Profiler for Python
The Python profiler is a decorator-based timing library. It tracks call counts, latency statistics and exceptions, and exports JSON, CSV or HTML. It does not document an API-upload path in the same way as the .NET, Java and Node.js agents, so treat it as an in-process measurement library rather than a production trace-collection agent.
Catalogue status. The Python profiler is present in the
profiler.pythonrepository. It is not part of the public profiler catalogue. The website's profiler catalogue intentionally excludes Python; do not present it as a published, installable agent.
Install
pip install reignite-profiler
Or add the profiler package directory to your project.
Use
from profiler import profile, get_session, JsonExporter
@profile
def calculate_fibonacci(n):
if n <= 1:
return n
return calculate_fibonacci(n - 1) + calculate_fibonacci(n - 2)
@profile
def process_data(items):
return [item * 2 for item in items]
fib_result = calculate_fibonacci(10)
data = process_data([1, 2, 3, 4, 5])
session = get_session()
JsonExporter.export_session(session, "profiling_results.json")
Async functions
from profiler import profile_async
import asyncio
@profile_async
async def fetch_data(url):
await asyncio.sleep(0.1)
return f"Data from {url}"
Sessions
from profiler import start_session, end_session
start_session("checkout")
run_checkout()
session = end_session()
print(f"Total calls: {session.total_calls}")
print(f"Duration: {session.duration_seconds:.2f}s")
Statistics
For each profiled method the session exposes call count, average, median, minimum, maximum, standard deviation and exception count:
session = get_session()
for name, profile in session.method_profiles.items():
print(f"{name}: {profile.call_count} calls, "
f"{profile.average_duration_ms:.2f} ms average")
Export
| Exporter | Output |
|---|---|
JsonExporter |
JSON, with to_json_string for in-memory use |
CsvExporter |
CSV |
HtmlExporter |
A standalone, sortable HTML report |
Limitations
- Timing is taken in-process and is subject to Python's GIL. It is appropriate for application-level timing (business logic, API calls, database queries), not for micro-benchmarking or GIL-contention analysis.
- For GIL-free profiling or micro-benchmarking, use
cProfileorpy-spy. - Disable the profiler in production with
disable_profiling()orenabled=Falseunless you have a specific reason to keep it on.