refactor(store): improve __flush_internals performance by pre-sorting derived deps

[Sheraff]

Assuming Derived stores are more often updated than created/destroyed, it should be more performant to pre-compute the deps order from registerOnGraph when we add them to __storeToDerived, than to re-sort every time we call __flush_internals

---

benchmark (tl;dr 3.7x)

function setup() {
  const a = new Store(1)
  const b = new Derived({ deps: [a], fn: () => a.state })
  b.mount()
  const c = new Derived({ deps: [a], fn: () => a.state })
  c.mount()
  const d = new Derived({ deps: [b], fn: () => b.state })
  d.mount()
  const e = new Derived({ deps: [b], fn: () => b.state })
  e.mount()
  const f = new Derived({ deps: [c], fn: () => c.state })
  f.mount()
  const g = new Derived({
    deps: [d, e, f],
    fn: () => d.state + e.state + f.state,
  })
  g.mount()

  g.subscribe(() => noop(g.state))

  return a
}

describe('Derived', () => {
  const a = setup()
  bench('New', () => {
    for (let i = 0; i < 1000; i++)
      a.setState(() => 2)
  })

  const b = setupOld()
  bench('Old', () => {
    for (let i = 0; i < 1000; i++)
      b.setState(() => 2)
  })
})

 ✓  @tanstack/store  tests/foo.bench.ts > Derived 1218ms
     name        hz     min     max    mean     p75     p99    p995    p999     rme  samples
   · New   1,938.60  0.4730  0.6508  0.5158  0.5170  0.5841  0.5978  0.6508  ±0.21%      970
   · Old     514.10  1.7354  4.2685  1.9451  1.9701  2.2839  2.3610  4.2685  ±1.04%      258

 BENCH  Summary

   @tanstack/store  New - tests/foo.bench.ts > Derived
    3.77x faster than Old

---

For completeness, because this PR makes creating a Derived store more expensive than before, we can also do a benchmark of that to assess the damage (tl;dr 1.05x slower). The measured slowdown here seems consistent, but the effect size is so small it could virtually be ignored.

 ✓  @tanstack/store  tests/foo.bench.ts > Derived: setup 1294ms
     name      hz     min      max    mean     p75      p99     p995     p999      rme  samples
   · New   117.94  3.3893  94.4454  8.4791  9.2112  94.4454  94.4454  94.4454  ±43.33%       64
   · Old   123.41  2.7881   111.03  8.1031  3.3671   111.03   111.03   111.03  ±56.50%       62

 BENCH  Summary

   @tanstack/store  Old - tests/foo.bench.ts > Derived: setup
    1.05x faster than New

[nx-cloud]

View your CI Pipeline Execution ↗ for commit ec20d56

Command	Status	Duration	Result
nx affected --targets=test:sherif,test:knip,tes...	✅ Succeeded	47s	View ↗
nx run-many --target=build --exclude=examples/**	✅ Succeeded	14s	View ↗

---

☁️ Nx Cloud last updated this comment at 2025-09-24 16:50:13 UTC

[pkg-pr-new]

@tanstack/angular-store

npm i https://pkg.pr.new/@tanstack/angular-store@236

@tanstack/react-store

npm i https://pkg.pr.new/@tanstack/react-store@236

@tanstack/solid-store

npm i https://pkg.pr.new/@tanstack/solid-store@236

@tanstack/store

npm i https://pkg.pr.new/@tanstack/store@236

@tanstack/svelte-store

npm i https://pkg.pr.new/@tanstack/svelte-store@236

@tanstack/vue-store

npm i https://pkg.pr.new/@tanstack/vue-store@236

commit: ec20d56