# PPF Implementation Tasks

## Legend

```
[ ] Not started
[~] In progress
[x] Completed
[!] Blocked/needs discussion
```

---

## Immediate Priority (Next Sprint)

### [x] 1. Unify _known_proxies Cache

**Completed.** Added `init_known_proxies()`, `add_known_proxies()`, `is_known_proxy()`
to fetch.py. Updated ppf.py to use these functions instead of local cache.

---

### [x] 2. Graceful SQLite Error Handling

**Completed.** mysqlite.py now retries on "locked" errors with exponential backoff.

---

### [x] 3. Enable SQLite WAL Mode

**Completed.** mysqlite.py enables WAL mode and NORMAL synchronous on init.

---

### [x] 4. Batch Database Inserts

**Completed.** dbs.py uses executemany() for batch inserts.

---

### [x] 5. Add Database Indexes

**Completed.** dbs.py creates indexes on failed, tested, proto, error, check_time.

---

## Short Term (This Month)

### [x] 6. Log Level Filtering

**Completed.** Added log level filtering with -q/--quiet and -v/--verbose CLI flags.
- misc.py: LOG_LEVELS dict, set_log_level(), get_log_level()
- config.py: Added -q/--quiet and -v/--verbose arguments
- Log levels: debug=0, info=1, warn=2, error=3
- --quiet: only show warn/error
- --verbose: show debug messages

---

### [x] 7. Connection Timeout Standardization

**Completed.** Added timeout_connect and timeout_read to [common] section in config.py.

---

### [x] 8. Failure Categorization

**Completed.** Added failure categorization for proxy errors.
- misc.py: categorize_error() function, FAIL_* constants
- Categories: timeout, refused, auth, unreachable, dns, ssl, closed, proxy, other
- proxywatchd.py: Stats.record() now accepts category parameter
- Stats.report() shows failure breakdown by category
- ProxyTestState.evaluate() returns (success, category) tuple

---

### [x] 9. Priority Queue for Proxy Testing

**Completed.** Added priority-based job scheduling for proxy tests.
- PriorityJobQueue class with heap-based ordering
- calculate_priority() assigns priority 0-4 based on proxy state
- Priority 0: New proxies (never tested)
- Priority 1: Working proxies (no failures)
- Priority 2: Low fail count (< 3)
- Priority 3-4: Medium/high fail count
- Integrated into prepare_jobs() for automatic prioritization

---

### [x] 10. Periodic Statistics Output

**Completed.** Added Stats class to proxywatchd.py with record(), should_report(),
and report() methods. Integrated into main loop with configurable stats_interval.

---

## Medium Term (Next Quarter)

### [x] 11. Tor Connection Pooling

**Completed.** Added connection pooling with worker-Tor affinity and health monitoring.
- connection_pool.py: TorHostState class tracks per-host health, latency, backoff
- connection_pool.py: TorConnectionPool with worker affinity, warmup, statistics
- proxywatchd.py: Workers get consistent Tor host assignment for circuit reuse
- Success/failure tracking with exponential backoff (5s, 10s, 20s, 40s, max 60s)
- Latency tracking with rolling averages
- Pool status reported alongside periodic stats

---

### [x] 12. Dynamic Thread Scaling

**Completed.** Added dynamic thread scaling based on queue depth and success rate.
- ThreadScaler class in proxywatchd.py with should_scale(), status_line()
- Scales up when queue is deep (2x target) and success rate > 10%
- Scales down when queue is shallow or success rate drops
- Min/max threads derived from config.watchd.threads (1/4x to 2x)
- 30-second cooldown between scaling decisions
- _spawn_thread(), _remove_thread(), _adjust_threads() helper methods
- Scaler status reported alongside periodic stats

---

### [x] 13. Latency Tracking

**Completed.** Added per-proxy latency tracking with exponential moving average.
- dbs.py: avg_latency, latency_samples columns added to proxylist schema
- dbs.py: _migrate_latency_columns() for backward-compatible migration
- dbs.py: update_proxy_latency() with EMA (alpha = 2/(samples+1))
- proxywatchd.py: ProxyTestState.last_latency_ms field
- proxywatchd.py: evaluate() calculates average latency from successful tests
- proxywatchd.py: submit_collected() records latency for passing proxies

