query-db-collection: a later refetch failure of one on-demand subset never reaches that subset's live query

オープン
#2,270 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

まだ誰も着手していません。

評価

難易度
4/5
見積もり時間
3〜5日
初心者へのやさしさ
48/100
issue の種類
バグ
明瞭さ
明確に書かれている
活発さ
活発
技術スタック
typescript
領域
database

調査の方向性

Start with the supplied Vitest reproduction and trace how query-db-collection records refetch failures around query.ts:3265-3280. Then follow how loadSubset:error reaches lastSubsetError in collection-subscriber.ts:108-116, and inspect the related work in #1764. Done when the owning live query exposes a failed refresh, retains its rows, and other subsets remain unaffected.

索引モデルが issue の本文から書いたものです。

説明

  • I've validated the bug against the latest version of DB packages

Describe the bug

With syncMode: "on-demand", once a subset has loaded, a later refetch of that subset that fails (e.g. after queryClient.invalidateQueries) is recorded only collection-wide (collection.utils.lastError). The live query that owns the subset stays ready, its utils.lastSubsetError stays undefined, and nothing tells it the data on screen could not be refreshed.

#1764 (fixed by #1756 / #1797) scoped failures of a subset's first load to the requesting subscription. This is the same expectation for a later load of a subset that already succeeded once.

Tested with @tanstack/db 0.15.2, @tanstack/query-db-collection 1.4.8, @tanstack/query-core 5.104.5 (no related commits on main since 0.15.2).

To Reproduce

Drop-in vitest test (no framework adapter):

import { expect, test } from "vitest";
import { createCollection, createLiveQueryCollection, eq } from "@tanstack/db";
import { QueryClient } from "@tanstack/query-core";
import { parseLoadSubsetOptions, queryCollectionOptions } from "@tanstack/query-db-collection";

const tick = () => new Promise((resolve) => setTimeout(resolve, 0));
const settle = async () => {
  for (let i = 0; i < 30; i++) await tick();
};

test("a later refetch failure is visible on the failed subset's live query, not just collection-wide", async () => {
  const queryClient = new QueryClient({ defaultOptions: { queries: { retry: false, gcTime: 60_000 } } });
  let failing = null;
  const reads = [];
  const collection = createCollection(queryCollectionOptions({
    id: "gap-1",
    queryClient,
    queryKey: ["gap-1"],
    syncMode: "on-demand",
    getKey: (row) => row.key,
    queryFn: (ctx) => {
      const { filters } = parseLoadSubsetOptions(ctx.meta?.loadSubsetOptions);
      const seq = filters.find((filter) => filter.field.join(".") === "seq")?.value;
      reads.push(seq);
      if (seq === failing) return Promise.reject(new Error("read failed"));
      return Promise.resolve([{ key: seq + "-1", seq }, { key: seq + "-2", seq }]);
    },
  }));
  const live = (seq) => createLiveQueryCollection((q) => q.from({ r: collection }).where(({ r }) => eq(r.seq, seq)));
  const a = live("a");
  const b = live("b");
  await Promise.all([a.preload(), b.preload()]);
  expect(a.toArray.map((row) => row.seq)).toEqual(["a", "a"]);

  failing = "a";
  await queryClient.invalidateQueries({ queryKey: ["gap-1"] });
  await settle();

  expect(reads.filter((seq) => seq === "a")).toHaveLength(2);
  expect(a.toArray.map((row) => row.seq)).toEqual(["a", "a"]);
  expect(collection.utils.lastError).toBeInstanceOf(Error);
  expect(a.status === "error" || a.utils.lastSubsetError instanceof Error).toBe(true);
  expect(b.status).toBe("ready");
  expect(b.utils.lastSubsetError).toBeUndefined();
});

Result: the last two a assertions fail; collection.utils.lastError is set.

Expected behavior

The failed subset's live query exposes the refetch failure (error status, or at least utils.lastSubsetError), while its rows stay (a failed refresh should not empty the view) and other subsets' live queries are unaffected.

Additional context

  • The error is recorded collection-wide at query-db-collection/src/query.ts:3265-3266, and marks the collection only while it is loading (:3278-3280). lastSubsetError is fed only by loadSubset:error (db/src/query/live/collection-subscriber.ts:108-116), which an observer refetch failure never emits.
  • Related: #672 (first-class error tracking), #1764.
  • Why it matters: an app showing paged server data (ratings, history) cannot tell a member "couldn't refresh, retry" for one view. We currently read the query cache directly to find a failed subset's error.
  • Happy to send a PR (with the oracle coverage CONTRIBUTING asks for) if you'd like to assign it; equally fine if you'd rather fix it in your own flow. The test above is free to adopt as-is.

The repro was written with an AI coding assistant and checked by hand against the installed sources.

主要言語
TypeScript
スター
4k
フォーク
277
平均マージ
1日 7時間
マージ済み PR(30日)
170

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

TanStack/db のほかの issue

TanStack/db の issue をすべて見る

似ている issue

TypeScript の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。