ConvolvedDistributions v0.4.2
Updates the compat to support the latest version of EpiAwareADTools
Merged pull requests:
- Honour custom numeric integration solver payloads (Convolved, Difference, Product, Ratio) — #92 (#149) (@seabbs-bot)
- ci(deps): bump the github-actions group with 9 updates (#151) (@dependabot[bot])
- docs: point release-notes header at GitHub Releases (#152) (@seabbs-bot)
- feat: add public quantile_initial_guess hook (#150) (#156) (@seabbs-bot)
- docs: point developer FAQ at Discourse/epinowcast, not GitHub Discussions (#157) (@seabbs-bot)
- fix: closed-form logcdf/ccdf/logccdf for uniform-window pairs (#158) (@seabbs-bot)
- Exact lattice quantile for discrete combinations (#159) (@seabbs-bot)
- Add an output-position mask to convolve_series (#160) (@seabbs-bot)
- feat: shared repeat mechanism, convolved(d, k) and product(d, k) (#165) (@seabbs-bot)
- test(uniform_window): cover Weibull partial_expectation guards (#169) (@seabbs-bot)
- docs: point remaining discussion mentions at Discourse/epinowcast (#171) (@seabbs-bot)
- Migrate Difference onto the solver-dispatch generics (#176) (@seabbs-bot)
- fix: forward discretisation keywords through the time-varying convolve_series path (#177) (@seabbs-bot)
- test: AD coverage for custom solver payloads (#161) (#179) (@seabbs-bot)
- fix: close the truncated testitem breaking main's test suite (#182) (@seabbs-bot)
- chore: migrate to Runic and fix main CI (#183) (@seabbs-bot)
- feat: migrate Product and Ratio onto the solver-dispatch generics (#184) (@seabbs-bot)
- fix: stop benchmark-history aborting on commits that predate a benchmarked verb (#185) (@seabbs-bot)
- ci: raise the docs build timeout to 90 minutes (#186) (@seabbs-bot)
- feat: make the difference, product and ratio pair hooks public (#187) (@seabbs-bot)
- fix: floor EpiAwareADTools at 0.2 and drop its internal _gamma_cdf (#189) (@seabbs-bot)
Closed issues:
- Broaden test coverage: cross-method consistency suite and pair breadth (#57)
- JuliaFormatter disagrees with the committed state of test/ADFixtures/src/ADFixtures.jl (#81)
- Mixed discrete+continuous Convolved silently returns pdf 0.0 everywhere (#115)
- Exact lattice quantile for discrete combinations (#116)
- Convolved hard-codes {Univariate, Continuous}, blocking discrete closed forms (pmf) (#120)
- Release docs are not building? (#125)
- Support a masking vector for convolve series of some kind (#127)
- Add convolution_power(d, k) for repeated self-convolution (#132)
- Migrate Difference/Product/Ratio onto the solver-dispatch generics from #80 (#133)
- Developer FAQ points at GitHub Discussions, which is disabled (both links 404) (#135)
- Point the release-notes docs page at GitHub Releases (#136)
- Mooncake AD failing on main (#145)
- Design: make convolved(delay, Uniform) a drop-in for primary-censored delays — add log-space cdf/ccdf/logccdf for uniform-window pairs (#147)
- Release v0.4.1: custom integration solver payloads honored (#148)
- Add a public extension point for custom quantile initial-guess generation on Convolved (#150)
- Remaining GitHub Discussions mentions outside developer/faq.md (#153)
- convolve_series(delays::AbstractVector, series) cannot forward discretisation kwargs to delay_masses (#154)
- AD coverage gap: custom solver payloads are never traced by any backend (#161)
- test/ad environment cannot be instantiated locally: EpiAwareADTools pin conflict (#162)
- Unit-test partial_expectation(::Weibull)'s non-positive-input guard (#167)
- Quality: formatting testitem requires Runic, but package uses JuliaFormatter (#168)
- EpiAwareADTools compat capped at 0.1 blocks the whole stack from adopting 0.2.0 and its logccdf_ad_safe(::Gamma) tail fix (#188)