Skip to content

Expose Count-Min totals for backend readouts (review) - #498

Closed
zzylol wants to merge 77 commits into
feat/histogram-quantilefrom
feat/cms-scaled-total
Closed

Expose Count-Min totals for backend readouts (review)#498
zzylol wants to merge 77 commits into
feat/histogram-quantilefrom
feat/cms-scaled-total

Conversation

@zzylol

@zzylol zzylol commented Sep 30, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Stacked on #493.

Supersedes #496, which carried the same change against integration/planner-for-backend; GitHub marked #496 merged when its head was incorporated into that branch, but it never reached main. Review this PR instead; #496 stays closed.

Why

Backend bare-count readouts need a sampling-aware Count-Min total without reproducing kernel arithmetic.

What

Expose CountMinSketchAccumulator::total() and the unsampled heap variant, using one row's mass scaled by the edge sampling probability. CountSketch has signed rows and intentionally has no total accessor.

Before this PR

A sketch containing ten sampled updates at p=0.25 required backend arithmetic to return the estimated total 40.

After this PR

The backend asks the Planner kernel for that total directly; collisions and merges preserve the mass.

Validation

Focused total tests pass (including evicted heap entries) for colliding keys, p=0.25, merged states, and an empty state.

🤖 Generated with Claude Code

zzylol and others added 30 commits September 30, 2026 18:02
@zzylol
feat: add shared summary kernels and typed physical values
4dc08e5 
Start `asap-physical-operators` with thin summary kernels over
`asap_sketchlib`, the kernel capability checks and the typed value model.

The boundary is:
- `asap_sketchlib` owns sketch algorithms and their state encodings.
- Kernels hold one population's in-memory state. They expose `merge`, a
  typed sketch readout (`estimate(&SketchQuery)`) and memory accounting.
  Exact states answer a typed `ExactReadout`; empty MIN/MAX read as `None`.
- Group-by belongs to physical operators.
- Deployments own wire decoding, delta frames, edge sampling and storage
  statistics.

So wire decoding, `SerializableToSink`, `AggregationType`, `aux_stats`,
`reset_to_empty` and the keyed/sum/min/max kernels are not carried over from
ASAPQuery-backend. The `asap_sketch_codec` crate is not carried over either;
it moves to `asap_sketchlib`. Hydra KLL remains as the Hydra shared-grouping
kernel. HLL uses sketchlib's classic estimator.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Rebased onto main: the Cargo.lock conflict with #478 (new asap-planner
crate) was resolved by regenerating workspace entries with
`cargo update -w --offline`.
@zzylol @claude
feat: add native physical operators and shared DAG runtime
fe80d7f 
Add the execution layer of `asap-physical-operators`:

- `plan`, `runtime` and `sources`: physical DAGs, run context, bounded
  backpressure, cancellation, memory reservations and raw-source scans.
- `expressions` and `operators`: relational/scalar operators, windows,
  temporal panes, current-series snapshots, and summary build, merge and
  readout. A readout is typed: a `SketchQuery`, or an `ExactReadout` whose
  counter lookback resolves to the run's evaluation range. Deserialized
  operators are validated before use.
- `readout`: readouts over merged exact summary states.

There is no physical planner yet; operators are built directly.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat: compile Planner selections into physical DAG candidates
0ced125 
Add `physical_planner`: reader-independent compilation of selected logical
candidates into precompute/query physical DAGs with typed materialization
frontiers, bounded frontier enumeration, workload cost selection, temporal
KLL pane compilation and PromQL row/value lowering. `dag` remains a
compatibility re-export.

Tests cover physical DAGs, plan recovery, precompute candidates and
populations, PromQL values and binaries, weighted TopK binding and
current-series heaps, plus Planner-to-execution integration tests. The design
doc describes the ownership boundary with deployments.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
refactor!: move pane construction out of the physical layer
ee9e0b1 
Pane geometry, pane population checks and per-pane scheduling are deployment
concerns. The physical layer keeps the computation the deployment binds:
per-input summary build, union, shared merge and readout.

- Remove `physical_planner::compile_temporal_pane_candidate` and its
  `TemporalPaneMaintenance`/`TemporalPaneCandidate`/`TemporalEntityIdentity`
  contract.
- Remove the `PaneInput` operator; `ScopeTimestamp` remains as a general
  operator in `operators/scope_timestamp`.
- Drop the `selected_temporal_lifecycle_compiles_panes_and_executes` E2E test
  and update the crate README and design doc.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: state what the physical layer does not own
fb92998 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
refactor: move PromQL series-identity resolution into asap-types
ea0966d 
Planner search needs the physical row representation to propose whole-root
physical alternatives, and asap-aware-mapping cannot depend on the physical
runtime crate. promql_rows::with_series_identity keeps its signature and
delegates.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat: list PromQL physical alternatives in PlanSpace
fb824f9 
The backend built one workload forest per per-query physical alternative by
calling SketchAlgorithmStrategy's proposal methods outside PlanSpace. Planner
now owns them: search_workload_with_targets asks each strategy's new
ReplacementStrategy::propose_for_root once per targeted root.
SketchAlgorithmStrategy resolves the series identity and proposes
current-series TopK, fixed-window and query-time Rate aggregation, and
realizations over per-series Rate state, finalized and deduplicated.

The candidates carry ReplacementProvenance::RootPhysicalRealization: DAG
assembly uses them verbatim, and global selection never commits them.
Existing candidates and their order are unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: describe per-root physical alternatives in candidate enumeration
c57d438 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
refactor: list only current-series heap alternatives in PlanSpace
a3b6f15 
PlanSpace decides what to compute; the summary maintenance lifecycle decides
node timing and the physical compiler reads it. The fixed-window and
query-time Rate aggregation candidates only rewrote a finalization node's
timing, so they are placement variants and no longer go through
propose_for_root. The backend still calls those methods directly.

SketchAlgorithmStrategy::propose_for_root now proposes only current-series
TopK heaps, which rank identity-carrying rows the logical root lacks. The
per-series Rate filter and the verbatim-assembly branch are removed: the
latter is unnecessary because non-Aggregate roots are already assembled
verbatim. RootPhysicalRealization stays so global selection never commits
an unvalidated identity-carrying readout.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: limit per-root PlanSpace alternatives to current-series heaps
c976d5c 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat: expose summary-maintenance lifecycle candidates for explicit bi…
166f257 
…nding

Split plan_summary_maintenance_lifecycles into enumeration and selection so
a deployment can price every lifecycle alternative per unique summary state
and bind its own choice. Planner selection is unchanged: it now enumerates
and then selects the cheapest complete combination through the same path.

SummaryMaintenanceLifecycleCandidates::select validates that each choice is
an alternative Planner could select, enforces schedule compatibility, and
obtains window frameworks and totals from the same complete-candidate hook.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: describe lifecycle candidate enumeration and binding
07244ea 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat: derive Post-ASAP execution timing from a lifecycle plan
590ec6d 
A chosen summary-maintenance lifecycle did not reach the DAG: timing was
fixed by realization strategies. SummaryMaintenanceLifecyclePlan::
execution_timed_dag assigns every node's timing from the selection: retained
(non-Ephemeral) states and their inputs at ingestion time, everything else at
query time. States without a selected lifecycle, and maintained populations
that lifecycle enumeration does not cover, are refused.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
test: carry a chosen lifecycle through timing to physical compilation
9b6dd96 
Rebase note: #483 removed the pane test this test was appended after; the
new test is appended to the #483 version of the file.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: state the lifecycle-owned timing layer contract
3b281da 
Replace the open question about placement ownership with the agreed four-layer
contract: logical Post-ASAP decides what to compute, the chosen lifecycle
decides timing, physical compilation partitions by timing, and the backend
prices lifecycle choices including store cost.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat: derive physical candidates from one compilation
7293332 
`compile_candidate` re-lowered the whole Post-ASAP DAG for every
materialization frontier, and `enumerate_frontiers` compiled it once more.
Number helper operators from their Planner node (`u64::MAX - node_id`,
at most one helper per node) so every boundary choice is a subgraph of
one lowering. `cut_candidate` partitions a `compile` result for one
frontier and `enumerate_compiled_frontiers` enumerates over it; both
produce candidates byte-identical to per-frontier recompilation.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: describe compile-once candidate cuts
d1a82ae 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
test: compare cut candidates by their serialized form
8b59b4b 
The #462 split no longer exposes PhysicalCandidate::encode; its serde form
gives the same byte-for-byte comparison.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat: derive the physical frontier from lifecycle timing
d9c794d 
The lifecycle layer assigns each node's timing; physical compilation now
reads it. `frontier_from_timing` returns the ingestion-time nodes read by
query-time nodes (or an ingestion-time root) and rejects a query-time node
feeding an ingestion-time one, so each lifecycle assignment is a
`cut_candidate` of one `compile` result. `enumerate_compiled_frontiers` is
private: placement comes from timing, and its only caller is
`enumerate_frontiers`.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
test: cut one compilation by chosen lifecycle timings
437a5dc 
For the KLL quantile and grouped Rate->Sum fixtures, ContinuouslyMaintained
and Ephemeral timed DAGs cut one compilation into exactly the candidates
`compile_candidate` builds. The hand-written timing frontier in the chosen
lifecycle test now uses `frontier_from_timing`.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: describe timing-derived candidate cuts
d133d35 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
refactor: allow several helper operators per Planner node
b89bd24 
Number helpers as u64::MAX - (node << 16) - index so a node that lowers to
an operator chain keeps deterministic, traversal-independent helper IDs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
test: expect retained frontier states read at query time or at the root
1a7cbf2 
With several retained states, only those read by a query-time node or
forming the root are cut points.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
fix: keep a selected grouped Sum over realized Rate readouts
0ca3062 
DAG assembly replaced any selected outer Sum over an inner aggregate with a
query-time exact Sum, even when the selected summary realizes the inner Rate
itself. The grouped Sum state therefore never reached the inventory, and no
lifecycle choice could move grouped Sum into precompute.

Assembly now keeps such a selected summary; the query-time residual still
applies when the outer summary would hide its inner aggregate in KeepPreAsap.
Default selection for sum by(job)(rate(...)) now yields Rate -> grouped Sum
state; the physical frontier test reads its query root accordingly.

Conflicts with earlier stack changes resolved to the integration tree:
- crates/integration-tests/tests/summary_maintenance_lifecycle_e2e.rs: c98281a Merge remote-tracking branch 'origin/feat/compile-once-cuts' into integration/planner-for-backend

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
refactor!: remove timing-only Rate placement candidates
ff18c03 
SketchAlgorithmStrategy::fixed_window_rate_candidates and
query_time_rate_aggregation_candidates returned the same logical DAG as the
ordinary heap or grouped Sum candidate with Rate finalization flipped between
ingestion and query time. Placement now comes only from a chosen lifecycle via
SummaryMaintenanceLifecyclePlan::execution_timed_dag.

compile_fixed_window_rate_aggregation takes that lifecycle-timed PostAsapDag
instead of a SummaryNode with baked-in timing. The fixed-window heap test binds
continuously maintained lifecycles; the grouped Sum placement pair is covered
by the lifecycle end-to-end test.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: describe grouped Rate->Sum placement as a lifecycle choice
eb5885c 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat: plan maintained populations through summary lifecycles
0ed79d9 
execution_timed_dag refused any plan with a MaintainPopulation node because
lifecycle enumeration covered SummaryAgg states only, so population timing
stayed fixed by the realization strategy.

Enumeration now also emits one deployment per unique maintained population
that does not feed a SummaryAgg, with the usual alternatives costed through
the caller's lifecycle hooks (unknown stays unknown). select and
execution_timed_dag treat it like summary state: retained at ingestion with
its raw input, Ephemeral rebuilt from raw input at query time. A population
feeding a SummaryAgg is that state's input and follows its timing. A plan
whose population deployment was removed is still refused.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
refactor: make maintained-population timing a lifecycle decision
4e01311 
The execution-data-state validator required MaintainPopulation at ingestion
time, and MaintainedPopulationStrategy hard-coded it there. Timing is now the
lifecycle's choice: the validator accepts either timing and keeps the
structural contracts (the population reads its matching raw input; its
readout is query-time over a population that supports it). The strategy
writes query time as the initial layout, as other realization strategies do.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
test: carry a chosen population lifecycle to physical compilation
761524e 
Conflicts with earlier stack changes resolved to the integration tree:
- crates/integration-tests/tests/summary_maintenance_lifecycle_e2e.rs: 51fd198 Merge #491 population lifecycle into Planner integration

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: describe maintained populations as lifecycle-planned state
76c503e 
Rebase note: keeps the #479 compile-once documentation and test row beside
the population text, as the integration branch does.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
zzylol and others added 23 commits September 30, 2026 18:37
@zzylol @claude
fix(mapping): type the summary state column of without aggregations
6715f96 
A without grouping's keys are the excluded labels, so the state column
index was the excluded-key count, which could fall past the kept labels.
The family was then never written into the schema and exporting the
candidate panicked with SummaryFamilySchemaMismatch.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat(physical): compile per-series vector arithmetic over stored read…
5e651f7 
…outs

Query-time Binary nodes over rows with a series identity, such as
avg_over_time as stored sum/count or rate(a)/rate(b), now compile to
series_labels and series_binary: one-to-one matching on labels without
the metric name, honoring on/ignoring when the IR carries them. A literal
operand drops the metric name and applies to each value.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
fix(physical): compensate PromQL sums and averages like Prometheus
88c0478 
Grouped sum/avg (including current-series readouts) and sum/avg_over_time
now use Kahan-Neumaier summation. An average switches to Prometheus'
incremental mean once the running sum would overflow, so it no longer
becomes +Inf where Prometheus returns a finite mean.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: record per-series arithmetic and compensated summation coverage
c543d8c 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
test: cover infinite sums and name removal on stored readouts; clarif…
000c1ef 
…y docs

Addresses review: pin Inf/-Inf sum and average cases, show that literal
arithmetic drops __name__ from a stored readout, correct the zero-start
comment, and note one-to-one-only coverage and the SQL SUM/AVG change.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
fix: reject duplicate label sets after dropping the metric name
d0f5f97 
PromQL arithmetic with a literal drops `__name__`; if two series then
share a label set, Prometheus errors ("vector cannot contain metrics with
the same labelset"). The per-series literal path returned both rows, and
the Fallback literal path kept `__name__` altogether.

Add `Operator::series_without_name`, a `SeriesLabels` rewrite that errors
on equal resulting label sets, and use it for both literal paths. Matching
and `without` relabeling still tolerate repeats: `series_binary` already
rejects right-hand and matched left-hand duplicates, and aggregation
merges them.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat(ir): carry PromQL bool comparisons as BinaryOpKind::CompareBool
f74c4b2 
The frontend dropped the bool modifier, so `a > bool 1` lowered like the
filter `a > 1`. A separate variant keeps bool off non-comparisons and
reaches both the query-level BinaryOp and the post-ASAP Binary payload.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat(physical): compile PromQL comparisons, set operators, and group …
13baddb 
…modifiers

series_binary now implements Prometheus' VectorBinop, VectorAnd/Or/Unless
and vector-scalar semantics for Fallback subtrees and query-time Binary
nodes, including scalar() operands. Range functions drop __name__ in the
Fallback and reject equal label sets.

Conflicts with earlier stack changes resolved to the integration tree:
- crates/asap-physical-operators/src/physical_planner/promql_rows.rs: a9b8fdd Merge #492 comparison and set coverage with shared series identity typing

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: record comparison and set operator coverage
a6b7337 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
integrate: widen shared series identity to every PromQL binary operator
50f4aef 
The comparisons, set operators, and group modifiers compiled here share
series identity through asap-types (moved there by #477), while
promql_rows.rs no longer keeps its own copy. Accept every BinaryOp kind
and group modifier in `with_promql_series_identity`, excluding only
operands that read the evaluation timestamp. Taken from integration
merge a9b8fdd.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
fix: reject PromQL fill modifiers instead of ignoring them
9d612ed 
VectorMatch has no fill field, so lowering silently dropped fill,
fill_left, and fill_right, changing query results. Reject them with a
fill-specific UnsupportedFeature error so the query falls back to exact
execution, and lower the testdata corpus floor to the measured 1485.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
integrate: record that PromQL fill modifiers are rejected
6eb8a84 
#492 listed `fill`, `fill_left`, and `fill_right` as modifiers the
frontend still ignores; this PR rejects them. Taken from integration
commit 0d132b9.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
fix(types): exclude a renamed sample value from without output labels
9c506d2 
A without aggregate over a nested aggregate kept the renamed value
column (e.g. `sum`) as a label. Identify the value like SampleValue
resolution does.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat: carry le and without (le) grouping for classic histogram_quantile
ad69de7 
HistogramQuantile now names its bucket-bound column, and the PromQL and
SQL frontends group it without (le) instead of by (). The output keeps
every input label except le; the compiler can now recover each histogram.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
fix(physical): match Prometheus bucketQuantile arithmetic
4f9fada 
Divide before scaling when interpolating, and use almost.Equal for the
small-delta fix so an infinite count is not merged into a finite one.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat(physical): compile classic histogram_quantile in PromQL fallbacks
da61e6e 
A without (le) HistogramQuantile aggregate compiles to a series operator
that groups rows by every label except le, applies bucketQuantile, and
drops le and __name__, rejecting equal result label sets.

