QueryWorkload streaming fields (RepeatingEntry, DataCharacteristics) are unused — should be wired in

[milindsrivastava1997]

Summary

crates/ir/src/workload.rs defines QueryWorkload.repeating_queries: Option<Vec<RepeatingEntry>> and .data_characteristics: Option<DataCharacteristics> (backed by RepeatingEntry, DataCharacteristics, DataDistribution, RepetitionInterval) for streaming/repeating workloads and sketch-sizing inputs.

Neither field has a producer or a consumer anywhere in the workspace:

• No frontend or batch-lowering path ever populates repeating_queries — lower_promql_batch/lower_sql_batch only read query_batch.
• data_characteristics's own doc comment claims it's "used by the cost model and sketch-parameter binder to size sketches" — plan::cost_model::CostModel::rank_candidates/size_params take no such input; there is no live call site anywhere.
• The only place either field is even referenced is one PromQL test file, which always sets both to None.

Ask

Either wire these into a real streaming-workload path and the cost model's sizing logic (per the doc comment's stated intent), or, if streaming support isn't near-term, consider removing them until there's a concrete consumer — following the same empirical-verification pattern as #181/#182 for QueryExpr::Ref/LetBinding/Window.

[zzylol]

Related open work

• Recurrence-aware optimization: cost shared maintenance by query repetition #287 is the concrete recurrence-aware consumer for RepeatingEntry and relevant DataCharacteristics.
• Closing this issue requires metadata to reach optimization/costing, not only a lowering helper.