Skip to content

Latest commit

 

History

History
117 lines (100 loc) · 5.79 KB

architecture.md

File metadata and controls

117 lines (100 loc) · 5.79 KB

Architecture

Docs Hub / Overview / Architecture

High-Level Data Flow

graph TB
    REQ[HTTP Request] --> SPC[Symfony Profiler Collectors<br/>Doctrine, Cache, Twig, Events, HttpClient, Stopwatch]
    SPC --> DC["OptimizationAdvisorDataCollector::lateCollect()<br/>priority -100"]
    DC --> DBA[DatabaseAnalyzer]
    DC --> CAA[CacheAnalyzer]
    DC --> TWA[TwigAnalyzer]
    DC --> EVA[EventAnalyzer]
    DC --> HCA[HttpClientAnalyzer]
    DC --> OSA[OtherSignalsAnalyzer]
    DC --> PFA[PerformanceAnalyzer]
    DBA --> ENG["AdvisorEngine::evaluate()<br/>14 detection rules"]
    CAA --> ENG
    TWA --> ENG
    EVA --> ENG
    HCA --> ENG
    OSA --> ENG
    ENG --> DEDUP[Deduplicate by fingerprint]
    DEDUP --> SORT[Sort by ROI descending]
    SORT --> CAP["Cap at max_items (default 200)"]
    CAP --> PANEL[Profiler Panel]
    PFA --> DC
Loading

Request Lifecycle

sequenceDiagram
    participant R as Request
    participant P as Symfony Profiler
    participant DC as DataCollector
    participant A as Analyzers (7)
    participant E as AdvisorEngine
    R ->> P: Request handled
    P ->> DC: collect(request, response)
    Note over DC: Stores route, controller, method, URI
    P ->> P: Other collectors finish
    P ->> DC: lateCollect()
    DC ->> A: analyze(profiler data)
    A -->> DC: signals (per category)
    DC ->> E: evaluate(db, cache, twig, events, http, messenger)
    E -->> DC: scored opportunities
    Note over DC: Build summary, store in $data
    DC -->> P: Ready for panel rendering
Loading

Origin Classification

All signals are classified into three origins. Only app signals generate opportunities.

flowchart TD
    SIG[Signal] --> TYPE{Signal type?}
    TYPE -->|DB query| DB_CHECK{Tables match<br/>infra prefixes?}
    DB_CHECK -->|pg_, information_schema| INFRA[infra]
    DB_CHECK -->|infra_db_tables config| INFRA
    DB_CHECK -->|SQL contains<br/>CURRENT_DATABASE etc .| INFRA
    DB_CHECK -->|No match| APP[app]
    TYPE -->|Cache pool| CACHE_CHECK{Pool name prefix?}
    CACHE_CHECK -->|profiler_cache_pool_prefixes| PROF[profiler]
    CACHE_CHECK -->|app_cache_pool_prefixes| APP
    CACHE_CHECK -->|Neither| INFRA
    TYPE -->|Twig template| TWIG_CHECK{Template prefix?}
    TWIG_CHECK -->|profiler_template_prefixes| PROF
    TWIG_CHECK -->|No match| APP
    TYPE -->|Event listener| EVT_CHECK{Listener class?}
    EVT_CHECK -->|profiler_event_namespace_prefixes| PROF
    EVT_CHECK -->|profiler_event_classes| PROF
    EVT_CHECK -->|app_namespace_prefix| APP
    EVT_CHECK -->|Other| INFRA
    TYPE -->|Messenger handler| MSG_CHECK{Handler namespace?}
    MSG_CHECK -->|app_namespace_prefix| APP
    MSG_CHECK -->|Other| INFRA
    TYPE -->|HTTP call| HTTP_APP[app]
Loading

Scoring Model

Each opportunity is scored on three dimensions (1-5 scale):

Dimension Description
Impact Expected performance improvement magnitude
Effort Implementation work required
Confidence Detection certainty (higher = fewer false positives)

Derived Metrics

Metric Formula Purpose
ROI impact * confidence / effort Sorting (highest first)
Quick Win effort <= 2 AND confidence >= 4 Low-hanging fruit flag
High Impact impact >= 4 High-value change flag
Fingerprint md5(code + "|" + evidence) Deduplication key
Optimization Score 100 - SUM(impact * confidence * 0.5) Request health (0-100)

Signal Flow Summary

Analyzer Data Source Output Key Detection Rules
DatabaseAnalyzer DebugDataHolder db PG_SLOW_QUERY_GROUP, PG_N_PLUS_ONE_SUSPECTED, PG_DUPLICATE_QUERY, PG_LARGE_RESULTSET, CACHE_CANDIDATE_DB_RESULTS, DOCTRINE_2LC_OPPORTUNITY
CacheAnalyzer CacheDataCollector cache CACHE_LOW_HITRATE_POOL
TwigAnalyzer Twig\Profiler\Profile twig TWIG_HOT_TEMPLATE, TWIG_DUP_RENDER
EventAnalyzer TraceableEventDispatcher events EVENTS_TOO_MANY_LISTENERS, EVENTS_SLOW_LISTENER
HttpClientAnalyzer HttpClientDataCollector http HTTP_SLOW_ENDPOINT, HTTP_DUP_CALL
OtherSignalsAnalyzer TraceRegistry other MESSENGER_SYNC_HEAVY
PerformanceAnalyzer Stopwatch performance (informational only)

← Overview | Next: Tech Stack →