---

### [x] 14. Export Functionality

**Completed.** Added export.py CLI tool for exporting working proxies.
- Formats: txt (default), json, csv, len (length-prefixed)
- Filters: --proto, --country, --anonymity, --max-latency
- Options: --sort (latency, added, tested, success), --limit, --pretty
- Output: stdout or --output file
- Usage: `python export.py --proto http --country US --sort latency --limit 100`

---

### [ ] 15. Unit Test Infrastructure

**Problem:** No automated tests. Changes can break existing functionality silently.

**Implementation:**
```
tests/
├── __init__.py
├── test_proxy_utils.py    # Test IP validation, cleansing
├── test_extract.py        # Test proxy/URL extraction
├── test_database.py       # Test DB operations with temp DB
└── mock_network.py        # Mock rocksock for offline testing
```

```python
# tests/test_proxy_utils.py
import unittest
import sys
sys.path.insert(0, '..')
import fetch

class TestProxyValidation(unittest.TestCase):
    def test_valid_proxy(self):
        self.assertTrue(fetch.is_usable_proxy('8.8.8.8:8080'))

    def test_private_ip_rejected(self):
        self.assertFalse(fetch.is_usable_proxy('192.168.1.1:8080'))
        self.assertFalse(fetch.is_usable_proxy('10.0.0.1:8080'))
        self.assertFalse(fetch.is_usable_proxy('172.16.0.1:8080'))

    def test_invalid_port_rejected(self):
        self.assertFalse(fetch.is_usable_proxy('8.8.8.8:0'))
        self.assertFalse(fetch.is_usable_proxy('8.8.8.8:99999'))

if __name__ == '__main__':
    unittest.main()
```

**Files:** tests/ directory
**Effort:** High (initial), Low (ongoing)
**Risk:** Low

---

## Long Term (Future)

### [x] 16. Geographic Validation

**Completed.** Added IP2Location and pyasn for proxy geolocation.
- requirements.txt: Added IP2Location package
- proxywatchd.py: IP2Location for country lookup, pyasn for ASN lookup
- proxywatchd.py: Fixed ValueError handling when database files missing
- data/: IP2LOCATION-LITE-DB1.BIN (2.7M), ipasn.dat (23M)
- Output shows country codes: `http://1.2.3.4:8080 (US)` or `(IN)`, `(DE)`, etc.

---

### [x] 17. SSL Proxy Testing

**Completed.** Added SSL checktype for TLS handshake validation.
- config.py: Default checktype changed to 'ssl'
- proxywatchd.py: ssl_targets list with major HTTPS sites
- Validates TLS handshake with certificate verification
- Detects MITM proxies that intercept SSL connections

### [x] 18. Additional Search Engines

**Completed.** Added modular search engine architecture.
- engines.py: SearchEngine base class with build_url(), extract_urls(), is_rate_limited()
- Engines: DuckDuckGo, Startpage, Mojeek (UK), Qwant (FR), Yandex (RU), Ecosia, Brave
- Git hosters: GitHub, GitLab, Codeberg, Gitea
- scraper.py: EngineTracker class for multi-engine rate limiting
- Config: [scraper] engines, max_pages settings
- searx.instances: Updated with 51 active SearXNG instances

### [x] 19. REST API

**Completed.** Added HTTP API server for querying working proxies.
- httpd.py: ProxyAPIServer class with BaseHTTPServer
- Endpoints: /proxies, /proxies/count, /health
- Params: limit, proto, country, format (json/plain)
- Integrated into proxywatchd.py (starts when httpd.enabled=True)
- Config: [httpd] section with listenip, port, enabled

### [x] 20. Web Dashboard

**Completed.** Added web dashboard with live statistics.
- httpd.py: DASHBOARD_HTML template with dark theme UI
- Endpoint: /dashboard (HTML page with auto-refresh)
- Endpoint: /api/stats (JSON runtime statistics)
- Stats include: tested/passed counts, success rate, thread count, uptime
- Tor pool health: per-host latency, success rate, availability
- Failure categories: timeout, proxy, ssl, closed, etc.
- proxywatchd.py: get_runtime_stats() method provides stats callback

