[WIP] DO NOT MERGE refactor(RemoveAny): remove any types #773
Conversation
Removing Any* types to make the code maintainable in the long run.
github-actions
bot
added
test:uipath-langchain
|
|
||
| def _is_coded_evaluator(self, evaluators: List[AnyEvaluator]) -> bool: | ||
| def _is_coded_evaluator( | ||
| self, evaluators: List[BaseEvaluator[Any, Any, Any]] |
There was a problem hiding this comment.
this method will always categorize evaluators as coded. might break reporting
CC @Chibionos
| ] | ||
|
|
||
| AnyEvaluationItem = Union[EvaluationItem, LegacyEvaluationItem] | ||
| AnyEvaluator = Union[LegacyBaseEvaluator[Any], BaseEvaluator[Any, Any, Any]] |
There was a problem hiding this comment.
Continuing the conversation #685 (comment) here - AnyEvaluator.
What is not clear to me is:
If a new class wants to be backward incompatible (intentionally) then extending NewClass from OldClass would be wrong, right ?
Are we saying that no matter what we would never have backward incompatibility ?
Thoughts @radu-mocanu @cristipufu @akshaylive @Chibionos ?
There was a problem hiding this comment.
I think there are two concerns here:
- Backwards compatibility of the programmatic interfaces
- Backwards compatibility of the schema to represent evaluation sets
We don't care about ##1 right now because of no usage. However, ##2 is needed.
In the programmatic interfaces, we should avoid having to represent Any* = A | B because the entire internals of the library needs to know about both the versions. Instead, we need to transform the legacy schema to the new schema during loading.
We had a conversation regarding this and the conclusion was that the Legacy* will be removed in a few weeks.
Also updated the title of the PR to prevent merge.
akshaylive
changed the title
3 participants
Removing Any* types to make the code maintainable in the long run.