示例
nanostores 1.4.2: Work out what nanostores' lazy mounting actually does: which argument of a listener is the changed key, why .get() fires onMount and re-runs it every second, why a computed store accepts .set(), and why .value and .get() disagree
已验证示例 — npm nanostores 1.4.2: Work out what nanostores' lazy mounting actually does: which argument of a listener is the changed key, why .get() fires…
sha256:2bad4d8c3646df6d3f3c67f5c83c45192b4aa22c2f9b886001b21a310347289f
本网络只提供一件事：能构建的样本。它在沙箱中运行并保留签名回执。它不评级、不担保——同样的代码能否在你的环境构建，它没有测量过。
提交了通过的契约回执的不同签名密钥数量。为 1 表示只有作者；大于 1 表示还有其他人构建过。密钥是自行生成的，背后没有注册身份，因此计的是密钥而非人。
MIT-0
执行证据
声明的环境与签名的运行分开呈现，你可以看到这个样本究竟运行了什么、在哪里运行。
- 证据依据
- 签名契约通过
- 验证回执
- 1
- 构建过它的签名密钥
- 1
声明的环境
node 22 linux x64 node 22 javascript npm
验证运行环境
| 环境 | 契约 | 阶段 | 运行日期 |
|---|---|---|---|
| node 22 · linux alpine/x64 · docker ed25519:a2ec939a4c60e243 | PASS | compile:SKIPPED · contract:PASS · load:PASS · resolve:PASS CONTAINER_RUN · node-typescript@1 |
2026-08-14 |
案例
HOW- 目标
- Work out what nanostores' lazy mounting actually does: which argument of a listener is the changed key, why .get() fires onMount and re-runs it every second, why a computed store accepts .set(), and why .value and .get() disagree
- 符号
-
- nanostores.atom
- nanostores.map
- nanostores.computed
- nanostores.batch
- nanostores.onMount
- nanostores.keepMount
- nanostores.cleanStores
- nanostores.listenKeys
- nanostores.deepMap
- nanostores.readonlyType
- nanostores.task
- nanostores.allTasks
- nanostores.STORE_UNMOUNT_DELAY
- Store.listen
- Store.subscribe
- Store.setKey
- Store.get
- Store.value
- Store.lc
- globalThis.nanostoresGlobal
- 环境
- node 22
- 创建时间
- 2026-08-14T20:16:45Z
契约
- assert every store, atom and map alike, calls its listeners as (value, oldValue, changedKey), so a map listener's second argument is the previous whole object and the changed key arrives third
- assert an atom fills all three slots and leaves the third undefined, which is why the changed key cannot be the second argument
- assert subscribe's priming call passes exactly one argument, so arguments.length is the only thing separating it from a real change whose previous value was undefined
- assert setKey(key, undefined) deletes the key rather than storing undefined, so `key in value`, Object.keys and JSON.stringify all lose the field
- assert deleting an already absent key is not a change and notifies nobody
- assert set() compares with === and nothing deeper, so an equal-looking new object is a change while the store's own current object is not
- assert readonlyType is an identity function returning the very same store object, so nothing about read-only-ness reaches runtime
- assert .set() on a computed store neither throws nor is ignored, and that the written value sticks across re-reads until one of its own dependencies really moves
- assert a fresh computed's .value property reads undefined while .get() returns the computed number, so the two disagree until the first mount
- assert .get() on an unmounted store runs the onMount initializer and returns with the listener count back at zero, so reading is not passive
- assert repeat reads inside the delay window do not re-mount, because the store is still flagged active
- assert STORE_UNMOUNT_DELAY is 1000 and the onMount cleanup runs on that timeout rather than when the last listener leaves
- assert that same delay makes a store which is only ever read with .get() re-run its mount side effect once the window expires
- assert batch() collapses many writes into a single listener call carrying the final value, the value from before the first write, and an undefined changed key
- assert the same two writes unbatched deliver two calls, each naming its own key
- assert listenKeys stays quiet for an unwatched key and fires for a watched one in both modes: batched, where it must diff old against new because the key argument is missing, and unbatched, where the key is present and Set membership decides
- assert the computed callback does not run at construction and does not re-run on repeat reads
- assert an unrelated store's change bumps the single epoch counter that nanostores parks on globalThis so separately bundled copies share it, leaving the recompute guard stale, and that the callback still does not re-run because the per-store argument comparison is unchanged
- assert the callback re-runs only when this store's own dependency actually moves
- assert a plain async arrow in computed() stores the Promise object itself as the store value, with no warning and no throw, because the async path is gated on a .t marker rather than on being a thenable
- assert the identical shape wrapped in task() does take that path, warning that Promise support in computed() is going away and leaving the store undefined at the moment get() returns
- assert awaiting allTasks() is still too early to read that value, because it resolves before the .then() computed chained onto the same promise, and the value only appears a turn later
- assert keepMount returns undefined because it discards the unsubscribe function, and that no dropMount export exists to undo it
- assert cleanStores throws "cleanStores() can be used only during development or tests" when NODE_ENV is production, rather than degrading to a no-op
- assert cleanStores outside production unmounts a pinned store synchronously, skipping the 1000 ms delay
- assert deepMap still exports from the core package but announces on construction that it moved to @nanostores/deepmap and will be removed in 2.0, and that its setKey takes a dotted path that descends rather than creating a literal dotted key
- assert that notice is deduped by exact message text, so a second deepMap is silent and the deprecation can be missed entirely
- assert the core package has no create, createStore or useStore, the names carried over from the better known zustand/jotai/redux family
文件
- csx.json
- package-lock.json
- package.json
- src/stores.mjs
- test/contract.mjs