How rlabbe/filterpy works
Project type and purpose
• This is an open-source Python library (“filterpy”) providing Kalman filters and related estimation filters (Extended KF, Unscented KF, g-h filters, Least Squares, discrete Bayes, etc.).
• It’s a modular, pedagogical codebase intended for import into user scripts, integration with numerical Python tools, and extension by researchers or students.
File structure
A. Top-level folders/files
– filterpy/ ← core source code, structured by filter category
– docs/ ← Sphinx documentation source
– conda/ ← Conda packaging recipes
– setup.py, setup.cfg, requirements.txt ← PyPI packaging & dependency metadata
– README.rst ← project overview, install & usage instructions
– tests (under each module) ← unit tests with pytest
– examples (under filterpy/examples) ← runnable demos
B. Core module breakdown
– filterpy/common ← helper utilities (discretization, kinematics)
– filterpy/kalman ← Kalman filters & smoothers
– filterpy/gh ← g-h filters
– filterpy/discrete_bayes
– filterpy/leastsq
– filterpy/memory
– filterpy/monte_carlo
– filterpy/stats
C. Build & publishing
– conda/recipes ← scripts and meta.yaml for conda-forge
– setup.py / MANIFEST.in ← PyPI distribution
– creating_a_release.rst ← release process docs
D. Documentation & CI
– docs/conf.py, Makefile ← Sphinx config
– Git triggers RTD builds on commit
– pytest invoked via “py.test”
Readme insights
• Emphasis on pedagogical clarity over micro-optimizations
• Dependencies: NumPy, SciPy, Matplotlib (optional), Python 3.5+ (2.7 legacy)
• Testing: basic asserts & pytest suite
• Distribution: PyPI & conda-forge