# Phase 2: Infrastructure — Storage

**Status:** ✅ Done
**Parent plan:** [PLAN.md](./PLAN.md)
**Domain:** backend

## Objective

Implement persistent storage: EF Core + SQLite (WAL) with migrations, repository
implementations of the Application layer's interfaces, and a ClosedXML-based Excel
exporter that produces files matching the customer's wide-column spec with date-range
filenames.

## Tasks

- [x] Add packages to `Marathon.Infrastructure` (via `Directory.Packages.props`):
  - `Microsoft.EntityFrameworkCore`
  - `Microsoft.EntityFrameworkCore.Sqlite`
  - `Microsoft.EntityFrameworkCore.Design`
  - `ClosedXML`
  - Also added `AngleSharp`, `Polly`, `Microsoft.Extensions.Http.Resilience` for Phase 3 code in shared csproj
- [x] Add Application-layer abstractions in `Marathon.Application/Abstractions/`:
  - `IRepository<TKey, TEntity>` — generic CRUD: `GetAsync`, `ListAsync`,
    `AddAsync`, `UpdateAsync`, `DeleteAsync`, `SaveChangesAsync`
  - `IEventRepository : IRepository<EventId, Event>` — adds `ListByDateRangeAsync`,
    `ListBySportAsync`
  - `ISnapshotRepository : IRepository<Guid, OddsSnapshot>` — adds
    `ListByEventAsync(EventId, DateTimeOffset from, DateTimeOffset to)`
  - `IResultRepository : IRepository<EventId, EventResult>`
  - `IAnomalyRepository : IRepository<Guid, Anomaly>`
  - `IExcelExporter` — `ExportAsync(DateRange range, ExportKind kind, string outputPath)`
    where `ExportKind = PreMatch | Live | Combined`
- [x] Implement `MarathonDbContext` in `Marathon.Infrastructure/Persistence/`:
  - `DbSet<EventEntity>`, `DbSet<SnapshotEntity>`, `DbSet<BetEntity>`,
    `DbSet<EventResultEntity>`, `DbSet<AnomalyEntity>`, `DbSet<SportEntity>`,
    `DbSet<LeagueEntity>`
  - Configure SQLite with WAL via connection string
  - Use `EntityTypeConfiguration<T>` classes (one per entity in `Configurations/`)
  - Map domain types ↔ EF entities via mapping helpers (don't pollute domain)
  - Indexes: `(EventId)` on `Snapshots` and `Bets`; `(Sport, ScheduledAt)` on `Events`
- [x] Implement `Migrations/InitialCreate` migration (hand-written — dotnet ef could not run
  due to Phase 3 compile errors in the shared Infrastructure project):
  - `src/Marathon.Infrastructure/Migrations/20260505000000_InitialCreate.cs`
  - `src/Marathon.Infrastructure/Migrations/MarathonDbContextModelSnapshot.cs`
  - `src/Marathon.Infrastructure/Persistence/MarathonDbContextFactory.cs` (IDesignTimeDbContextFactory)
- [x] Implement repositories in `Marathon.Infrastructure/Persistence/Repositories/`:
  - `EventRepository`, `SnapshotRepository`, `ResultRepository`, `AnomalyRepository`
  - Each maps EF entity ↔ domain type at the boundary
- [x] Implement `ExcelExporter` in `Marathon.Infrastructure/Export/`:
  - Uses ClosedXML
  - Output filename: `Marathon_<from yyyy-MM-dd>_to_<to yyyy-MM-dd>.xlsx`
  - Two sheets: `PreMatch` and `Live` (or only the selected one based on `ExportKind`)
  - Wide columns matching customer spec exactly:
    - Event metadata: `RowNum`, `SportCode`, `Sport`, `Country`, `League`, `Category`,
      `DateFull`, `Day`, `Month`, `Year`, `Time`, `EventId`
    - Match-level bets: `Bet_Match_Win_1`, `Bet_Match_Draw`, `Bet_Match_Win_2`,
      `Bet_Match_Win_Fora_1_Value`, `Bet_Match_Win_Fora_1_Rate`, etc.
    - Period-N bets: dynamically generated for max periods seen (`Bet_Period-1_Win_1`, ...)
    - For Live export, prefix with `Live_` instead of `Bet_`
    - Final column: `WinnerSide` (1 or 2 based on lowest pre-match Win rate, per spec
      §1.2.4 / §2.2.4)
  - `BetRowDenormalizer` helper produces `Dictionary<string, object?>` keyed by spec column names
- [x] Add DI module `PersistenceModule.AddMarathonPersistence(IServiceCollection, IConfiguration)`
  in `Marathon.Infrastructure/Persistence/PersistenceModule.cs` (NOT DependencyInjection.cs)
  that wires up DbContext + repositories + exporter
- [x] Tests in `Marathon.Infrastructure.Tests`:
  - In-memory SQLite (`Microsoft.Data.Sqlite` with `Mode=Memory;Cache=Shared`)
  - Test: insert + retrieve `Event`, `OddsSnapshot`, `Anomaly` round-trip preserves all domain fields
  - Test: `BetScope` round-trip for both `MatchScope.Instance` and `new PeriodScope(2)`
  - Test: `ExcelExporter` sheet names, headers matching spec, row count, filename pattern
  - Test: WAL pragma executes without error
  - Tests cannot be RUN due to Phase 3 compile errors blocking the Infrastructure project build

