You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
feat(promql): preserve and plan topk over sum_over_time/count_over_time (ASAPQuery #706 parity) #342
Add semantic planning parity with ProjectASAP/ASAPQuery#706 for Top-K over sum_over_time and count_over_time.
Reproduction
Both queries lower successfully through frontend-promql, but the normal summary-selection path used by ASAPQuery-backend currently produces no legal summary candidate for at least the value-ranked case:
topk(5, sum_over_time(m[1m]))
topk by (job) (5, count_over_time(m[1m]))
ASAPQuery #706 defines an important semantic distinction:
topk(sum_over_time(...)) ranks each series by the sum of sample values (value-weighted heap);
topk(count_over_time(...)) ranks each series by the number of samples (count-weighted heap).
The current AggIntent::TopK / SketchQuery::TopK representation carries k but does not carry the ranking/weighting basis, so downstream physical compilation cannot preserve this distinction without re-parsing PromQL.
Scope
Represent the Top-K ranking basis in canonical pre-ASAP and post-ASAP IR using language-neutral semantics.
Activity