Debug: Python slow startup / import time

Python application takes too long to start up, especially CLI tools or serverless functions.

Profile and reduce import time:

# 1. Measure import time
python3 -X importtime -c 'import myapp' 2> import.log
# Shows tree of imports with cumulative time
# Sort by time:
sort -t'|' -k2 -n import.log | tail -20

# 2. Profile startup
python3 -m cProfile -s cumulative -c 'import myapp' | head -30

Optimization strategies:

# 1. Lazy imports (move imports to function level)
# BAD: imports pandas on startup even if unused
import pandas as pd

def process_csv(path):
    return pd.read_csv(path)

# GOOD: import only when needed
def process_csv(path):
    import pandas as pd  # Only imported when function called
    return pd.read_csv(path)

# 2. Use importlib for conditional imports
import importlib
def get_parser():
    return importlib.import_module('lxml.etree')

# 3. Avoid importing entire packages
# BAD
import scipy  # Imports everything (~500ms)
# GOOD
from scipy.stats import norm  # Just what you need

# 4. Common slow imports to lazy-load:
# pandas: ~200-400ms
# numpy: ~100-200ms
# matplotlib: ~300-500ms
# scipy: ~300-500ms
# torch: ~500-2000ms
# tensorflow: ~2000-5000ms

1. For CLI tools: use lazy-loader package or __getattr__ pattern
2. For serverless: pre-import in the module scope (runs during cold start, cached for warm starts)

Python import time directly affects CLI responsiveness and serverless cold starts. Heavy imports like pandas or torch can add seconds to startup.

Python CLI tools, serverless functions, or apps with slow startup