don
/
neon-sprawl
mirror of https://github.com/ViPro-Technologies/neon-sprawl
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
neon-sprawl/docs/plans/NEO-127-implementation-plan.md

14 KiB
Raw Permalink Blame History

NEO-127 — Implementation plan

Story reference

Field Value
Key NEO-127
Title E7M2-04: RewardRouterOperations (apply QuestRewardBundle)
Linear https://linear.app/neon-sprawl/issue/NEO-127/e7m2-04-rewardrouteroperations-apply-questrewardbundle
Module E7.M2 — RewardAndUnlockRouter · Epic 7 Slice 2 · backlog E7M2-04
Branch NEO-127-e7m2-reward-router-operations
Precursor NEO-125 Done — catalog load + bundle validation · NEO-126 Done — IRewardDeliveryStore + RewardDeliveryEvent
Pattern NEO-105 — EncounterCompletionOperations simulate-all + sequential apply + compensating rollback; NEO-107 — store TryRecord after grants commit
Blocks NEO-128 (QuestStateOperations.TryMarkComplete wiring)
Client counterpart none — server engine (E7M2-prototype-backlog)

Kickoff clarifications

No clarifications needed. Linear AC, E7M2-04 backlog, E7M2 kickoff decisions, and landed NEO-105/NEO-126 code settle:

  • Namespace Game/Rewards/ — co-locate router with IRewardDeliveryStore (NEO-126); bundle DTOs stay on QuestDefRow in Game/Quests/ (NEO-125).
  • Grant order — item grants first, skill XP second; compensating item rollback on inventory or skill-apply failure (NEO-105 precedent). Prototype bundles have ≤1 skill row; no XP rollback helper exists.
  • Skill XP path — call SkillProgressionGrantOperations.TryApplyGrant with MissionRewardSkillXpConstants.MissionRewardSourceKind (not the fire-and-forget MissionRewardSkillXpGrant wrapper) so deny envelopes map to router reason codes.
  • Idempotent replay — IRewardDeliveryStore.TryGet before apply; return success with existing RewardDeliveryEvent without re-granting (backlog “idempotent replay via store”).
  • Empty bundle — null or both-empty grant lists → success with empty snapshots and TryRecord (NEO-126 allows empty arrays; CI guarantees all four prototype quests have bundles).
  • Store race — if grants succeed but TryRecord returns false, rollback this call's item + skill XP grants and return the winner's store event (mirror EncounterCompletionOperations when TryMarkCompleted loses); do not retain duplicate grants on the loser.

Goal, scope, and out-of-scope

Goal: Server-internal apply for quest completionRewardBundle rows (items + skill XP) with structured outcomes and durable idempotency via IRewardDeliveryStore.

In scope (from Linear + E7M2-04):

  • RewardRouterOperations.TryDeliverQuestCompletion in server/NeonSprawl.Server/Game/Rewards/.
  • Item grants via PlayerInventoryOperations (simulate-all pre-flight, sequential apply, compensating rollback).
  • Skill XP via SkillProgressionGrantOperations.TryApplyGrant with fixed mission_reward source kind.
  • RewardDeliveryReasonCodes + RewardDeliveryResult (success, deny reason, grant snapshots, RewardDeliveryEvent?).
  • Unit + integration tests (AAA): happy path (skill-only + item+skill bundles), inventory deny, idempotent store replay, apply-phase item rollback, skill deny passthrough.
  • server/README.md router section.

Out of scope (from Linear):

  • Quest state transition hooks (NEO-128 E7M2-05).
  • HTTP DTOs / completionRewardSummary (NEO-129 E7M2-06).
  • E9.M1 telemetry hook comments (NEO-130 E7M2-07).
  • Godot client, Bruno (optional on NEO-128).

Acceptance criteria checklist

  • Bundle rows apply atomically where possible; compensating rollback on partial failure before store record.
  • Skill XP uses mission_reward source kind only.

Implementation reconciliation (shipped)

  • Operations: RewardRouterOperations.TryDeliverQuestCompletion — item pre-flight + apply, skill XP via TryApplyGrant (mission_reward), compensating item rollback + per-grant perk tracking + TryRevertAppliedSkillGrants on TryRecord race loss, IRewardDeliveryStore.TryRecord.
  • Types: RewardDeliveryResult, RewardDeliveryReasonCodes.
  • Tests: 14 AAA tests in RewardRouterOperationsTests (…, TryRecord race rollback, concurrent race no double-grant, perk revert on simulated race loss, concurrent level-up winner perk preserved).
  • Docs: server/README.md reward router section; E7.M2 module anchor + alignment register updated.

