query-db refetch oracle: a window row is missing from the source Collection after a held delete is released #2214
Send to my agentSummary#
The query-db refetch oracle fails on main at bd10e564e for one random seed. The history has a held delete, two upserts, a refetch, and the release of the held delete. After it, a row sorts inside the live window, but the source Collection does not have it.
Reproduction#
pnpm install
pnpm --filter @tanstack/db-ivm build
pnpm --filter @tanstack/db build
pnpm --filter @tanstack/db-sqlite-persistence-core build
cd packages/query-db-collection
TANSTACK_DB_ORACLE_SEED=-116668131 \
TANSTACK_DB_ORACLE_PATH="71:3:0:2:1:1:2:3:0:5:5:4:0:0:0:0:3:1:2:4:4:7" \
TANSTACK_DB_ORACLE_PROPERTY=query-db.refetch-ordered-prefix \
npx vitest run tests/refetch-ordered-prefix-oracle.test.ts
The random campaign of query-db.refetch-ordered-prefix found the seed. In my runs, approximately 1 of 3 random campaigns failed.
Shrunk counterexample#
{
"scenario": { "encoding": "cursor", "functionKey": false, "terms": 1, "orderKey": "rank", "limit": 1, "offset": 0 },
"initial": [
{ "id": 0, "rank": 0, "marked": false },
{ "id": 1, "rank": 0, "marked": false }
],
"history": [
{ "type": "hold", "kind": "delete", "id": 1 },
{ "type": "upsert", "id": 2, "rank": 0 },
{ "type": "upsert", "id": 1, "rank": 1 },
{ "type": "refetch" },
{ "type": "release" }
]
}
Failure#
coverage after release at step 4: rows [{"id":2,"rank":2,"label":"a2"}] sort at or before the window's end but are [null] in the source Collection
The coverage law applies after the release of the held write. Every row that sorts at or before the end of the served window must then be in the source Collection. Here, row 2 sorts inside the window, but the source Collection does not contain it.
Scope#
- The failure also occurs with the live-query sources of #2194 and of its merge base, so it does not come from #2194.
- I did not find the cause. The shrunk history has a held delete of row 1 while row 2 is inserted, row 1 moves, and a refetch runs. That points at the refetch and the release of the held write, possibly near #2206.
This issue was written by Isaac.
Sign in to join this conversation.