Release v0.14.2 · FlavioCFOliveira/GoGraph

GoGraph v0.14.2 — see CHANGELOG.md for the per-release entry.

0.14.2 — 2026-09-15

9 commits — 3 fixes, 4 documentation changes and 2 merges. Counted at 92c47fda, the
branch tip before release preparation; the release-preparation commits that carry this entry
are necessarily not in their own count. One sprint delivered this window — 361, bug fixing
campaign 26.09.15 — and all 12 of its tasks closed. 25 files changed, 3 235 insertions,
103 deletions (git diff --shortstat v0.14.1..92c47fda).

The 12 closed tasks and the 3 fix commits that carry them match in both directions — twelve
ids in the sprint, twelve distinct ids across the commit bodies, empty both ways, with each fix
commit's Refs footer naming exactly four. This is not a 1:1 task-to-commit mapping and is
not claimed as one: the sprint was worked as three batches of four tasks sharing a scope, which
is what the project's batching rule asks for. The remaining six commits are four documentation
commits, the sprint merge and the v0.14.1 back-merge.

What the insertion count is, because 3 235 lines is not 3 235 lines of engine. Bucketed by
path from git diff --numstat v0.14.1..92c47fda:

bucket	files	insertions	deletions
`*_test.go` (regression tests)	13	2 254	10
engine `*.go` (non-test, shipped)	5	354	30
`internal/sim/` (the simulator's oracle, non-test)	4	318	15
root Markdown (`CLAUDE.md`, `README.md`)	2	255	46
`docs/` (prose)	1	54	2
total	25	3 235	103

The shipped engine change is 354 insertions across five files, and one of those five is
godoc. bolt/server/serve.go (+23/−4) changes comments only and has no compiled effect, so
four files carry the whole of this release's behaviour: cypher/api.go (+181/−21),
bolt/server/errors.go (+66/−2), cypher/exec/set.go (+53/−3) and cypher/exec/set_all.go
(+31/−0). Against that stand 2 254 lines of regression test across 13 files — the shape a
bug-fixing sprint should have, and the sprint's own rule: every fix closes with a test that
fails on the old behaviour and passes on the corrected one.

One defect runs through this entire release: the module said one thing and did another.
Every item below is an instance of it. A SET reported success and discarded the value
(#2816). An undirected SET reported two relationship writes and made one (#2817). A
DROP INDEX … IF EXISTS reported a removal that never happened (#2818). A statement that
failed and rolled back still reported the counters for the work it had just undone (#2823).
Three faults the client had committed reached it as "An internal error occurred" (#2819).
Three godocs asserted intent the code did not implement (#2811, #2824), and one counter
counted more than its name admits (#2791). This is a release about the gap between what
GoGraph reports and what GoGraph does, and it closes that gap in every place the sprint
found it.

Two of the twelve tasks refuted their own premise and shipped no fix (#2815, #2820). Both
are reported in full under Notes rather than quietly dropped, because a report that turns
out not to reproduce at HEAD is a result, and the regression coverage written to establish
it ships with the release.

go.mod and go.sum are byte-identical to v0.14.1 — same pinned toolchain, same
dependency set at the same versions (git diff v0.14.1..92c47fda -- go.mod go.sum is empty).
There is nothing to report under a dependency heading and nothing is invented to fill one.

The exported API is unchanged, and that is what makes this a PATCH. Five functions were
added and all five are unexported — relStorageDirection, isEntityPropertyValue,
exprValueIsEntity, errEntityPropertyValue and dropIndexCounter. No exported identifier
was added, removed or re-signed, and no exported struct gained a field. Under
docs/semver.md a new exported identifier is the MINOR trigger and there is
not one, so 0.14.2 is a PATCH by the rule rather than by decision — which is the opposite of
v0.14.1, where a MINOR surface was shipped as a PATCH deliberately. There is therefore no
Added section in this entry, and the heading is omitted because there is genuinely nothing
to put under it, not because the surface went unexamined.

No change is marked breaking — no commit in the window carries a BREAKING CHANGE footer
or a ! subject, verified over the full commit bodies rather than the subjects alone. Three
behaviours a caller can observe do change, and every one of them replaces a wrong answer with
a right one; they are itemised under Fixed and revisited in
release-notes/v0.14.2.md.

The openCypher TCK gate is unchanged at 3 897/3 897. const tckExecutionBaseline = 3897
in cypher/tck/runner_test.go is untouched and no .feature file changed in this window,
so the scenario population is the one v0.14.1 was measured against.

