Omry Yadan
Omry Yadan
@Jasha10, good to merge?
@Jasha10, this is a breaking change (that would also require a parallel fix to Hydra. 1. News fragment. 2. I think it's really best if the effort is focused on...
I think the best solution would be to "vendor" OmegaConf with a specific version of antlr, and have Hydra depend on it. That would require some rewriting some imports in...
> Could an option be to replace the `anttrl4` java binary by [antlr-tools](https://github.com/antlr/antlr4-tools)? I don't see how that would do anything. The antlr java binary is just a matter of...
> I see, so the binary dependency is not runtime, but build-time only. Not quite. The antlr runtime is checking that it's working with code generated by a generator from...
I think this needs tests to show what it's actually doing and to ensure it does not break in the future. Also, worth considering any performance impact.
We are actually considering introducing a headers mechanism to files similar to `# @package abc` in Hydra (#anchor Such a mechanism could be used to support allow_objects or other OmegaConf...
I will let @Jasha10 decide on accepting this. I am not too thrilled because it will make the headers support more complicated and the mechanism there would in principle allow...
@bzczb, patch welcome (along with a unit test).
This would break all code that depends on the returned type from to_container being dict. Isn't there a standard way to recursively convert a dict to SimpleNamespace?