Technical approach

1. Result + reason types (Game/Rewards/)

RewardDeliveryReasonCodes — stable strings for router denies:

Code When
already_delivered IRewardDeliveryStore.TryGet hit before apply (idempotent replay — success path, not a deny).
inventory_store_missing Inventory snapshot unavailable.
inventory_full, invalid_item, invalid_quantity Passthrough from PlayerInventoryReasonCodes via PlayerInventoryOperations.
unknown_skill, invalid_amount, source_kind_not_allowed Passthrough from SkillProgressionSnapshotApi reason constants via TryApplyGrant.
progression_store_missing SkillProgressionGrantApplyKind.ProgressionStoreMissing.

RewardDeliveryResult — readonly record:

Field Role
Success Grants committed (includes idempotent replay).
ReasonCode Set when Success is false.
GrantedItems Applied item snapshot (RewardItemGrantApplied).
GrantedSkillXp Applied skill XP snapshot (RewardSkillXpGrantApplied).
DeliveryEvent Populated on success (fresh or replay).
DeliveredAt From event on success.

2. RewardRouterOperations.TryDeliverQuestCompletion

Signature (static ops, mirror encounter completion):

public static RewardDeliveryResult TryDeliverQuestCompletion(
    string playerId,
    string questId,
    QuestRewardBundleRow? bundle,
    IItemDefinitionRegistry itemRegistry,
    IPlayerInventoryStore inventoryStore,
    ISkillDefinitionRegistry skillRegistry,
    IPlayerSkillProgressionStore skillProgressionStore,
    ISkillLevelCurve levelCurve,
    PerkUnlockEngine perkUnlockEngine,
    IPlayerPerkStateStore perkStore,
    IRewardDeliveryStore deliveryStore,
    TimeProvider timeProvider)

Flow:

  1. Normalize ids via RewardDeliveryIds; fail closed on empty player/quest id (deny with appropriate code or empty reason — match store fail-closed).
  2. Idempotent replay — if deliveryStore.TryGet, return Success: true with existing event and grant snapshots (no inventory/skill mutation).
  3. Resolve grants — treat null bundle as empty item + skill lists.
  4. Item pre-flight — clone inventory snapshot; TrySimulateAddStack for each bundle.ItemGrants row in catalog order.
  5. Item apply — foreach row TryAddStack; accumulate RewardItemGrantApplied; on deny/store-missing, return deny without skill apply.
  6. Skill XP apply — foreach bundle.SkillXpGrants row, TryApplyGrant(playerId, skillId, amount, MissionRewardSourceKind, …):
    • On Denied, CompensatingRemoveItems for applied item rows and return mapped ReasonCode from outcome.Response!.ReasonCode.
    • On ProgressionStoreMissing, compensating item rollback + deny.
    • On Granted, append RewardSkillXpGrantApplied.
  7. Record — build RewardDeliveryEvent with DeliveredAt = timeProvider.GetUtcNow(), IdempotencyKey = RewardDeliveryIds.MakeIdempotencyKey(...), grant snapshots.
  8. deliveryStore.TryRecord — if true, return fresh success; if false (race), compensating rollback of this call's item + skill XP grants, then return TryGet winner event with already_delivered (or deny if store miss).

Compensating item rollback — private helper mirroring EncounterCompletionOperations.CompensatingRemoveGrants: reverse-order TryRemoveStack for applied item rows only (no skill XP undo).

Compensating skill progression rollback — track perks unlocked per skill grant (CollectPerksUnlockedSince); on TryRecord race loss, TryRevertAppliedSkillGrants reverses each applied skill XP row via negative delta and removes only perks this call unlocked, then ReevaluateAfterLevelUp at post-rollback levels re-applies path-auto perks the winner's XP still qualifies for (avoids full snapshot restore clobbering a concurrent winner).