### [x] 21. Dashboard Enhancements (v2)

**Completed.** Major dashboard improvements for better visibility.
- Prominent check type badge in header (SSL/JUDGES/HTTP/IRC with color coding)
- System monitor bar: load average, memory usage, disk usage, process RSS
- Anonymity breakdown: elite/anonymous/transparent proxy counts
- Database health indicators: size, tested/hour, added/day, dead count
- Enhanced Tor pool: total requests, success rate, healthy nodes, avg latency
- SQLite ANALYZE/VACUUM functions for query optimization (dbs.py)
- Database statistics API (get_database_stats())

### [x] 22. Completion Queue Optimization

**Completed.** Eliminated polling bottleneck in proxy test collection.
- Added `completion_queue` for event-driven state signaling
- `ProxyTestState.record_result()` signals when all targets complete
- `collect_work()` drains queue instead of polling all pending states
- Changed `pending_states` from list to dict for O(1) removal
- Result: `is_complete()` eliminated from hot path, `collect_work()` 54x faster

---

## Profiling-Based Performance Optimizations

**Baseline:** 30-minute profiling session, 25.6M function calls, 1842s runtime

The following optimizations were identified through cProfile analysis. Each is
assessed for real-world impact based on measured data.

### [x] 1. SQLite Query Batching

**Completed.** Added batch update functions and optimized submit_collected().

**Implementation:**
- `batch_update_proxy_latency()`: Single SELECT with IN clause, compute EMA in Python,
  batch UPDATE with executemany()
- `batch_update_proxy_anonymity()`: Batch all anonymity updates in single executemany()
- `submit_collected()`: Uses batch functions instead of per-proxy loops

**Previous State:**
- 18,182 execute() calls consuming 50.6s (2.7% of runtime)
- Individual UPDATE for each proxy latency and anonymity

**Improvement:**
- Reduced from N execute() + N commit() to 1 SELECT + 1 executemany() per batch
- Estimated 15-25% reduction in SQLite overhead

---

### [ ] 2. Proxy Validation Caching

**Current State:**
- `is_usable_proxy()`: 174,620 calls, 1.79s total
- `fetch.py:242 <genexpr>`: 3,403,165 calls, 3.66s total (proxy iteration)
- Many repeated validations for same proxy strings

**Proposed Change:**
- Add LRU cache decorator to `is_usable_proxy()`
- Cache size: 10,000 entries (covers typical working set)
- TTL: None needed (IP validity doesn't change)

**Assessment:**
```
Current cost:     5.5s per 30min = 11s/hour = 4.4min/day
Potential saving: 50-70% cache hit rate = 2.7-3.8s per 30min = 5-8s/hour
Effort:           Very low (add @lru_cache decorator)
Risk:             None (pure function, deterministic output)
```

**Verdict:** LOW PRIORITY. Minimal gain for minimal effort. Do if convenient.

---

### [x] 3. Regex Pattern Pre-compilation

**Completed.** Pre-compiled proxy extraction pattern at module load.

**Implementation:**
- `fetch.py`: Added `PROXY_PATTERN = re.compile(r'...')` at module level
- `extract_proxies()`: Changed `re.findall(pattern, ...)` to `PROXY_PATTERN.findall(...)`
- Pattern compiled once at import, not on each call

**Previous State:**
- `extract_proxies()`: 166 calls, 2.87s total (17.3ms each)
- Pattern recompiled on each call

**Improvement:**
- Eliminated per-call regex compilation overhead
- Estimated 30-50% reduction in extract_proxies() time

---

### [ ] 4. JSON Stats Response Caching

**Current State:**
- 1.9M calls to JSON encoder functions
- `_iterencode_dict`: 1.4s, `_iterencode_list`: 0.8s
- Dashboard polls every 3 seconds = 600 requests per 30min
- Most stats data unchanged between requests

**Proposed Change:**
- Cache serialized JSON response with short TTL (1-2 seconds)
- Only regenerate when underlying stats change
- Use ETag/If-None-Match for client-side caching

