[SPARK-29914] ML models append metadata in `transform`/`transformSchema`

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 3.0.0
• Fix Version/s: 3.1.0
• Component/s: ML
• Assignee: Ruifeng Zheng
• Reporter: Ruifeng Zheng
• Created: 15/Nov/19 10:54
• Updated: 25/Dec/19 01:40
• Resolved: 04/Dec/19 08:40

Description

There are many impls (like `Binarizer`/`Bucketizer`/`VectorAssembler`/`OneHotEncoder`/`FeatureHasher`/`HashingTF`/`VectorSlicer`/...) in `.ml` that append appropriate metadata in `transform`/`transformSchema` method.

However there are also many impls return no metadata in transformation, even some metadata like `vector.size`/`numAttrs`/`attrs` can be ealily inferred.

Issue Links

• links to: GitHub Pull Request #26547
• links to: GitHub Pull Request #26767
• links to: GitHub Pull Request #26999