planner/optimizer — `candidate_gen.rs::enumerate_candidates` currently falls back to EXACT-only whenever an AQE needs more than one statistic (`aqe.requirements.statistics.len() != 1`), e.g. `avg` which decomposes to `[Sum, Count]`. This is marked with a `// ponytail:` comment as "not supported in v1."
Need to decide: should multi-statistic AQEs be supported by pairing two sketches (one per statistic), and if so, what does candidate generation / cost modeling / query method assignment look like for a pair instead of a single sketch? If staying EXACT-only for now is fine, this issue tracks that as a known gap rather than a silent limitation.
Found during code review of PR #407.
planner/optimizer — `candidate_gen.rs::enumerate_candidates` currently falls back to EXACT-only whenever an AQE needs more than one statistic (`aqe.requirements.statistics.len() != 1`), e.g. `avg` which decomposes to `[Sum, Count]`. This is marked with a `// ponytail:` comment as "not supported in v1."
Need to decide: should multi-statistic AQEs be supported by pairing two sketches (one per statistic), and if so, what does candidate generation / cost modeling / query method assignment look like for a pair instead of a single sketch? If staying EXACT-only for now is fine, this issue tracks that as a known gap rather than a silent limitation.
Found during code review of PR #407.