perf(shadow-indexer): skip payload rewrite on canonical-only upsert

[wbj-cb]

Summary

Skip the ~176KB JSONB payload rewrite when a shadow_blocks upsert only needs to stamp a canonical_hash.

Background

insert_rows upserts blocks with ON CONFLICT (number) DO UPDATE. Two kinds of conflict reach the DO UPDATE:

1. New candidate hash for a height (reorg replacing the stored block) — genuinely new data.
2. Same-hash redelivery carrying a canonical_hash — the block is unchanged; we only want to record that it became canonical.

Previously both cases assigned payload = EXCLUDED.payload unconditionally. In case (2) this reassigns the large JSONB payload to a byte-identical value, but Postgres still rewrites the row's TOAST chunks and emits full-page WAL for the tuple — hundreds of KB of write amplification just to set one hash column.

Change

Guard payload and created_at with a CASE keyed on whether the hash actually changed:

ON CONFLICT (number) DO UPDATE SET
  hash = EXCLUDED.hash,
  canonical_hash = EXCLUDED.canonical_hash,
  created_at = CASE WHEN shadow_blocks.hash <> EXCLUDED.hash
      THEN EXCLUDED.created_at ELSE shadow_blocks.created_at END,
  payload = CASE WHEN shadow_blocks.hash <> EXCLUDED.hash
      THEN EXCLUDED.payload ELSE shadow_blocks.payload END,
  updated_at = now()
WHERE shadow_blocks.hash <> EXCLUDED.hash
   OR EXCLUDED.canonical_hash IS NOT NULL

New hash (case 1): payload/created_at land wholesale, exactly as before.

• Same-hash canonical stamp (case 2): payload is reassigned to itself. Postgres keeps the existing TOAST pointer, so no payload rewrite and no full-page WAL for the payload — only canonical_hash/updated_at change.

The WHERE clause is unchanged: redeliveries that neither change the hash nor add a canonical hash are still dropped.

Impact

• Removes payload rewrite + full-page WAL amplification on the canonical-stamping path, which is the common steady-state conflict.
• No behavioral change to stored data: identical rows result in both cases.

Testing

• cargo test -p base-shadow-indexer-db — 4/4 pass
• cargo clippy -p base-shadow-indexer-db — clean
• cargo check -p base-shadow-indexer-db — clean

Notes

• No schema/migration change; SQL-only.
• Follow-up perf work (payload compression tuning, insert-only payload side table) is tracked separately.

[cb-heimdall]

🟡 Heimdall Review Status

Requirement	Status	More Info
Reviews	🟡 0/1	Denominator calculation Show calculation 1 if user is bot 0 1 if user is external 0 2 if repo is sensitive 0 From .codeflow.yml 1 Additional review requirements Show calculation Max 0 0 From CODEOWNERS 0 Global minimum 0 Max 1 1 1 if commit is unverified 0 Sum 1

[depot-code-access]

✅ All benchmarks green — 14 within ±2% (deterministic instruction counts). View run

Benchmark details (14)

Benchmark	Base (target)	Head (this PR)	Δ instructions
batch_queue/drain/drain_cached_span_batches	242,027	242,027	+0.0%
batch_transaction/encode_in_place/encode_in_place	4,199,759	4,199,759	+0.0%
batch_transaction/temporary_frame_buffers/temporary_frame_buffers	8,408,350	8,408,350	+0.0%
flashblock_decode/decode/brotli	3,296,484	3,296,484	+0.0%
flashblock_decode/decode/plain_json	2,280,194	2,280,194	+0.0%
flz/compress_len/real_contract_call	43,148	43,148	+0.0%
flz/compress_len/synthetic_0	38,205	38,205	+0.0%
flz/compress_len/synthetic_1	54,682	54,682	+0.0%
flz/compress_len/synthetic_2	147,976	147,976	+0.0%
flz/data_gas	43,059	43,059	+0.0%
flz/tx_estimated_size	43,056	43,056	+0.0%
frame_parse/decode/single_4kib	1,031	1,031	+0.0%
frame_parse/parse_frames/few_large	1,053,062	1,053,062	+0.0%
frame_parse/parse_frames/many_small	154,763	154,763	+0.0%

[github-actions]

Review: No issues found.

The SQL change is correct and well-targeted:

• NULL safety: hash is NOT NULL in the schema, so shadow_blocks.hash <> EXCLUDED.hash never hits three-valued NULL logic. No edge case.
• Case analysis: The three conflict scenarios (new hash, same-hash canonical stamp, same-hash redelivery without canonical) all produce identical stored outcomes to the prior version. The only difference is that case 2 now avoids a TOAST rewrite by self-assigning payload and created_at instead of reassigning from EXCLUDED.
• WHERE clause: Unchanged and still correctly drops pure redeliveries (same hash, no canonical hash).
• No block-production sensitivity: This touches the shadow indexer DB (observability for reorged blocks), not any block production, execution, or payload assembly path.

Clean optimization with no behavioral change.

[depot-code-access]

Base Std historical fork tests

Fork	Result	Passed	Failed	Skipped	base/base	base-anvil	base-std
Beryl	pass	616	0	13	d0e0d4f2	cecddfa5	4658f1b7
Cobalt	pass	721	0	14	d0e0d4f2	fb00db40	e30b3421

View run