# Database Migration Design

## Overview

This document describes the approach for migrating legacy SQL Server objects to DbUp migration scripts.

## Script Naming Convention

Scripts follow the pattern: `NNN_Description.sql`

```
001_CreateSearchTable.sql           (existing)
002_CreateDataUpdateTable.sql       (existing)
003_CreateBranchTable.sql
...
025_CreateLotUsageHistTable.sql
026_CreateWorkOrderView.sql
...
032_CreateLastDataUpdatesView.sql
033_CreateWorkOrderFilterParameterType.sql
...
039_CreateItemOperationMisFilterParameterType.sql
040_CreateSubmitSearchProcedure.sql
...
043_CreateResetPartialSearchesProcedure.sql
044_CreateMatchMisFunction.sql
```

### Number Ranges

| Range | Category | Count |
|-------|----------|-------|
| 001-002 | Already done (Search, DataUpdate) | 2 |
| 003-025 | Tables | 23 |
| 026-032 | Views | 7 |
| 033-039 | Table-valued parameter types | 7 |
| 040-043 | Stored procedures | 4 |
| 044 | Functions | 1 |

## Execution Order

Objects must be created in dependency order:

```
1. Reference tables (no FK dependencies)
   - Branch, StatusCode, FunctionCode, ProfitCenter, WorkCenter
   - Item, JdeUser, OrgHierarchy, RouteMaster, MisData

2. Core tables (depend on reference tables)
   - Lot, LotLocation, WorkOrder_Curr, WorkOrder_Hist
   - WorkOrderStep_Curr, WorkOrderStep_Hist
   - WorkOrderTime_Curr, WorkOrderTime_Hist
   - WorkOrderComponent_Curr, WorkOrderComponent_Hist
   - LotUsage_Curr, LotUsage_Hist

3. Views (depend on tables)
   - Union views: WorkOrder, WorkOrderStep, WorkOrderTime, WorkOrderComponent, LotUsage
   - Aggregation views: WorkOrderTotalScrap, LastDataUpdates

4. Types (no dependencies, but used by procedures)
   - All 7 TVP types

5. Procedures and Functions (depend on tables, views, types)
   - SubmitSearch, StartSearch, CompleteSearch, ResetPartialSearches
   - MatchMis function
```

## Schema Mapping

### Data Type Decisions

| Legacy | New | Rationale |
|--------|-----|-----------|
| DATETIME | DATETIME2(7) | Better precision, recommended for new development |
| VARCHAR | VARCHAR | Keep as-is for JDE/CMS compatibility |
| NVARCHAR | NVARCHAR | Keep as-is |
| VARBINARY(MAX) | VARBINARY(MAX) | Keep for Excel storage (per user decision) |

**Note:** The DATETIME → DATETIME2(7) conversion requires updating the database-schema spec to reflect this decision. This is a deliberate modernization choice.

### Index Strategy

- Primary keys: Defined in table creation scripts
- Foreign keys: NOT created (legacy doesn't have them, cache tables)
- Clustered indexes: On primary keys
- Non-clustered indexes: Included in table creation scripts (match legacy exactly)

## DbUp Configuration

The existing `DatabaseMigrator.cs` configuration is appropriate:
- Uses `WithTransaction()` for atomic migrations
- Uses `WithScriptsEmbeddedInAssembly()` for embedded resources
- Uses `EnsureDatabase.For.SqlDatabase()` to create DB if needed

No changes needed to the migrator itself.

## Verification Approach

1. **Script syntax**: Run against local SQL Server container
2. **Object existence**: Query sys.tables, sys.views, sys.procedures
3. **Schema accuracy**: Compare column definitions to legacy
4. **Codex review**: Cross-reference with specs

## File Structure

```
NEW/src/JdeScoping.Database/
├── Scripts/
│   ├── 001_CreateSearchTable.sql          (existing)
│   ├── 002_CreateDataUpdateTable.sql      (existing)
│   ├── 003_CreateBranchTable.sql          (new)
│   ├── ...
│   └── 046_CreateMatchMisFunction.sql     (new)
├── DatabaseMigrator.cs                     (existing, no changes)
└── JdeScoping.Database.csproj             (existing, no changes)
```