Le Truc 2.5 doesn't add a feature you'll reach for on its own. It exists to make the next Cause & Effect release land without breaking your code.

The gap that forced bad patterns#

Cause & Effect's reactive primitives up to 1.4 covered nine signal types. They were indexed along two axes: what shape the data has (single value, keyed list, keyed record), and where the value comes from (mutable source, sync derivation, async derivation, external push). That four-by-three grid had gaps. The one that mattered most: you couldn't derive a Store or a List from a Task. Fetch some JSON, key it, and turn it into a reactive list or object — there was no direct path.

So people wrote this instead:

const user = createTask(async () => fetchUser(id.get()))
const store = createStore({ name: '', email: '' })
createEffect(() => store.set(user.get()))

That's an imperative write inside an effect, syncing one signal into another by hand. We've always discouraged this pattern. It duplicates state. The effect also has to run on every change, instead of Cause & Effect propagating only what actually changed. But until now, it was the only door.

What Cause & Effect 1.5 does about it#

Cause & Effect 1.5 backports the fix without breaking anything yet. Three new functions — deriveSignal, deriveList, and deriveStore — replace the old split between createComputed/createMemo and the collection-specific derivation functions. Each one derives a signal of its matching shape from any source: a sync function, an async function, an external push, or another signal. That closes exactly the gap above — a Task can become a Store or a List directly, no manual effect required.

The old names — createComputed, List, Store, createCollection, and others tied to the previous nine-type taxonomy — still work in 1.5. They're marked @deprecated and point at their replacement. Nothing breaks yet. That's the point of a 1.x release. The renames are visible now. The removal happens later, in Cause & Effect 2.0.

What Le Truc 2.5 does about it#

Le Truc 2.5 re-exports every one of those new names next to the deprecated one it replaces. You can start using deriveSignal, deriveList, MutableList, MutableStore, and the rest today, without waiting for a major version bump on either library. Cause & Effect's MIGRATION-2.0.md has the full rename table. A codemod rewrites your imports automatically:

bun tools/codemod-v2.ts 'src/**/*.ts' --module @zeix/le-truc

Run it, review the diff, done. The one rename it can't do for you: createComputed → deriveSignal also renames the value option to initial — that one needs a manual pass.

Why the names had to change#

Two of the renames are more than cosmetic. Today, List means the mutable type — the thing you call .add() and .remove() on. Under the new taxonomy, List becomes the readonly base, and MutableList is the mutable one. Store gets the same flip: createStore used to return Store; from 2.0 it returns MutableStore, and Store becomes the readonly base.

That's a real breaking change, not a rename you can skip. Code that types a variable List<T> and calls .add() on it breaks at compile time — silently wrong until you rebuild, not silently wrong at runtime. Cause & Effect's team considered other options, including keeping List as the mutable name and introducing Collection as the readonly base. They dropped that path: "Collection" reads to most people, and to most language models, as something like a reactive Map, not a keyed list. A confusing name forever seemed worse than a well-flagged breaking rename now.

What's still ahead#

Le Truc 3.0 will follow Cause & Effect 2.0 once it ships. The deprecated re-exports go away, and the type surface collapses from nine names to six. If you migrate to the new names now, that jump is a version bump, not a rewrite. The full reasoning behind the type collapse — including why Collection doesn't survive as a name and why Sensor merges into State — is in Cause & Effect's ADR 0018.

One thing the codemod can't flag for you: isSignal and isMutableSignal keep their names in 2.0 but narrow what they match, from "any reactive value" to "single-value signal only". If you guard an unknown value with either today, check it by hand before you rely on it after the jump. The sooner your imports point at the new names, the smaller Le Truc 3.0 will feel when it arrives.