AD-safe evaluation hooks
The problem
The stock Distributions.jl evaluators for a Gamma are not differentiable in the distribution's shape and scale. cdf(::Gamma) routes through SpecialFunctions.gamma_inc, whose ChainRule leaves the shape partial unimplemented, and logcdf/logccdf route through StatsFuns._gammalogcdf/_gammalogccdf, which have no Dual shape method. A downstream numeric kernel that evaluates a Gamma component's CDF, survival, or density inside a differentiated integrand therefore breaks under every supported AD backend.
The fix
Five hooks give a wrapper package one sanctioned surface to evaluate a component through:
Each is an extension point. The generic method falls through to the stock Distributions function, so a distribution that already differentiates cleanly is untouched. The Gamma methods route through the gamma-CDF derivative so a Gamma component stays differentiable in its parameters.
using EpiAwareADTools, Distributions
d = Gamma(2.0, 1.0)
cdf_ad_safe(d, 3.0), logccdf_ad_safe(d, 3.0)(0.8008517265285442, -1.6137056388801092)Extending the hooks
A wrapper package adds a method for its own component type. For example, a package whose modified density is not AD-safe under a given backend adds a pdf_ad_safe method for its type, and the numeric kernel that calls pdf_ad_safe then differentiates cleanly:
import EpiAwareADTools: pdf_ad_safe
pdf_ad_safe(d::MyModifiedDist, t::Real) = my_ad_safe_density(d, t)Upstream target
The family exists because the Gamma CDF is not differentiable in its parameters upstream — the same gap the gamma-CDF derivative fills (SpecialFunctions.jl issue #531). The hooks are deleted once a Gamma CDF differentiable in its parameters ships upstream and wrapper packages can call cdf/logcdf directly.