Skip to content

deps: adopt lower-allocation OTLP decoder stack - #176

Merged
casselc merged 1 commit into
mainfrom
perf/sdk-scalar-defaults-20261020
Oct 20, 2026
Merged

casselc merged 1 commit into
mainfrom
perf/sdk-scalar-defaults-20261020

Conversation

@casselc

@casselc casselc commented Oct 20, 2026

Copy link
Copy Markdown
Contributor

Summary

Update the root OTel SDK from 26186ce to 3366a33, composing the reviewed optimizations from casselc/otel #65 and #66 with the existing stock-codec/cache capabilities.

The decoder reuses original scalar defaults and scans its seven fixed AnyValue fields without constructing a lazy filtered sequence. Profiling identified both as avoidable per-attribute allocation. These are ordinary library changes on stock Jolt; no compiler patch or new dependency is needed. Existing validation, error handling, optional codec/parser defaults and Durable confirmation rules stay unchanged.

Update the exact resolved-SDK source guard; retain all older negative controls. The minimal embedded profile is not repinned. Document why a global native-library override is inappropriate when testing root and minimal profiles with different native versions.

Validation

  • Final combined SDK: official stock Jolt 0.8.23 / Chez 10.4.1, 476 tests / 5,484 assertions, zero failures/errors.
  • Final Oscope headless suite: 361 tests / 3,539 assertions, zero failures/errors. Leave JOLT_CHDB_LIB unset so root and unchanged minimal fixtures use their own versioned assets. An earlier globally forced 26.9 override correctly failed the minimal fixture's 26.7.3 version guard; no source guard was relaxed.
  • Final real config-selected stock-reader fixture on native chDB 26.9.0: 45 assertions, three genuine reader accepts / zero declines, actual native stream/codec use, and separate default-reader Durable recovery terminal/settled/valid.
  • Claude read-only review approved the repin conditional on those final consumer gates; both are now complete. The underlying scalar and fixed-field SDK changes also have independent reviews and source/behavior controls.

These application gates establish functional integration, not a throughput/tail gain. Decoder-only studies and the earlier scalar-only socket allocation comparison retain their original source scopes. Pre-existing typed benchmark provenance drift is tracked separately in #175; no historical measurements are relabeled by this repin.

Includes a changelog entry. No S3 performance, app packaging, or expanded persistence guarantee is claimed.

@casselc
casselc merged commit 2615d30 into main Oct 20, 2026
7 checks passed
@casselc
casselc deleted the perf/sdk-scalar-defaults-20261020 branch October 20, 2026 20:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant