Release Notes

v0.5.1 — Direct CNRS-H Chain Rule

• Added cnrs.cnrs_h_chain, a finite-order CNRS-H coefficient-space chain-rule layer.
• Added EGF-series composition compose_series(f, g, order=N).
• Added direct chain-rule checks D(f ∘ g) versus (Df ∘ g) * Dg without using CnrsDual.
• Added tests/test_cnrs_h_chain_rule.py.
• Added examples/science_workflows/cnrs_h_native_chain_rule_demo.py.
• Scope: finite truncated CNRS-H series; this is a structural coefficient-calculus implementation, not a full analytic-continuation theorem.

Validation: 811 passed, 6 xfailed.

v0.5.1 — Symbolic-to-CNRS-H Bridge

Introduces the first tested bridge between the symbolic calculus layer and CNRS-H coefficient calculus. Supported symbolic expressions can now be converted into finite CNRS-H exponential-generating-function representations, differentiated or integrated structurally, and compared against symbolic differentiation/integration paths.

Added

• cnrs/cnrs_h_bridge.py with conservative symbolic-to-CNRS-H conversion.
• cnrs_h_from_symbolic(expr, var, order, env) for supported expressions around the expansion point var = 0.
• Commutation checks for symbolic differentiation vs. CNRS-H digit-shift differentiation.
• Commutation checks for conservative symbolic integration vs. CNRS-H digit-shift integration, with integration constants handled explicitly.
• tests/test_cnrs_h_bridge.py covering polynomials, scale laws, exponentials, sine/cosine, symbolic parameters, unsupported cases, and bridge comparison helpers.
• examples/science_workflows/symbolic_to_cnrs_h_demo.py.

Supported bridge subset

• Constants and the expansion variable.
• Polynomials built from +, -, *, and non-negative integer powers.
• Division by expressions independent of the expansion variable.
• exp, sin, and cos of affine arguments such as k*s, s/L, or k*s+b.
• Scale laws such as A*exp(k*s) with numeric parameter values supplied through env.

Scope note

This is a minimal bridge, not a full symbolic series engine. Unsupported expressions raise UnsupportedBridgeExpression rather than silently producing unreliable expansions.

Validation

811 passed, 6 xfailed

The 6 expected failures remain known representational-limit tests, not regressions. Existing reliable-domain warnings from scale-law and oscillator tests remain expected.