All 9 commits in this window are unsigned (git log --pretty=%G? returns N for every
one). docs/release.md asks for this to be re-measured per release rather
than assumed; this is that measurement, and it has not changed.

Fixed

Cypher — the effect a statement reports now equals the effect it applied

A SET no longer discards a node-, relationship- or path-valued property while reporting
success (#2816). isStorableProperty rejected only maps, so five write paths each read its
ok=false as "no value was produced" and moved on. The defect was wider than reported: on
the REPLACE forms the dropped entry still cleared the existing keys, so SET a = {other: b}
answered ok with propertiesRemoved=1 and left the node with no properties at all —
silent destruction of data the caller never asked to remove. All five sites now raise
InvalidPropertyType. Gated by cypher/set_entity_property_value_test.go.
An undirected SET over a reciprocal relationship pair now writes both relationships
(#2817). resolveRelBinding keyed the edge-property mutation on the row's traversal order,
and the reverse hop of an undirected pattern is the swap of the edge's storage order — so
both rows addressed the same stored edge, one edge was stamped twice and the other never
touched. id(r) returns two distinct relationships, so under openCypher 9 relationship
isomorphism the statement owes two rows and two writes: the counter was right and the write
was missing. relStorageDirection normalises the target to storage order. Gated by
cypher/set_undirected_reciprocal_test.go and
cypher/exec/set_rel_storage_direction_test.go.
DROP INDEX … IF EXISTS no longer counts a removal that never happened (#2818). Both
branches of runDropIndex bumped IndexesRemoved unconditionally, so absorbing a missing
index reported indexesRemoved: 1 over an index listing identical before and after.
dropIndexCounter now reads for existence before the operator runs and reports the effect
actually applied. WAL behaviour is unchanged — nothing was ever written on the absorbed
path; only the report was wrong. This fix has a measured cost, stated under Performance.
A statement that failed and rolled back no longer reports the counters for the work it
undid (#2823). rollbackUnderBarrier correctly reversed every eager mutation and then left
r.counters pointing at them, so a caller that inspected the result of a failed statement was
told about writes that no longer existed. It is not MERGE-specific as reported — a plain
multi-assignment SET and a CREATE … WITH … SET reproduce it identically. The evidence for
the corrected behaviour is unanimous across three independent sources. The TCK corpus never
asks for counters from a statement that raised. Of the 1 615 scenario blocks in
cypher/tck/features/, 192 carry a "should be raised" step and 1 423 carry a side-effect
assertion, and the intersection is empty. The 1 423 counts both forms the TCK uses, because
both assert about side effects: 244 state them as an explicit the side effects should be: table and 1 179 use no side effects, which asserts just as positively that the
statement changed nothing (244 + 1 179 = 1 423 exactly, so no block carries both forms). It is
still empty on the executed population, where those 1 615 blocks expand to the 3 897 scenarios
the baseline counts: 695 raising, 3 202 asserting side effects, 0 both. A Bolt FAILURE
message carries no stats field, and the reference driver returns no summary at all when the
query failed. Gated by
cypher/failed_statement_counters_test.go.

Bolt — three client faults the server was masking

Three errors a client can act on now reach it as client errors instead of
"An internal error occurred" (#2819), and they were masked for three different reasons.
txn.ErrFieldTooLong had no FailureCode case at all; the unsupported-DDL refusal arrives
untyped, as a wrapped fmt.Errorf; and validatePreExisting's NOT NULL arm wrapped only the
bare sentinel where errors.As needed the type. They are now
Neo.ClientError.Statement.ArgumentError, Neo.ClientError.Statement.SyntaxError and
Neo.ClientError.Schema.ConstraintValidationFailed respectively. Each code follows
precedent already in the module rather than being invented for the occasion, and the
reasoning is recorded beside the branch that uses it:
Neo.ClientError.General.LimitExceeded was considered for the over-long field and
rejected, because rmp #2561 established that it does not appear in Neo4j's status codes
at all. Every forwarded
message was checked to carry only static text, the client's own tokens and a format
constant — no server path, no internal identifier — and that check is now asserted
permanently rather than performed once, in
bolt/server/named_client_faults_wire_test.go.

Instruments that could not see what they measured

The deterministic simulator can now see a DDL counter (#2822). No DDL counter reached any
oracle, because the simulator discarded the Result — which is precisely why #2818's oracle
criterion had been met only vacuously, by an oracle structurally incapable of observing
the quantity it was said to check. The new oracle derives its expectation from the engine's
own index and constraint registries, read either side of the statement, and it was proved
to have teeth by a mutant: restoring the pre-#2818 unconditional increment turns it red.
internal/sim/ddl_counters_oracle.go, with internal/sim/ddl_counters_oracle_test.go.
A metrics wire test that failed about one run in seven now cannot (#2821). This was not
a server race. ObserveLatency interns the histogram before it records into it, so a scrape
landing between those two steps emits a complete, well-formed _count 0 line — which the
test's substring wait accepted as "the metric has appeared" before asserting on the zero it
had just raced. The wait now requires the value, not the line.

Changed

Benchmark execution has been removed from the release path, by decision, in the
release-preparation commits — which are therefore not in this entry's 9-commit count.
make release-accuracy no longer requires docs/benchmarks/VERSION.md to exist (its other
five checks are untouched), and make release-preflight no longer runs
scripts/run_headline_bench.sh. The correctness gate is unchanged: release-accuracy
plus the full make ci — go test -race ./..., the TCK =100 % baseline, golangci-lint,
govulncheck and the coverage gate — still stands between a tag and a publication.
This is a change of gate, not of standard. Measurement still decides every performance
question in this project and scripts/bench_gate.sh still compares a candidate against its
baseline locally, before a change lands; what stopped is producing a per-release campaign as a
precondition for tagging. docs/release.md and the release.yml comments
are corrected to describe the gate that now exists.

Performance

This release ships a measured benchmark record, and that record is explicitly
underpowered — read what it does not support before quoting it.
docs/benchmarks/v0.14.2.md compares v0.14.1 and v0.14.2
first-hand on one host, in three interleaved arms with a same-binary noise floor measured in
the same rounds. The campaign was designed as six rounds and was stopped by decision after
one complete round plus a partial second, when benchmarking left the release path. At
n ≤ 2 per arm no p-value is computed anywhere in that document, and benchstat's
confidence interval is unavailable. A row that separates but does not clear the campaign's
own false-positive envelope is reported there as not adjudicable at this power — not as
unchanged — and that distinction is preserved here.

Three packages are proved unchanged rather than measured unchanged. search,
search/centrality and graph/index/count compile to byte-identical test binaries
(sha256, -trimpath) at v0.14.1 and at v0.14.2. Those three host the release gate's
entire headline set — Dijkstra_PostWarmup, Dijkstra_Large, BFSDirectionOpt_PowerLaw,
Yen_K100, Brandes_RandomGraph — so the headline benchmarks execute identical machine
code at both releases. A hash establishes equality; a benchmark could only ever fail to
detect a difference. Their measured rows are used as a cross-arm noise floor, not as results.
The DROP INDEX … IF EXISTS existence read costs +115 ns and +2 allocations on the
absorbed path — 1.003 µs → 1.117 µs, +11.4 % (#2818). This is the one time cost in the
release established beyond any question of statistical power: arm A measured 1 000 and
1 006 ns while arms B/B2 measured 1 101–1 150 ns, four samples against two with the
ranges fully disjoint, in both rounds and in both arm orders. It is published rather than
buried because it is what a correctness fix cost. In proportion it is a DDL statement's
cost, not a query path's: DROP INDEX runs once per statement under the engine's exclusive
DDL lock, and on the real-removal path (H_CreateDropIndex, which pays a CREATE INDEX in
the same iteration) the added read does not rise above the floor at all — 19.02 → 18.77 µs
with allocations unchanged at 108.
#2817's storage-order normalisation costs allocations, as exact per-operation constants:
+1 alloc/op on a one-row directed relationship SET, and +10 allocs/op and +458 B/op
on a two-row undirected statement. Every sample within each arm is equal in both arms, so
these hold at any sample size. Part of the +10 is not overhead but the second edge write
v0.14.1 wrongly skipped — at v0.14.1 both rows wrote to the same pair, which was the
bug. That split is a hypothesis consistent with the diff, not a measured attribution:
separating the two would need a third arm with the normalisation disabled, and none was
built. No time cost is resolvable on either shape at this power.
#2816's SET entity-value rejection shows no measurable cost. The three node-SET
shapes — scalar, REPLACE-map, and the plan-miss form that forces a fresh plan build — hold
their allocation counts exactly at 101, 130 and 102 across both arms, and their timings
sit inside the false-positive envelope. The guard is a type switch on a value the evaluator
already holds.
bolt/server changed and did not move. Its edit is on the error path; none of its 24
benchmark functions names FailureCode, sanitiseErr or isClientFaultErr, so no benchmark
reaches the new code and none was expected to. All 27 rows were run anyway: block geomean
+0.10 % against a +0.32 % same-binary floor.
Of the 212 comparable rows in the four packages whose binary did change, two clear the
calibration and 27 separate without clearing it. The second of the two,
exec/BuildReverse_vs_E/E=1000000 at −8.61 %, is an improvement that no changed symbol
explains and is reported as unexplained rather than claimed — nothing in set.go or
set_all.go is on the reverse-adjacency build path.

