Skip to content

Improve append-only inference for input connectors #149

@zxqfd555

Description

@zxqfd555

Is your feature request related to a problem? Please describe.
Input connectors often infer append-only properties of the collections they read from, but these properties are not always inferred correctly, even though this could (and should) be improved.

In particular:

  • Collections read from Kafka are almost always append-only (except possible UpsertSession exception, not even exposed to the users).
  • The Python connector should also be append-only by default.

Marking these collections as append-only would allow the runtime engine to apply more aggressive memory optimizations. When a collection is append-only, significantly more efficient memory management strategies are available.

Describe the solution you'd like

  • The Python connector append-only by default.
  • The Kafka connector append-only by default.

Describe alternatives you've considered
N/A.

Additional context
This change should be fully backward-compatible.
In practice, the collections produced by these connectors are already append-only; the issue is that this property is currently not explicitly set.

Metadata

Metadata

Assignees

Labels

enhancementNew feature or request

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions