quantile-python

1.1 verified Fri May 01 auth: no python

Python Implementation of Graham Cormode and S. Muthukrishnan's Effective Computation of Biased Quantiles over Data Streams in ICDE'05. Version 1.1 is the latest stable release. The library has a low release cadence with no recent updates; it provides an efficient algorithm for computing biased quantiles over data streams.

pip install quantile-python

Common errors

error ModuleNotFoundError: No module named 'quantile-python' ↓

cause Importing the package using hyphens instead of underscores.

fix

Use 'import quantile_python' instead of 'import quantile-python'.

error AttributeError: module 'quantile_python' has no attribute 'QuantileEstimator' ↓

cause Misspelled class name (e.g., quantileestimator) or wrong import style.

fix

Use 'from quantile_python import QuantileEstimator'.

Warnings

gotcha The import path uses underscores (quantile_python), not hyphens. Importing with hyphens (quantile-python) will raise ModuleNotFoundError. ↓

fix Use 'from quantile_python import QuantileEstimator'.

gotcha The epsilon parameter controls the memory-accuracy tradeoff; very small epsilon (e.g., 0.0001) may cause excessive memory usage or slow insertions. ↓

fix Start with epsilon=0.01 or higher for large streams.

deprecated The library has not been updated since 1.1 (released ~2022). It depends on Python standard libraries only but may not support newer Python versions if breaking changes occur. ↓

fix Test with your Python version; consider switching to alternative libraries like tdigest if compatibility issues arise.

Imports

QuantileEstimator
```
from quantile_python import QuantileEstimator
```
The correct import uses underscores, not hyphens.
StreamSummary
```
from quantile_python import StreamSummary
```
StreamSummary is another key class for stream quantile estimation.

Quickstart

Demonstrates creating a QuantileEstimator, inserting values, and querying quantiles.

from quantile_python import QuantileEstimator

# Create an estimator with a desired error epsilon (epsilon = 0.001 means about 0.1% error)
estimator = QuantileEstimator(epsilon=0.001)
for value in range(1000):
    estimator.insert(value)

# Get quantiles: 0.5 (median), 0.9, 0.99
print("Median:", estimator.query(0.5))
print("90th percentile:", estimator.query(0.9))
print("99th percentile:", estimator.query(0.99))