The published concurrency ladders were not re-measured. The -test.cpu=1,8,64,256,1024
ladders v0.14.1 published were in the six-round design and were dropped with the campaign,
not by choice: the concurrency levels this project publishes are unmeasured for this
release. Latency percentiles at those levels are unmeasured for the ninth consecutive cycle.

Documentation

MaxInFlightPerConnection's godoc carried three false claims and is corrected (#2811) —
the code it said was emitted, the retry classification it asserted, and the session state it
described after the cap trips. Of the three godocs reported stale in that task, two had
already been corrected by #2806 in the v0.14.1 window; that is stated rather than
re-claimed as new work. Gated by bolt/server/inflight_cap_doc_agreement_test.go.
runDDLOpCounted's godoc asserted intent rather than code and is corrected, with its
siblings swept in the same pass (#2824).
docs/bolt.md now declares the properties-set counter divergence (#2791): the Bolt
properties-set counter carries removals as well as additions. It is documented as a
protocol-conformance divergence, with its reproduction, because the incumbent has no
separate counter either — changing the number would diverge from the implementation every
client is written against. No counter code changed — the only non-test change under
bolt/ in this whole window is FailureCode's three new branches (#2819) and a godoc
block — which is why this appears here and not under Fixed.

Compliance

100 % openCypher TCK-compliant at the execution level — 3 897 / 3 897 scenarios,
against const tckExecutionBaseline = 3897, which is untouched. No .feature file changed in
this window, so the population is the one v0.14.1 was measured against. This release
preserves the compliance; it does not extend it. #2823's fix was decided against the TCK
corpus rather than against intuition: of 1 615 scenario blocks, 192 raise an error and
1 423 assert side effects — 244 as an explicit table, 1 179 as no side effects — and
no block does both, nor does any of the 3 897 scenarios those blocks expand to.
100 % ACID-compliant. #2816 and #2817 are Atomicity and Consistency failures in the
narrow sense that matters to a caller — a statement that reports an effect it did not apply
leaves the caller's model of the graph inconsistent with the graph — and #2823 closes a result
that described state a rollback had already removed. No durability path changed in this
window: store/wal, store/recovery, store/checkpoint, store/snapshot and store/txn
are byte-identical to v0.14.1, and #2818 alters a counter, never a WAL frame.

Notes

Pre-1.0 stability. This is a 0.y.z release. The public Go API may change without a
major-version bump until 1.0.0; pin the exact version you depend on.
Module path. The Go module path is github.com/FlavioCFOliveira/GoGraph with no /vN
suffix, which is Semantic-Import-Versioning-correct for a 0.x line.
Two premises were refuted, and neither shipped a fix. #2815 reported a
relationship-property read divergence; it could not be reproduced at HEAD on any path —
embedded API, Bolt 4.4 and 5.0, autocommit and managed transactions, directed and undirected,
multigraph and simple, WAL-backed, and across a store reopen. #2820 reported a
DROP CONSTRAINT twin of #2818; that twin does not exist, because dropConstraintLocked
must resolve the name to a kind, label and property before it can act and returns the empty
result on an unresolved name before the counting wrapper is reached. Both shipped as
permanent regression coverage (cypher/rel_accessor_agreement_test.go,
bolt/server/rel_accessor_agreement_wire_test.go,
cypher/drop_constraint_if_exists_counter_test.go) so that the behaviour each was said to
violate is now defended by a test, whatever the original report meant.
No store migration, and no on-disk format change. A store written by v0.14.1 opens
unchanged; nothing in this window touches a snapshot, a manifest or a WAL record layout.
Examples are not part of the module. examples/ is an exercise harness; the module
neither imports nor depends on it. No example changed in this window.
Gates not run for this release. The soak and nightly layers were not run, and this release
carries no production certification of its own — the most recent was taken at the
v0.11.0 commit. The benchmark campaign behind
docs/benchmarks/v0.14.2.md was stopped after one complete
round of six, and the concurrency ladders were dropped with it.

Navigation Menu

Repository navigation

v0.14.2

Choose a tag to compare

Sorry, something went wrong.