Pragmatic approach to parsing importtime import profiles

While tuna is a fantastic way to visualize slowdowns in your python code, it lacks a native python-parsing API to allow for CI-based checks to determine or automatically flag if code is suddenly much slower (such as due to new imports).

import-timer acts as a middle-ground. Create a profile using python -X importtime, then read it in with our utilities to help you parse and interrogate the graph by hand, establish baselines, and then let you apply these easily later in your tests.

pip install import-timer

1. In an interactive instance, create an importtime log using python -X such as:

import subprocess
output = subprocess.run(["python3", "-X", "importtime", "-c", command], capture_output=True, text=True)

2. Load in the logs via import-timer:

from import-timer import read_import_profile

data = read_import_profile(output)

3. Interrogate your log using our helpers.

Will calculate the total time each tree in the data takes up by summing all of the times calculated down each node. This is an inplace operation, and will add a new key (total_time) to each node.

from import_timer import read_import_profile, calculate_total_time

data = read_import_profile(...)
total_time = calculate_total_time(data)
print(total_time)

Will sort all children of the passed in tree inplace based on the total time

from import_timer import read_import_profile, calculate_total_time

data = read_import_profile(...)
calculate_total_time(data)
sort_nodes_by_total_time(data)
print(data["children"][0]["total_time"])

A useful function that will take all of the nodes in data and then limit them to a threshold of which paths to return. Helpful in cases where you only want import paths which take up >30% of the time, or a particular minimum amount of seconds to start debugging

from import_timer import read_import_profile, calculate_total_time

data = read_import_profile(...)
calculate_total_time(data)
percentage_threshold = 20  # Threshold as a percentage of total time
threshold_time = total_time * (percentage_threshold / 100)  # Convert percentage to actual time threshold
max_depth = 7 # How deep in the tree do we want to report back paths
important_paths = get_paths_above_threshold(
    data_copy, threshold_time, max_depth
)
print(important_paths)

Explore our tests/ to see more, or this PR where this library was first utilized