3. Skill XP source kind enforcement

Always pass MissionRewardSkillXpConstants.MissionRewardSourceKind at apply time — content rows do not carry sourceKind (NEO-124/NEO-43). Startup catalog gates already require skills to allow mission_reward; router still maps source_kind_not_allowed if a deny occurs.

4. Tests (RewardRouterOperationsTests)

Use InMemoryWebApplicationFactory + resolved DI (same pattern as EncounterCompletionOperationsTests):

Test Coverage
TryDeliverQuestCompletion_ShouldApplyGatherIntroSkillXp_WhenBundleIsSkillOnly prototype_quest_gather_intro-shaped bundle → salvage +25; store recorded; replay via second call returns same event without double XP.
TryDeliverQuestCompletion_ShouldApplyItemAndSkillXp_WhenOperatorChainBundle survey_drone_kit ×1 + salvage +50; inventory + progression assertions.
TryDeliverQuestCompletion_ShouldDenyInventoryFull_WithoutStoreRecord Fill bag; deny inventory_full; TryGet false.
TryDeliverQuestCompletion_ShouldRollbackItems_WhenSkillApplyDenied Item grant succeeds then skill deny (inject invalid amount or use test-only bundle row) → item removed; no store record.
TryDeliverQuestCompletion_ShouldReturnExistingEvent_WhenAlreadyDelivered Pre-seed IRewardDeliveryStore; call router → success, no inventory/XP change.
TryDeliverQuestCompletion_ShouldUseMissionRewardSourceKind Assert grant succeeds for catalog skill; optional negative with skill lacking allowlist (unit-level bundle row).

Optional host smoke: resolve RewardRouterOperations is static — assert factory provides all injected stores/registries used in one happy-path call.

5. Docs

  • server/README.md — new Reward router (NEO-127) section: operation entry point, reason codes, apply order, idempotency + store interaction, pointer to NEO-128 wiring.
  • documentation_and_implementation_alignment.md — E7.M2 row: router apply landed (NEO-127).
  • E7_M2_RewardAndUnlockRouter.md — implementation anchor for RewardRouterOperations.

Files to add

Path Purpose
server/NeonSprawl.Server/Game/Rewards/RewardDeliveryReasonCodes.cs Stable router deny reason strings.
server/NeonSprawl.Server/Game/Rewards/RewardDeliveryResult.cs Structured outcome + grant snapshots + optional event.
server/NeonSprawl.Server/Game/Rewards/RewardRouterOperations.cs TryDeliverQuestCompletion apply + rollback + store record.
server/NeonSprawl.Server.Tests/Game/Rewards/RewardRouterOperationsTests.cs AAA happy path, deny, rollback, idempotent replay tests.
docs/plans/NEO-127-implementation-plan.md This plan.

Files to modify

Path Rationale
server/README.md Document RewardRouterOperations, reason codes, idempotency flow.
docs/decomposition/modules/documentation_and_implementation_alignment.md E7.M2 alignment — router apply (NEO-127).
docs/decomposition/modules/E7_M2_RewardAndUnlockRouter.md Add NEO-127 implementation anchor after delivery store bullet.

Tests

Target Coverage
server/NeonSprawl.Server.Tests/Game/Rewards/RewardRouterOperationsTests.cs Skill-only + item+skill happy paths; inventory_full fail-closed; item compensating rollback on skill deny; store idempotent replay; mission_reward apply path.
Optional — extend existing host/DI test Factory resolves all router dependencies (smoke only).
Manual verification dotnet test server/NeonSprawl.Server.Tests --filter RewardRouterOperationsTests; dev server boot unchanged until NEO-128 wires quest complete.

Open questions / risks

Question or risk Agent recommendation Status
Multi-row skill XP partial apply Prototype bundles have ≤1 skill row; implement sequential apply with item rollback on any skill deny; defer XP compensating rollback unless a future story requires multi-row bundles. adopted
MissionRewardSkillXpGrant vs TryApplyGrant Use TryApplyGrant directly for deny envelopes; keep MissionRewardSkillXpGrant as HTTP/convenience wrapper only. adopted
Null bundle on future quests Treat as empty grants + record delivery for idempotency; NEO-128 may skip router call when bundle absent — both safe. adopted
Quest-state rollback on deny NEO-128 scope per E7M2 fail-closed decision; NEO-127 returns structured deny only. deferred

None beyond the above.