Conflicts with earlier stack changes resolved to the integration tree:
- crates/asap-physical-operators/src/operators/mod.rs: 2bea2f6 Merge #493 classic histogram quantile coverage
- series_labels.rs and tests/promql_fallback.rs: placement from 2bea2f6
  without the later c39156e/15ab72b changes of this PR

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
test: select and execute histogram_quantile through candidate search
bf2f79e 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
docs: record classic histogram_quantile coverage
b0a1557 
Conflicts with earlier stack changes resolved to the integration tree:
- docs/develop_docs/physical-compile-coverage.md: 2bea2f6 Merge #493
  classic histogram quantile coverage, without the paragraph that 6d895b7
  in this PR adds later. The comparison section from #492 stays, the
  Remaining list drops the entries both PRs now cover, and the totals
  count both PRs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
fix(physical): find no bucket for a NaN histogram rank, as Prometheus
0d60a1c 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
fix(physical): skip out-of-range histogram bounds like ParseFloat
f25df53 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
fix(sql): keep the histogram bridge a single-histogram aggregate
e61ca14 
The SQL marker is the projection's only column, so grouping without (le)
would add undeclared output columns.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
integrate: bind the histogram input layout in series_labels
9b53ea7 
This PR's histogram_quantile branch reads a function-level `left_layout`,
but #492 restructured `series_labels::execute` to bind it inside each
binary branch. Bind it from the operator input in the histogram branch,
and format the widened `Kind` match arm. Taken from integration commit
0d132b9.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol @claude
feat(physical): compile Count-Min bare counts to a Planner readout
112148d 
A deployment reading a stored Count-Min state's total had to call a
kernel accessor itself, so the computation lived outside the compiled
plan. The Planner already expresses that readout as
SketchQuery::PointCount { value: None }; it failed to compile because
plain Count-Min was not a native state.

Plain Count-Min is now a native stored state for merge and bare-count
readout (not for building from rows, which still fails closed). The
kernel answers the bare count through AggregateCore::estimate with one
row's mass, so colliding items keep their weight. The readout is Int64,
matching the Planner's unit-weight count output, and rejects a
non-integral total.

Count-Min with heap stays without a compiled bare count: its native
state is WeightedFrequency, which carries no total.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol
zzylol force-pushed the feat/histogram-quantile branch from fbeddda to 9b53ea7 Compare September 30, 2026 20:24
@zzylol
zzylol force-pushed the feat/cms-scaled-total branch from cc766fe to 112148d Compare September 30, 2026 20:24
@zzylol
zzylol force-pushed the feat/histogram-quantile branch 2 times, most recently from 50c9d6d to ae5f517 Compare October 1, 2026 23:14
zzylol added a commit that referenced this pull request Oct 1, 2026
@zzylol @claude
feat: extend physical compile coverage for deployment computation (#498-
79f3587 
#507)

Count-Min totals, typed series identity, time() operands, non-finite values, compensated exact Sum, label_replace, @ start()/end(), stored-series filters and set operators, deriv/predict_linear, and typed maintenance arithmetic. The compensated Sum keeps its plain Sum name: there is no persisted format to version yet.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@zzylol

zzylol commented Oct 1, 2026

Copy link
Copy Markdown
Contributor Author

Included in the squash merge of #507 into main (79f3587).

@zzylol zzylol closed this Oct 1, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@zzylol