PerfTrace 🔍

PerfTrace is a unified performance tracing and profiling CLI for Python applications.

It provides detailed insights into function execution, context/module performance, CPU and memory usage, and system metrics, with rich statistical summaries and export support — all through a clean, production-ready command-line interface.

PerfTrace is developer-centric, explicit, and lightweight.
It focuses on performance analysis, not error or exception tracking.

---

✨ Key Features

• 🔍 Function & context-level profiling
• 📊 Statistical metrics (min / max / avg / p90 / p95 / p99 / std dev)
• 🕒 Recent, daily, and historical analysis
• 🐢 Slowest / fastest execution detection
• 🧠 System & memory monitoring
• 📁 Export to CSV & JSON
• 🩺 Health diagnostics (doctor)
• ⚙️ Configurable storage backends (DuckDB & PostgreSQL)
• 🧩 Function, class, and context-manager based instrumentation

---

📦 Installation

pip install perftrace

Requirements

• Python 3.11+

---

🚀 Getting Started

perftrace help

Recommended first commands:

perftrace summary
perftrace doctor
perftrace stats-function <FUNCTION_NAME>

---

🧠 How PerfTrace Works

PerfTrace works in two clear phases:

1. Instrumentation
   Developers explicitly mark functions, classes, or code blocks using decorators or context managers.

2. Analysis
   The CLI queries stored metrics and generates summaries, statistics, and exports.

PerfTrace runs only when your code runs — there are no background agents, daemons, or always-on processes.

---

🧩 Instrumenting Your Code

🎛 Selecting Metrics to Collect

PerfTrace allows you to explicitly control which performance metrics are collected.

You can either:

• provide a list of specific metrics, or
• use "all" to collect every supported metric

Supported Metrics

Metric	Description
cpu	CPU usage and CPU time
memory	Memory usage and deltas
execution	Execution time
file	File I/O activity
garbagecollector	Garbage collection activity
ThreadContext	Thread and execution context
network	Network activity

Using Specific Metrics

from perftrace import perf_trace_metrics

@perf_trace_metrics(profilers=["cpu", "memory", "execution"])
def compute():
    return [i for i in range(100000)]

Using All Metrics

@perf_trace_metrics(profilers="all")
def full_trace():
    return [i for i in range(100000)]

Class-Level Profiling

from perftrace import perf_trace_metrics_cl

@perf_trace_metrics_cl(profilers=["cpu", "execution"])
class MyProcessor:
    @staticmethod
    def step1(x):
        return x + 1

    def step2(self, y):
        return y * 2

Context-Based Profiling

from perftrace import PerfTraceContextManager

with PerfTraceContextManager(
    context_tag="work",
    cls_collectors=["cpu", "memory"]
):
    work = [x ** 2 for x in range(100000)]

---

📖 Complete CLI Reference

General Commands

Command	Description
version	Show PerfTrace version
help	Display help
doctor	Run health checks
summary	Overall performance summary
list	List available functions and contexts

Function Commands

Command	Description
show-function <name>	Detailed trace data
stats-function <name>	Statistical metrics
recent-function	Recently executed functions
search-function <name>	Search execution history
count-function	Execution frequency
slowest	Slowest execution
fastest	Fastest execution

Context Commands

Command	Description
show-context <name>	Context trace data
stats-context <name>	Statistical metrics
recent-context	Recently executed contexts
search-context <name>	Search history
count-context	Execution frequency

Time-Based

Command	Description
today	Executions today
history	Historical data

System & Memory

Command	Description
system-status	System status
system-info	System information
system-monitor	Real-time monitoring
memory	Memory usage

Export

Command	Description
export-csv	Export database to CSV
export-json	Export database to JSON
export-function-csv	Export function CSV
export-context-csv	Export context CSV
export-function-json	Export function JSON
export-context-json	Export context JSON

---

⚙️ Configuration (YAML)

Config file locations:

• Linux / macOS: ~/.perftrace/config.yaml
• Windows: %USERPROFILE%\.perftrace\config.yaml

DuckDB (Default)

database:
  engine: duckdb
  duckdb:
    path: ./data/default.duckdb

PostgreSQL (Optional)

database:
  engine: postgresql
  postgresql:
    host: localhost
    port: 5432
    user: postgres
    password: your_password

Interactive setup:

perftrace set-config

Verify:

perftrace doctor

---

🔍 Positioning (PerfTrace vs APM Tools)

PerfTrace complements APM tools by providing:

• precise function-level metrics
• lightweight, on-demand profiling
• local and CI-friendly analysis

PerfTrace is not a distributed tracing or error-tracking system.

---

📄 License

MIT License