That means the static compiler can already only compile the use-constructs as lexical categories right now. The dynamic version uses a thread local scope to "see" categories during runtime. For the dynamic version to be like the lexical version you would have to know the calling context. And even though thread-local is such a context, it is too wide scoped for this. I am not saying we cannot implement that. I am saying that the current infrastructure we have in place does not provide us with that. That means there will be most likely bigger changes to the callsite caching and here the old callsite caching code is actually in the way. Again... all doable... but this will be a change to a dozens of files or so in a difficult area, just to provide the additional information on the callsite.
bye Jochen
