v.2.3.0 - state dtype unification

✅ Release Notes (GitHub / PyPI)

🚀 c4dynamics 2.2.0 — Deterministic State Type & Safer Assignments

This release brings a major improvement to the reliability of the State object.
To ensure consistent numerical precision and avoid dtype-related bugs, the state vector is now always stored internally as float64.

🔒 Fixed Internal Precision

Regardless of how variables are created or assigned, they are stored as np.float64.
This eliminates ambiguity and ensures predictable results in simulation, filters, dynamics, and optimization routines.

🔄 Deterministic Assignment Semantics

• All assignments to X or to individual fields are automatically cast to float64.
• Assignments with incorrect length raise a clear exception.
• All mutation paths update the underlying array reliably (previous versions could fail silently).

💡 Why this matters

• More stable Kalman filters
• Predictable integrators
• No platform-dependent dtype quirks
• Better compatibility with SciPy & NumPy internals
• Zero surprises when assigning Python ints, numpy ints, lists, or arrays

Full Changelog: v2.2.0...v2.3.0