opendp

library0.15.1pypypi✓ verified 83d ago

Python bindings for the OpenDP Library, a modular collection of tools for differential privacy. Current version 0.15.1, released approximately quarterly. Provides a high-level API for privacy-preserving data analysis.

security

pip install opendp

github ↗homepage ↗

INSTALL

IMPORT

SIG · OPENDP

opendp

securitypythonv0.15.1

harness data pending

Install & Compatibility

Where this runs

No compatibility data collected yet for this library.

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

Measurement

✓ from opendp.typing import Measurement

✗ from opendp import Measurement

Measurement is in opendp.typing

then_count

✓ from opendp.transformations import then_count

✗ from opendp import then_count

Transformations are in opendp.transformations

base_laplace

✓ from opendp.measurements import base_laplace

✗ from opendp import base_laplace

Measurements are in opendp.measurements

Simple differential privacy count query using OpenDP's compositor and transformation chain.

import opendp from opendp.transformations import make_split_dataframe, make_select_column, then_count, then_clamp, then_resize, then_variance, then_mean from opendp.measurements import then_laplace, then_gaussian from opendp.domains import vector_domain, atom_domain from opendp.metrics import symmetric_distance from opendp.typing import L2Sensitivity context = opendp.Context.compositor( data=... , # your data privacy_unit=opendp.parameters.unit('epsilon=1, delta=0.01') ) # Example: count with Laplace noise meas = (opendp.preprocessing.make_split_dataframe(",", ["A", "B"])) >> \ (opendp.transformations.make_select_column("A", TOA=str)) >> \ (then_count()) >> \ (then_laplace(scale=1.0)) released = meas(meas.arg) print(released)

Debug

Known issues

breakingBreaking change in 0.15: Many functions now require explicit typing and domain/metric arguments. Old patterns like `base_laplace()` without explicit `TO` or `L2Sensitivity` fail.

fix

Ensure every measurement/transformation includes explicit type parameters. E.g., `then_laplace(scale=1.0, TO=int)`.

affects: >=0.15

deprecated`make_count` and similar direct measurements are replaced by `then_count` and chaining operations.

fix

Use transformation chains ending with measurement, e.g., `then_count() >> then_laplace(scale=1.0)`.

affects: >=0.9

gotchaThe Context API is not yet stable and may change. Avoid relying on it for production code.

fix

Use the lower-level combinator API (e.g., `make_chain_mt`) for stability.

affects: <1.0

gotchaPrivacy parameters epsilon and delta must be passed as a single string, not separate arguments.

fix

Use `opendp.parameters.unit('epsilon=1, delta=0.01')`.

affects: >=0.12

Errors

Common errors & fixes

AttributeError: module 'opendp' has no attribute 'transformations'

Older import pattern; opendp submodules are not directly exposed.

fix

Use `from opendp.transformations import ...` instead of `opendp.transformations.make_count`.

TypeError: base_laplace() missing 1 required positional argument: 'scale'

Base measurements require scale as an argument; old code might have used default.

fix

Always provide scale: `base_laplace(scale=1.0)`.

NotImplementedError: This operation is not supported.

Using a transformation or measurement before preparing context or chaining correctly.

fix

Ensure the chain consists of transformations ending with a measurement; use `>>` operator.

ImportError: cannot import name 'make_count' from 'opendp.transformations'

`make_count` was removed or renamed to `then_count`.

fix

Use `from opendp.transformations import then_count`.

Upgrade

Version history

0.15.1latest on PyPI · released May 29, 2026

Audit

Dependencies

No dependency data recorded yet.

Agent activity

15 hits · last 30 days

node

Resources

githubhttps://github.com/opendp/opendp ↗

homepagehttps://opendp.org ↗