metricchrono 0.5.0

Python bindings for MetricChrono with native acceleration and a pure-Python fallback

MetricChrono Python

Production Python bindings for MetricChrono. The PyPI package uses the bundled metricchrono-ffi native library via ctypes when it is present, and otherwise falls back to a byte-identical pure-Python implementation — so import metricchrono works with no native library at all.

Install

python3 -m pip install metricchrono

pip install metricchrono downloads a pre-built binary wheel and needs no Rust toolchain or Cargo. Wheels are published for Linux (manylinux x86_64 and aarch64), macOS (Intel and Apple Silicon), and Windows (x86_64).

Only the source-distribution (sdist) fallback build needs Cargo plus a Rust toolchain — used when pip cannot find a matching wheel for the target platform.

Source-tree usage

To use the wrapper directly from a checkout without installing a wheel, no native build is required: the pure-Python backend works from the source tree. Build the native library from the repository root only when you want the fast path, then point Python at it:

cargo build -p metricchrono-ffi --release
export METRICCHRONO_FFI_LIB=target/release/libmetricchrono_ffi.dylib

Use .so on Linux and .dll on Windows.

Backend selection defaults to METRICCHRONO_BACKEND=auto: Python uses the native library when the wheel or source-tree cdylib is present and loadable, otherwise it falls back to the pure-Python backend. Force a backend with METRICCHRONO_BACKEND=python or METRICCHRONO_BACKEND=native.

Example

import metricchrono as mc

ladder = mc.geometric_ladder(0.5, 1.0, 2.0, 4, 0.5, 1.0)
print(mc.ladder_distance(3.0, ladder))

with mc.EventLog(2) as log:
    log.append(10, [1.0, 0.0])
    log.append(11, [1.0, 1.0])
    print(log.next_event(0, 0))

API Surface

• Kernel and ladders: Tier, tick_distance, ladder_distance, geometric_ladder, Ladder.
• Serialization: tier_from_schema, tier_to_schema, ladder_from_schema, ladder_to_schema, tick_vector_from_schema, tick_vector_to_schema, consensus_result_from_schema.
• Smooth and adaptive helpers: smooth_tick_distance, smooth_ladder_distance, adaptive_ladder_distance, ZoomDecision.
• Event memory: EventLog.
• Consensus: weighted_consensus, coherence_residuals, simple_weight_update.