mirror of
https://github.com/SigNoz/signoz.git
synced 2026-09-19 01:40:41 +01:00
0fe45e7114
Some checks failed
build-staging / prepare (push) Has been cancelled
build-staging / js-build (push) Has been cancelled
build-staging / go-build (push) Has been cancelled
build-staging / staging (push) Has been cancelled
cacheci / tests (push) Has been cancelled
Release Drafter / update_release_draft (push) Has been cancelled
#### Description Stacked on #12715. Makes the list-view ("all fields", empty `selectFields`) attributes bag read from the `attributes` JSON column, evolution-aware, so the bag stays complete once we stop dual-writing the legacy maps. - The empty-`selectFields` list path was a static, map-only column list (`attributes_string/number/bool`). It now picks attribute homes by column evolution for the query window: legacy maps before the JSON rollout, the `attributes` JSON column after it, both across it — via a new `MetadataStore.GetColumnEvolutions` and `tracestelemetryschema.BulkAttributeColumns`, mirroring the per-key selection. - `mergeSpanAttributeColumns` merges the natively-decoded JSON column (relies on #12555's `JSONValue` scan — no `toJSONString`, no second unmarshal) into the flat bag: nested paths flatten back to dotted keys and win over the legacy maps on collision, so the response shape is identical across the rollout with a single decode and no re-marshal. - Per-key group-by/aggregation is unchanged — it already resolves JSON-only after the rollout — so those queries keep working once map dual-write stops. Follow-up (out of scope, tracked by #12555's known gaps): the waterfall/flamegraph struct readers need the same `JSONValue` type before span attributes can go JSON-only; `resources_string` moves to the `resource` JSON column later. #### Additional Information Tests: bulk select across before/after/straddle windows; group-by after rollout reads JSON not the map; consume flatten/merge (nested flattening, type preservation, JSON-wins-on-collision). Assisted-by: Claude Opus 4.8