Issues

Range queries: evaluate `@ end()`/`@ start()` subtrees once per request and share them

ProjectASAP/ASAPPlanner#1251 · opened

on Nov 8, 2026

Bounded scan windows, retention and Stage 3 pricing for subqueries, @ anchors and ranges over projected rows

ProjectASAP/ASAPPlanner#1249 · opened

on Nov 7, 2026

Executor: subqueries over expressions other than a selector or a per-series range function (step-grid evaluator)

ProjectASAP/ASAPPlanner#1248 · opened

on Nov 7, 2026

SQL INTERSECT / EXCEPT fail to lower (DataFusion semi / anti join)

ProjectASAP/ASAPPlanner#1245 · opened

on Nov 7, 2026

1

PromQL info(): carry the info metric as an input so the executor can join it

ProjectASAP/ASAPPlanner#1244 · opened

on Nov 7, 2026

1

Stage 3 selects no plan when a scalar expression reads an aggregate (ObjectiveDiffersFromPlanCost)

ProjectASAP/ASAPPlanner#1243 · opened

on Nov 7, 2026

1

absent_over_time is lowered as a per-series aggregate, so it has no exact plan

ProjectASAP/ASAPPlanner#1240 · opened

on Nov 7, 2026

1

ClickhouseSQL drops backslashes in string literals (ClickBench q29 regex lowers wrong)

ProjectASAP/ASAPPlanner#1231 · opened

on Nov 6, 2026

Planner: error models and parameter lists for Hydra (and other unsized families), so Pass 1 can offer them

ProjectASAP/ASAPPlanner#1228 · opened

on Nov 6, 2026

Executor: store Disk-medium materialized views and one-evaluation results on disk

ProjectASAP/ASAPPlanner#1217 · opened

on Nov 5, 2026

Ingest-time builds: drop the Repeated requirement; replace Predictability with an allow_ingest_time_compute flag

ProjectASAP/ASAPPlanner#1211 · opened

on Nov 4, 2026

Make units explicit in workload JSON keys and time-valued fields (fixed_interval → fixed_interval_ms, …)

ProjectASAP/ASAPPlanner#1210 · opened

on Nov 4, 2026

Stage 1: split Pass 1 and Pass 2 into separate crates

ProjectASAP/ASAPPlanner#1209 · opened

on Nov 4, 2026

[rename] Give better names to each "stage" and each "pass"

ProjectASAP/ASAPPlanner#1208 · opened

on Nov 4, 2026

1

Tracking: implement the optimization formulation (memo candidates, properties, edges y_{u,c,i}, MILP) as named code per stage

ProjectASAP/ASAPPlanner#1207 · opened

on Nov 4, 2026

1

[future work] Stage 1: window candidates with properties — tumbling window, sliding window with merge, sliding window with subtraction

ProjectASAP/ASAPPlanner#1206 · opened

on Nov 4, 2026

DAG viewer / stage_pipeline: print memo candidates with properties, every producer–consumer edge with its feasibility and rejection reason, and every stage’s pruned candidates

ProjectASAP/ASAPPlanner#1205 · opened

on Nov 4, 2026

Stage 3 MILP: implement the formulation variables x, y, z, t and rows (A)–(I) as named code over explicit edges

ProjectASAP/ASAPPlanner#1204 · opened

on Nov 4, 2026

Logical memo candidate as (one operator over groups, logical properties θ) with a requirement per input; no private nodes

ProjectASAP/ASAPPlanner#1203 · opened

on Nov 4, 2026

1

asap_planner JSON drops metric-type warnings before Stage 3

ProjectASAP/ASAPPlanner#1202 · opened

on Nov 4, 2026

Producer–consumer edges (u, c, i): make the edge set explicit after Pass 2 and Stage 2, with α^log and α^phys, as the domain of y_{u,c,i}

ProjectASAP/ASAPPlanner#1201 · opened

on Nov 3, 2026

Physical properties: replace Materialization with timing × lifetime (None / OneEvaluation / AcrossEvaluations) and a schedule for shareable results, with r^phys and α^phys

ProjectASAP/ASAPPlanner#1200 · opened

on Nov 3, 2026

1

Stage 3 MILP vs the memo formulation: one choice costs several binaries, sizes and targets are groups, and the group graph is cyclic

ProjectASAP/ASAPPlanner#1199 · opened

on Nov 3, 2026

Planning is about 20x slower since the full tumbling-window-width domain (#1185)

ProjectASAP/ASAPPlanner#1197 · opened

on Nov 3, 2026

Clean up tools left broken or unused by asap_planner (#1161)

ProjectASAP/ASAPPlanner#1186 · opened

on Nov 2, 2026

1

Navigation Menu

Repository navigation

Issues

Range queries: evaluate `@ end()`/`@ start()` subtrees once per request and share them

Bounded scan windows, retention and Stage 3 pricing for subqueries, @ anchors and ranges over projected rows

Executor: subqueries over expressions other than a selector or a per-series range function (step-grid evaluator)

SQL INTERSECT / EXCEPT fail to lower (DataFusion semi / anti join)

PromQL info(): carry the info metric as an input so the executor can join it

Stage 3 selects no plan when a scalar expression reads an aggregate (ObjectiveDiffersFromPlanCost)

absent_over_time is lowered as a per-series aggregate, so it has no exact plan

ClickhouseSQL drops backslashes in string literals (ClickBench q29 regex lowers wrong)

Planner: error models and parameter lists for Hydra (and other unsized families), so Pass 1 can offer them

Executor: store Disk-medium materialized views and one-evaluation results on disk

Ingest-time builds: drop the Repeated requirement; replace Predictability with an allow_ingest_time_compute flag

Make units explicit in workload JSON keys and time-valued fields (fixed_interval → fixed_interval_ms, …)

Stage 1: split Pass 1 and Pass 2 into separate crates

[rename] Give better names to each "stage" and each "pass"

Tracking: implement the optimization formulation (memo candidates, properties, edges y_{u,c,i}, MILP) as named code per stage

[future work] Stage 1: window candidates with properties — tumbling window, sliding window with merge, sliding window with subtraction

DAG viewer / stage_pipeline: print memo candidates with properties, every producer–consumer edge with its feasibility and rejection reason, and every stage’s pruned candidates

Stage 3 MILP: implement the formulation variables x, y, z, t and rows (A)–(I) as named code over explicit edges

Logical memo candidate as (one operator over groups, logical properties θ) with a requirement per input; no private nodes

asap_planner JSON drops metric-type warnings before Stage 3

Producer–consumer edges (u, c, i): make the edge set explicit after Pass 2 and Stage 2, with α^log and α^phys, as the domain of y_{u,c,i}

Physical properties: replace Materialization with timing × lifetime (None / OneEvaluation / AcrossEvaluations) and a schedule for shareable results, with r^phys and α^phys

Stage 3 MILP vs the memo formulation: one choice costs several binaries, sizes and targets are groups, and the group graph is cyclic

Planning is about 20x slower since the full tumbling-window-width domain (#1185)

Clean up tools left broken or unused by asap_planner (#1161)