**Assessment:**
```
Current cost:     ~5.5s per 30min (JSON encoding overhead)
Potential saving: 60-80% = 3.3-4.4s per 30min = 6.6-8.8s/hour
Effort:           Medium (add caching layer to httpd.py)
Risk:             Low (stale stats for 1-2 seconds acceptable)
```

**Verdict:** LOW PRIORITY. Only matters with frequent dashboard access.

---

### [ ] 5. Object Pooling for Test States

**Current State:**
- `__new__` calls: 43,413 at 10.1s total
- `ProxyTestState.__init__`: 18,150 calls, 0.87s
- `TargetTestJob` creation: similar overhead
- Objects created and discarded each test cycle

**Proposed Change:**
- Implement object pool for ProxyTestState and TargetTestJob
- Reset and reuse objects instead of creating new
- Pool size: 2x thread count

**Assessment:**
```
Current cost:     ~11s per 30min = 22s/hour = 14.7min/day
Potential saving: 50-70% = 5.5-7.7s per 30min = 11-15s/hour = 7-10min/day
Effort:           High (significant refactoring, reset logic needed)
Risk:             Medium (state leakage bugs if reset incomplete)
```

**Verdict:** NOT RECOMMENDED. High effort, medium risk, modest gain.
Python's object creation is already optimized. Focus elsewhere.

---

### [ ] 6. SQLite Connection Reuse

**Current State:**
- 718 connection opens in 30min session
- Each open: 0.26ms (total 0.18s for connects)
- Connection per operation pattern in mysqlite.py

**Proposed Change:**
- Maintain persistent connection per thread
- Implement connection pool with health checks
- Reuse connections across operations

**Assessment:**
```
Current cost:     0.18s per 30min (connection overhead only)
Potential saving: 90% = 0.16s per 30min = 0.32s/hour
Effort:           Medium (thread-local storage, lifecycle management)
Risk:             Medium (connection state, locking issues)
```

**Verdict:** NOT RECOMMENDED. Negligible time savings (0.16s per 30min).
SQLite's lightweight connections don't justify pooling complexity.

---

### Summary: Optimization Priority Matrix

```
┌─────────────────────────────────────┬────────┬────────┬─────────┬───────────┐
│ Optimization                        │ Effort │ Risk   │ Savings │ Status
├─────────────────────────────────────┼────────┼────────┼─────────┼───────────┤
│ 1. SQLite Query Batching            │ Low    │ Low    │ 20-34s/h│ DONE
│ 2. Proxy Validation Caching         │ V.Low  │ None   │ 5-8s/h  │ Maybe
│ 3. Regex Pre-compilation            │ Low    │ None   │ 5-8s/h  │ DONE
│ 4. JSON Response Caching            │ Medium │ Low    │ 7-9s/h  │ Later
│ 5. Object Pooling                   │ High   │ Medium │ 11-15s/h│ Skip
│ 6. SQLite Connection Reuse          │ Medium │ Medium │ 0.3s/h  │ Skip
└─────────────────────────────────────┴────────┴────────┴─────────┴───────────┘

Completed: 1 (SQLite Batching), 3 (Regex Pre-compilation)
Remaining: 2 (Proxy Caching - Maybe), 4 (JSON Caching - Later)

Realized savings from completed optimizations:
  Per hour:   25-42 seconds saved
  Per day:    10-17 minutes saved
  Per week:   1.2-2.0 hours saved

Note: 68.7% of runtime is socket I/O (recv/send) which cannot be optimized
without changing the fundamental network architecture. The optimizations
above target the remaining 31.3% of CPU-bound operations.
```

---

## Potential Dashboard Improvements

### [ ] Dashboard Performance Optimizations

**Goal:** Ensure dashboard remains lightweight and doesn't impact system performance.

**Current safeguards:**
- No polling on server side (client-initiated via fetch)
- 3-second refresh interval (configurable)
- Minimal DOM updates (targeted element updates, not full re-render)
- Static CSS/JS (no server-side templating per request)
- No persistent connections (stateless HTTP)

