ScatterNet.jl

Layout

src/
  ScatterNet.jl          thin aggregator (includes Interfaces + AtomicRadii + the two groups)
  Interfaces.jl          RadiiSource / FormFactorSource markers (shared)
  AtomicRadii/
    AtomicRadii.jl       ion parsing + table (loaded once) + fallback chain
    atomic_radii.sqlite3
  Molecule/
    Molecule.jl          module Molecule
    Cache.jl             Lazy{T} + force
    Molecules.jl         create + r/theta/phi/coords/radii/vols/elms/name
    SASA.jl              Shrake-Rupley SASA + hydration-shell patch geometry
  Scattering/
    Scattering.jl        module Scattering
    SphFuncs.jl          sphHarm, sphBess          (SphericalHarmonics.jl, Bessels.jl)
    FormFactorXrayDB.jl  compute_form_factors      (xraydb via PythonCall)
py/FormFact_py.py        Python form-factor tiers
test/                    test suites + Aqua/JET

Modules

src/ScatterNet.jl includes Interfaces, AtomicRadii, then two grouped submodules, each its own folder with a parent module … end that includes its files in dependency order: Molecule (Cache, Molecules, SASA) and Scattering (SphFuncs, FormFactorXrayDB). AtomicRadii is a sibling of Molecule, not nested inside it: it only implements Interfaces.RadiiSource, the abstract backend Molecule.Molecules.create consults, so it has no dependency on Molecule in the other direction and ships its own bundled atomic_radii.sqlite3 next to it. Reach a leaf as ScatterNet.Molecule.SASA etc. export lists the public surface; _-prefixed names are internal. Swappable backends (RadiiSource, FormFactorSource) are abstract types + a documented method contract, in place of OCaml functors.

Run

julia --project=. -e 'using Pkg; Pkg.instantiate()'
julia --project=. -e 'using Pkg; Pkg.test()'      # xraydb tests self-skip without the CondaPkg env

PythonCall/CondaPkg build the xraydb+numpy env on first pyimport; force it with julia --project=. -e 'using CondaPkg; CondaPkg.resolve()'.