card-client

Card Client

Architecture Overview

┌─────────────────────────────────┐
│     Client/Presentation Layer    │
│         ServiceManager           │
└──────────────┬──────────────────┘
               │
┌──────────────▼──────────────────┐
│         Facade Layer            │
│  Single API for all operations  │
└──────────────┬──────────────────┘
               │
        ┌──────┴──────┬──────────┐
        │             │          │
┌───────▼──────┐  ┌───▼────┐  ┌──▼───┐
│   Catalog    │  │ Tracker│  │Router│
│ (Data Layer) │  │        │  │      │
└───────┬──────┘  └────────┘  └──────┘
        │
   ┌────┴────────────┬───────────┐
   │                 │           │
┌──▼───────────┐  ┌──▼────────┐  │
│Record Store  │  │Blob Store │  │
│  (SQLite)    │  │ (Pebble)  │  │
└──────────────┘  └───────────┘  │

Key Components

Core Services

• Facade (/facade) - High-level API that orchestrates all operations

  • VaultManager - Manages vault lifecycle with RWMutex protection
  • QueryService - Read operations
  • SyncService - Import/update operations
  • ExportService - Card export with templating
  • FavoriteService - Favorite management
  • CacheManager - Operation result caching

• Catalog (/store/catalog) - Unified interface for data operations

  • Wraps Record Store (metadata) and Blob Store (binary files)
  • Ensures transactional consistency between stores

• Dual Storage Strategy

  • SQLite (/store/record/erecord) - Structured metadata with Ent ORM
  • Pebble KV (/store/blob/pblob) - Binary card files with versioning

Project Structure

card-client/
├── client/          # Application initialization
├── facade/          # High-level API facade
├── store/           # Data access layer
│   ├── resource/    # Domain models (Record, Filter, etc.)
│   ├── catalog/     # Unified store interface
│   ├── record/      # Record storage abstraction
│   └── blob/        # Binary storage abstraction
├── library/         # Multi-vault management
├── preferences/     # Configuration (Viper-based)
├── credentials/     # Secure credential handling (keyring)
├── operation/       # Async operation tracking
├── tracker/         # Concurrency control
└── generate/        # Code generation tools

Stack

• Go 1.26.0 - Core language
• Ent ORM (v0.14.5) - SQL schema generation and queries
• SQLite - Record metadata storage
• Pebble (v2.1.4) - High-performance KV store for binary files
• Viper - Configuration management
• ZeroLog - Structured logging

Architecture Highlights

1. Hybrid Storage Design

• SQLite for queryable metadata (fast searches, transactions)
• Pebble KV for binary files (efficient versioning, fast access)
• Prevents large blobs from impacting query performance

2. Concurrency Model

• Vault-level RWMutex: Exclusive writes, concurrent reads
• Per-card mutex tracking: Fine-grained locking for updates
• Context propagation: Cancellable long-running operations
• Async operations: Goroutines for imports/exports

3. Vault System

• Multiple isolated vaults (separate databases)
• Lazy loading (only when accessed)
• Hot-swappable without restart
• Each vault has its own SQLite + Pebble databases

4. Version Tracking

• Keeps multiple versions of each card (default: 5)
• Composite keys: RID (Record ID) + timestamp
• Enables history and rollback capabilities

5. Operation Tracking

• In-memory registry of ongoing operations
• Progress reporting with success/failure metrics
• Context-based cancellation support
• Nano ID generation with UUID fallback

Important Gotchas

1. Code Generation

• Ent ORM generates code - Don't edit files in /store/record/erecord/ent/
• Post-processing required - Custom ID types (RID, CID) need manual fixes
• Run go generate ./generate after schema changes

2. Transaction Handling

• Nested transactions - Catalog coordinates Record + Blob transactions
• Context propagation - Always pass context for proper transaction scope
• Defer cleanup - Use defer for mutex unlocks and transaction rollbacks

3. Concurrency Pitfalls

• Always lock vault before operations (RLock for reads, Lock for writes)
• Use tracker for per-card operations to prevent race conditions
• Defer unlocks to ensure cleanup even on panics
• One vault active at a time per ServiceManager

4. Storage Limitations

• SQLite connection pool - Configure max connections carefully
• Pebble batch size - Large imports may need batching
• Version retention - Old versions auto-pruned (keep max 5)
• FTS indexing - Full-text search requires explicit UpsertFTS calls

5. Error Handling

• Check integrity - Use Record.Integrity() to validate completeness
• Sync status - Track SyncStatus enum (unchanged/updated/failed)
• Operation reports - Always check success/failure counts
• Context cancellation - Handle context.Canceled gracefully

6. Performance Considerations

• Lazy loading - Vaults only loaded when needed
• Batch operations - Use transactions for bulk inserts
• Index usage - SQLite indexes on source, URLs, timestamps
• Thumbnail generation - On-the-fly from PNG content (may be slow)

7. Security Notes

• Credentials - Stored in system keyring, not in config files
• Path traversal - Validate export paths to prevent directory escape

Development Tips

Adding New Storage Backend

1. Implement record.Builder or blob.Builder interface
2. Register in library's storage maps
3. Update Manifest to support new storage type

Adding New Facade Service

1. Create service struct in /facade
2. Add to Facade composite struct
3. Wire in ServiceManager initialization

Debugging Concurrency Issues

• Enable debug logging for mutex operations
• Use race detector: go run -race
• Check operation registry for stuck operations
• Monitor goroutine count for leaks

Testing Considerations

• Each test can use separate vault
• Use enttest package for test database setup
• Mock external dependencies (router, fetcher)
• Clean up test vaults to prevent disk bloat

Common Operations Flow

Import Cards

URLs → Validate → Fetch metadata → Check duplicates →
Insert record → Store PNG → Update FTS → Report

Update Cards

Lock card → Fetch latest → Compare versions →
Update if changed → Store new version → Unlock → Report

Query Cards

RLock vault → Build filter → SQL query →
Transform results → RUnlock → Return Box[T]

Build & Run

# Generate Ent code
go generate ./generate

# Set sqlite FTS5 flag
export GOFLAGS="-tags=sqlite_fts5"

# Build
go build ./...

# Run tests
go test ./...

# Run with race detector (development)
go run -race .

Dependencies

See go.mod for full dependency list. Key external packages:

• github.com/digital-foxy/card-fetcher - Protocol handlers
• github.com/digital-foxy/card-parser - PNG parsing
• github.com/digital-foxy/toolkit - Shared utilities
• entgo.io/ent - ORM framework
• cockroachdb/pebble - KV store