**Future considerations:**
- [ ] Add rate limiting on /api/stats endpoint
- [ ] Cache expensive DB queries (top countries, protocol breakdown)
- [ ] Lazy-load historical data (only when scrolled into view)
- [ ] WebSocket option for push updates (reduce polling overhead)
- [ ] Configurable refresh interval via URL param or localStorage
- [ ] Disable auto-refresh when tab not visible (Page Visibility API)

### [ ] Dashboard Feature Ideas

**Low priority - consider when time permits:**
- [ ] Dark/light theme toggle
- [ ] Export stats as CSV/JSON from dashboard
- [ ] Historical graphs (24h, 7d) using stats_history table
- [ ] Per-ASN performance analysis
- [ ] Geographic map visualization (requires JS library)
- [ ] Alert thresholds (success rate < X%, MITM detected)
- [ ] Mobile-responsive improvements
- [ ] Keyboard shortcuts (r=refresh, t=toggle sections)

---

## Completed

### [x] Work-Stealing Queue
- Implemented shared Queue.Queue() for job distribution
- Workers pull from shared queue instead of pre-assigned lists
- Better utilization across threads

### [x] Multi-Target Validation
- Test each proxy against 3 random targets
- 2/3 majority required for success
- Reduces false negatives from single target failures

### [x] Interleaved Testing
- Jobs shuffled across all proxies before queueing
- Prevents burst of 3 connections to same proxy
- ProxyTestState accumulates results from TargetTestJobs

### [x] Code Cleanup
- Removed 93 lines dead HTTP server code (ppf.py)
- Removed dead gumbo parser (soup_parser.py)
- Removed test code (comboparse.py)
- Removed unused functions (misc.py)
- Fixed IP/port cleansing (ppf.py)
- Updated .gitignore

### [x] Rate Limiting & Instance Tracking (scraper.py)
- InstanceTracker class with exponential backoff
- Configurable backoff_base, backoff_max, fail_threshold
- Instance cycling when rate limited

### [x] Exception Logging with Context
- Replaced bare `except:` with typed exceptions across all files
- Added context logging to exception handlers (e.g., URL, error message)

### [x] Timeout Standardization
- Added timeout_connect, timeout_read to [common] config section
- Added stale_days, stats_interval to [watchd] config section

### [x] Periodic Stats & Stale Cleanup (proxywatchd.py)
- Stats class tracks tested/passed/failed with thread-safe counters
- Configurable stats_interval (default: 300s)
- cleanup_stale() removes dead proxies older than stale_days (default: 30)

### [x] Unified Proxy Cache
- Moved _known_proxies to fetch.py with helper functions
- init_known_proxies(), add_known_proxies(), is_known_proxy()
- ppf.py now uses shared cache via fetch module

### [x] Config Validation
- config.py: validate() method checks config values on startup
- Validates: port ranges, timeout values, thread counts, engine names
- Warns on missing source_file, unknown engines
- Errors on unwritable database directories
- Integrated into ppf.py, proxywatchd.py, scraper.py main entry points

### [x] Profiling Support
- config.py: Added --profile CLI argument
- ppf.py: Refactored main logic into main() function
- ppf.py: cProfile wrapper with stats output to profile.stats
- Prints top 20 functions by cumulative time on exit
- Usage: `python2 ppf.py --profile`

### [x] SIGTERM Graceful Shutdown
- ppf.py: Added signal handler converting SIGTERM to KeyboardInterrupt
- Ensures profile stats are written before container exit
- Allows clean thread shutdown in containerized environments
- Podman stop now triggers proper cleanup instead of SIGKILL

### [x] Unicode Exception Handling (Python 2)
- Problem: `repr(e)` on exceptions with unicode content caused encoding errors
- Files affected: ppf.py, scraper.py (3 exception handlers)
- Solution: Check `isinstance(err_msg, unicode)` then encode with 'backslashreplace'
- Pattern applied:
  ```python
  try:
      err_msg = repr(e)
      if isinstance(err_msg, unicode):
          err_msg = err_msg.encode('ascii', 'backslashreplace')
  except:
      err_msg = type(e).__name__
  ```
- Handles Korean/CJK characters in search queries without crashing