# Testing and Benchmarks

This document provides an overview of the OWILIX test suite and benchmark results.

## Test Package Structure

```{eval-rst}
.. automodule:: tests.owilix.core
   :members:
   :noindex:
```

### Database Executor Tests

```{eval-rst}
.. automodule:: tests.owilix.core.db
   :members:
   :noindex:
```

### Fsspec Filesystem Tests

```{eval-rst}
.. automodule:: tests.owilix.core.fsspec
   :members:
   :noindex:
```

### Repository Tests

```{eval-rst}
.. automodule:: tests.owilix.core.repository
   :members:
   :noindex:
```

### Manager Tests

```{eval-rst}
.. automodule:: tests.owilix.core.manager
   :members:
   :noindex:
```

## Quick Start

```bash
# Run all unit tests (fast, no network)
uv run pytest tests/owilix/core/ -m "not integration" -v

# Run integration tests (requires LEXIS auth)
uv run pytest tests/owilix/core/ -m integration -v -s

# Run the remote search CLI/UI checks
uv run pytest tests/owilix/cli/test_remote_search_cli.py tests/owilix/core/tasks/test_search.py -q

# Run targeted network smoke checks used in release prep
uv run pytest tests/owilix/cli/test_smoke.py -q -m integration -k "remote_doctor or remote_ls_latest or remote_search"

# Run specific benchmark
uv run python tests/owilix/core/db/benchmark_comprehensive.py
```

## Benchmark Summary

### Async Executor Performance

| Query Type | Sync | Async | Improvement |
|------------|------|-------|-------------|
| COUNT(*) | 25.0s | 13.5s | **+46%** |
| SELECT | 23.6s | 13.2s | **+44%** |
| GROUP BY | 24.9s | 13.4s | **+46%** |
| WHERE filter | 26.6s | 13.4s | **+50%** |
| **Average** | | | **+46%** |

### HTTP Connection Pooling

| Method | Time (20 ops) | vs Baseline |
|--------|---------------|-------------|
| requests (no Session) | 3.95s | baseline |
| requests.Session | 1.52s | **2.6x faster** |
| httpx sync | 1.46s | **2.7x faster** |
| httpx async | 0.38s | **10x faster** |

### Key Findings

- **Use COUNT(\*)** - COUNT(id) is ~13% slower
- **prefetch=2** is optimal for sync executor
- **max_concurrent=10** is optimal for async
- **Async mode** best for batch processing, not interactive use