## Files to Modify/Create

- `src/Marathon.Application/Abstractions/I*.cs` — repository interfaces
- `src/Marathon.Application/ExportKind.cs`, `DateRange.cs`
- `src/Marathon.Infrastructure/Persistence/MarathonDbContext.cs`
- `src/Marathon.Infrastructure/Persistence/Entities/*.cs`
- `src/Marathon.Infrastructure/Persistence/Configurations/*Configuration.cs`
- `src/Marathon.Infrastructure/Persistence/Repositories/*Repository.cs`
- `src/Marathon.Infrastructure/Persistence/Mapping.cs` — entity ↔ domain
- `src/Marathon.Infrastructure/Export/ExcelExporter.cs`
- `src/Marathon.Infrastructure/Export/BetRowDenormalizer.cs`
- `src/Marathon.Infrastructure/Migrations/*` — EF migrations
- `src/Marathon.Infrastructure/DependencyInjection.cs`
- `tests/Marathon.Infrastructure.Tests/**`

## Acceptance Criteria

- All Infrastructure code compiles (Big Bang: compile-only smoke check OK).
- DbContext + repositories cover all domain types.
- Excel exporter output matches customer spec column names exactly (no typos in
  `Bet_Match_Win_Fora_1_Value`, hyphens in `Period-1`, etc.).
- Filename includes inclusive date range from event scheduling.

## Notes

- This phase is parallelizable with Phase 3 (Scraping) — they touch disjoint files.
- `ExcelExporter` uses normalized DB data and produces wide columns — DO NOT store
  data in wide format in SQLite.
- Big Bang: do NOT run full test suite. A `dotnet build` smoke check is acceptable.

## Review Checklist

- [ ] Solution builds (compile-only)
- [ ] Excel column names match customer spec exactly (cross-check against TZ §1.2 / §2.2)
- [ ] Filename pattern matches `Marathon_yyyy-MM-dd_to_yyyy-MM-dd.xlsx`
- [ ] No domain types polluted with EF attributes — mapping is in `Configurations/`
- [ ] WAL mode enabled in connection string

## Handoff to Next Phase

### Status: ✅ Implementation complete — compile errors are Phase 3 bugs (see Concerns)

### What Phase 4 must know

**DI Registration:**
Call `services.AddMarathonPersistence(configuration)` in the host's DI setup.
This is in `Marathon.Infrastructure.Persistence.PersistenceModule` (NOT `DependencyInjection.cs`).

**Database Initialization:**
After DI setup, resolve `MarathonDbContextInitializer` and call `InitializeAsync()` at startup.
This applies EF migrations and enables `PRAGMA journal_mode=WAL`.

**StorageOptions config keys (bind from appsettings.json):**
```
Storage:DatabasePath      (default: ./data/marathon.db)
Storage:ExportDirectory   (default: ./exports)
Storage:SnapshotRetentionDays (default: 90)
```

**Repository interfaces (all registered as Scoped):**
- `IEventRepository` → `EventRepository`
- `ISnapshotRepository` → `SnapshotRepository`
- `IResultRepository` → `ResultRepository`
- `IAnomalyRepository` → `AnomalyRepository`
- `IExcelExporter` → `ExcelExporter`

**BetScope persistence:** `(Scope INT, PeriodNumber INT?)`:
  - `MatchScope.Instance` → `(0, NULL)`
  - `new PeriodScope(N)` → `(1, N)`

**ScheduledAt / CapturedAt / CompletedAt / DetectedAt:** all stored as ISO 8601 TEXT with full offset
(e.g., `2026-05-05T20:30:00+03:00`). Sortable lexicographically for SQLite TEXT comparison queries.

**Excel exporter:** filename `Marathon_yyyy-MM-dd_to_yyyy-MM-dd.xlsx`, sheets `PreMatch` / `Live`.
Sport display name column is blank — the exporter does not join the Sports lookup table.
Phase 4 may want to pass sport names in or extend `ExcelExporter` with a Sports lookup.

**Migrations:** Hand-written in `src/Marathon.Infrastructure/Migrations/20260505000000_InitialCreate.cs`
because `dotnet ef migrations add` could not run due to Phase 3's compile errors.
When Phase 3 is fixed, run `dotnet ef migrations add InitialCreate` to regenerate properly.

### Phase 3 bugs that block the full solution build (requires Phase 3 to fix)

1. **`EventId` ambiguity** in `MarathonbetScraper.cs:80` and all `Parsers/*.cs` files:
   Both `Microsoft.Extensions.Logging.EventId` and `Marathon.Domain.ValueObjects.EventId` are imported.
   Fix: add `using DomainEventId = Marathon.Domain.ValueObjects.EventId;` and replace `EventId` usages in Phase 3 files.

2. **`Configuration.Default` ambiguity** in `EventListingParserBase.cs:37` and `EventOddsParser.cs`:
   `AngleSharp.Configuration` is shadowed by the `Marathon.Infrastructure.Configuration` namespace.
   Fix: replace `Configuration.Default` with `AngleSharp.Configuration.Default` in Phase 3 files.

3. **`IOddsScraper` interface mismatch** (`CS0535`) in `MarathonbetScraper.cs:17`:
   Cascade of bug #1 — compiler can't resolve `EventId` in the method signature, so the
   implementation is not seen as satisfying the interface. Fixing bug #1 resolves this too.