Choose a pressure eviction policy — Connect the previous skills

PART I

Warm-up

Reconnect with a useful idea before starting choose a pressure eviction policy. In one or two sentences, describe the input, output or host boundary from the previous exercise that this lesson builds on. If you are arriving directly, skim the worked example first; there is no locked prerequisite.

A rough note is enough. You can revise it after the example.

PART II

Instruction & guided example

Connect policy to fairness across mods

An eviction ranking applies only after eligibility is established. Least-recent-use, largest-first and expiry-first policies optimize different goals and can choose different victims. Make the objective explicit, calculate the effect and keep active or recovery-critical resources outside the candidate list.

Exercise cache ledger under a 100 MiB budget: A30 last used 1 minute ago, expires in 20; B20 last used 10 minutes ago, expires in 2; C25 last used 4 minutes ago, expires in 5; D15 active/retained. Usage 90. A/B/C are regenerable and eligible; D is never eligible while its owner is active. Rebuild costs are A60 s, B5 s, C20 s.

Supplied case: Mod M owns A30 and C25; mod N owns B20. Global LRU repeatedly evicts N’s cache after N is idle.

Worked reasoning: The policy may satisfy the shared budget yet impose uneven rebuild costs. Review whether the product contract needs per-mod quotas or a weighted fairness rule. State the desired tradeoff before inventing implementation machinery.

Contrast and diagnosis: Budget correctness alone does not establish fairness or predictable performance for every mod.

Worked reasoning and supplied artifact

// Disposable exercise fixture / host-contract notes; not executable API syntax.
// Supplied input: Mod M owns A30 and C25; mod N owns B20. Global LRU repeatedly evicts N’s cache after N is idle.
// Worked prediction: The policy may satisfy the shared budget yet impose uneven rebuild costs. Review whether the product contract needs per-mod quotas or a weighted fairness rule. State the desired tradeoff before inventing implementation machinery.
// Rejected alternative: Budget correctness alone does not establish fairness or predictable performance for every mod.

Illustrative example from the proposed WillowScript design. The language and WUIX/WMATH/WDB/WAUDIO/WTMP services are not implemented in the supplied source. This site has no interpreter, VM or native host and does not execute or certify the snippet.

Try the guided change

Reconstruct the supplied worked case first: list its inputs, intermediate decisions and final artifact. Compare each row with the explanation and identify exactly where the rejected alternative fails. Then use the changed inputs in Part IV for independent work.

• Write the supplied inputs and initial state before following the example.
• Reconstruct each intermediate calculation or host decision, then compare with the worked artifact.
• Explain why the contrasting alternative fails before attempting the changed inputs in Part IV.

PART III

Take a breather

Step back after the guided example. Look away from the screen, stretch, or reread your prediction. This is an optional learning pause, not a timer or an assessment gate.

When you return, try to name the lesson’s key idea without copying its wording. If that feels difficult, return to Part II and trace one line rather than restarting the entire lesson.

PART IV

Independent practice

Project context: Resource budget dashboard. Build a small, understandable change around choose a pressure eviction policy.

Prepare a two-mod usage/rebuild table and propose one measurable fairness criterion, including a pressure case where global LRU and your rule differ.

What to produce

Prepare a two-mod usage/rebuild table and propose one measurable fairness criterion, including a pressure case where global LRU and your rule differ. Include input → intermediate decision/state → predicted outcome, with any required host rule stated explicitly.

• State the intended behavior in a sentence before editing.
• Use only source-backed syntax and the host services whose contract you have been given.
• Record a normal case and a boundary or unavailable-resource case.
• Label hand predictions separately from anything actually measured in a supplied toolchain.

Syntax colors are a reading aid, not a compiler. The line you are editing stays uncolored until you move to a different line; capitalization is never changed. Drafts stay on this browser and code is not executed.

import tmp;
const cache = tmp.create({
    name: "report-cache",
    storage: "ssd",
    maxSize: "2GB",
    lifetime: { duration: "30m", refreshOnAccess: true },
    cleanup: { automatic: true, onCrashRecovery: true }
});

PART V

Self-check

Use this check to notice what to revisit. It does not award a private course grade, block another page or claim to verify compiled code. Every answer has a learning explanation.

You may keep exploring even if you are unsure. Use the hint, compare the trace, and revise your explanation.

PART VI

Recap & small task

Bring the idea back to the mod feature you are building. A useful recap connects the mechanism to a decision you can make in your own code.

• Connect policy to fairness across mods
• Explain the decision for the supplied case, then independently work the changed fixture.
• Separate source-backed behavior, host requirements and untested predictions.

One small recap task

In one or two sentences, name the decisive rule in “Connect policy to fairness across mods” and predict the outcome of one changed input from your independent fixture. State which host behavior still needs real verification.

This marks your personal practice record only. It is not a certification or an official workspace assessment.