SDK TimeAverage/Total count pre-first-sample time when the first sample is Uncertain (late-start interval) #821

New Issue

2026-11-11T00:05:57-05:00

dohertj2 commented

2026-11-11 00:05:57 -05:00

Found during #810. When data starts inside an interval and the first sample is Uncertain, the SDK's TimeAverage/Total also count the time before the first sample (as if ramping from 0). Example QP over [0,20): SDK Total 188.96 vs Part 13 163.96. Our SQL backends give the Part 13 answer; the conformance suite pins it by hand because the oracle wraps the SDK calculators.

Pin the divergence in SdkAggregateParityTests with an Uncertain-first late-start dataset.
Check the frozen live AggregateFilter calculators (src/Server/ZB.MOM.WW.OtOpcUa.OpcUaServer/Compatibility/Aggregates/) and decide together with #809.
Add the reverse-order variant of the late-start case for QuestDB (only forward is covered).

Found during #810. When data starts inside an interval and the first sample is Uncertain, the SDK's TimeAverage/Total also count the time before the first sample (as if ramping from 0). Example QP over [0,20): SDK Total 188.96 vs Part 13 163.96. Our SQL backends give the Part 13 answer; the conformance suite pins it by hand because the oracle wraps the SDK calculators. - Pin the divergence in `SdkAggregateParityTests` with an Uncertain-first late-start dataset. - Check the frozen live AggregateFilter calculators (`src/Server/ZB.MOM.WW.OtOpcUa.OpcUaServer/Compatibility/Aggregates/`) and decide together with #809. - Add the reverse-order variant of the late-start case for QuestDB (only forward is covered).

dohertj2 added the type:followup priority:p3 labels 2026-11-11 00:05:57 -05:00

dohertj2 referenced this issue

2026-11-11 00:49:03 -05:00

Historian conformance suites have no Uncertain samples for PercentGood/DurationGood #810

dohertj2 referenced this issue

2026-11-11 00:49:10 -05:00

Live AggregateFilter PercentGood/DurationGood count Uncertain as Good, disagreeing with Part 13 and HistoryRead #809

dohertj2 commented

2026-11-11 03:31:16 -05:00

Measured during the #815–#833 batch (#820/#821/#822 work). Dataset: interval QP [0,20), first sample Uncertain at +5 s.

Calculator	Total	TimeAverage	DurationGood	PercentGood
Part 13 / our HistoryRead	163.96	10.93	8000	40
Live frozen AggregateFilter (SDK 1.5.378)	188.96	9.45	15000	75
SDK 2.0.1 stock	188.96	9.45	8000	40

So stock SDK 2.0.1 would fix #809 (DurationGood/PercentGood counting Uncertain as Good) but not this issue: TimeAverage/Total still count the pre-first-sample time. Keeping open for the decision (override the live calculator, or accept and document the divergence).

Measured during the #815–#833 batch (#820/#821/#822 work). Dataset: interval QP [0,20), first sample Uncertain at +5 s. | Calculator | Total | TimeAverage | DurationGood | PercentGood | |---|---|---|---|---| | Part 13 / our HistoryRead | 163.96 | 10.93 | 8000 | 40 | | Live frozen AggregateFilter (SDK 1.5.378) | 188.96 | 9.45 | 15000 | 75 | | SDK 2.0.1 stock | 188.96 | 9.45 | 8000 | 40 | So stock SDK 2.0.1 would fix #809 (DurationGood/PercentGood counting Uncertain as Good) but **not** this issue: TimeAverage/Total still count the pre-first-sample time. Keeping open for the decision (override the live calculator, or accept and document the divergence).

dohertj2 referenced this issue from a commit

2026-11-11 05:24:23 -05:00

feat(historian): serve DurationGood/DurationBad on the SQL backends; pin the #821 late-start facts (#820, #821)

dohertj2 referenced this issue from a commit

2026-11-11 05:24:23 -05:00

docs(historian)+infra: shared QuestDB fixture, correct SQL test credential, DurationGood/Bad + #821 docs (#822, #820, #821)

dohertj2 commented

2026-11-11 11:28:51 -05:00

All three items are on master (e7b485c0, docs 1a4d46a7) and re-verified 2026-11-11 against the live fixtures (Sql.Common 283, QuestDb.Integration 249, SqlServer.Integration 211, LiveAggregateLateStartFactsTests 1 — all green): the SDK divergence is pinned in SdkAggregateParityTests (Uncertain-first late start, Good-first control), the live AggregateFilter calculators are pinned by LiveAggregateLateStartFactsTests, and the reverse-order late-start case is in the shared QualityMixConformance (QuestDB + SqlServer). The remaining decision (whether live calculators should change) is #809 — findings copied there.

All three items are on master (`e7b485c0`, docs `1a4d46a7`) and re-verified 2026-11-11 against the live fixtures (Sql.Common 283, QuestDb.Integration 249, SqlServer.Integration 211, `LiveAggregateLateStartFactsTests` 1 — all green): the SDK divergence is pinned in `SdkAggregateParityTests` (Uncertain-first late start, Good-first control), the live AggregateFilter calculators are pinned by `LiveAggregateLateStartFactsTests`, and the reverse-order late-start case is in the shared `QualityMixConformance` (QuestDB + SqlServer). The remaining decision (whether live calculators should change) is #809 — findings copied there.

dohertj2 closed this issue

2026-11-11 11:28:51 -05:00

dohertj2 referenced this issue

2026-11-11 11:28:52 -05:00

Live AggregateFilter PercentGood/DurationGood count Uncertain as Good, disagreeing with Part 13 and HistoryRead #809

dohertj2 referenced this issue from a commit

2026-11-11 14:53:22 -05:00

fix(aggregates): live TimeAverage/Total leave the span before an Uncertain first value out (#821)

dohertj2 referenced this issue from a commit

2026-11-11 14:53:22 -05:00

Merge #821 live TimeAverage/Total late-start with Uncertain first sample matches HistoryRead (Part 13 partial interval)

Sign in to join this conversation.