Skip to content

kups.core.neighborlist.fixed

Empty and fixed-edge :class:NeighborList implementations.

These cover graph-construction cases without cutoff search: always-empty edge sets for point clouds and state-provided fixed edge topology for bonded interactions, optionally filtered to a patch-affected subset. They satisfy the standard :class:NeighborList[D] protocol so a unified graph constructor can ask any neighbor list for edges, regardless of how those edges were obtained.

The public call contract treats keys as the self-graph/output table. queries is keyword-only and reserved for true bipartite neighbor-list queries; fixed topology is a self-graph implementation and does not use queries. queried_keys is keyword-only, mutually exclusive with queries, and names affected keys ids after the caller has already written updated particle data into keys. FixedEdgesNeighborList returns only fixed topology rows touched by those ids. Its implementation is a normal selector -> mask -> compactor pipeline over the fixed edge rows; shifts are computed from the current particle positions during selection.

EmptyNeighborList

Neighbor list that emits an :class:Edges[D] with zero rows.

The degree field is the runtime arity carried by the emitted edges; it must match the type parameter D.

Attributes:

Name Type Description
degree int

Edge arity (Literal[0] for point clouds, higher for unified graph constructors that need a degree-aware empty NL).
Source code in src/kups/core/neighborlist/fixed.py 
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
@dataclass
class EmptyNeighborList[D: int]:
    """Neighbor list that emits an :class:`Edges[D]` with zero rows.

    The ``degree`` field is the runtime arity carried by the emitted edges;
    it must match the type parameter ``D``.

    Attributes:
        degree: Edge arity (``Literal[0]`` for point clouds, higher for
            unified graph constructors that need a degree-aware empty NL).
    """

    degree: int = field(static=True, default=0)

    @overload
    def __call__(
        self,
        keys: Table[ParticleId, NeighborListPoints],
        systems: Table[SystemId, NeighborListSystems],
        *,
        queries: Table[ParticleId, NeighborListPoints],
    ) -> Edges[D]: ...

    @overload
    def __call__(
        self,
        keys: Table[ParticleId, NeighborListPoints],
        systems: Table[SystemId, NeighborListSystems],
        *,
        queried_keys: Index[ParticleId] | None = None,
    ) -> Edges[D]: ...

    @jit
    def __call__(
        self,
        keys: Table[ParticleId, NeighborListPoints],
        systems: Table[SystemId, NeighborListSystems],
        *,
        queries: Table[ParticleId, NeighborListPoints] | None = None,
        queried_keys: Index[ParticleId] | None = None,
    ) -> Edges[D]:
        assert queries is None or queried_keys is None, (
            "Neighbor-list calls cannot combine queries with queried_keys."
        )
        del queries, systems, queried_keys
        shift_inner = self.degree - 1 if self.degree > 1 else 0
        return Edges(
            indices=Index(keys.keys, jnp.zeros((0, self.degree), dtype=int)),
            shifts=jnp.zeros((0, shift_inner, 3), dtype=int),
        )

FixedEdgesNeighborList

Neighbor list for a fixed topology edge set.

Full self-graph calls return all fixed topology rows with shifts computed from the current particle positions. Affected self-graph calls pass keyword-only queried_keys after updated particle data has been written into keys; the neighbor list returns only fixed rows touched by those affected keys ids. queries is reserved for true bipartite neighbor-list implementations and is not a fixed-edge update mechanism.

Attributes:

Name Type Description
indices Index[ParticleId]

Fixed edge topology. Shifts are intentionally not stored; they are computed from the call's current particle positions.
avg_edges Capacity[int] | None

Update-only average affected-edge capacity per affected keys id. Full calls ignore this field and use the stored topology length; affected calls default to the full edge-buffer size when it is not provided.
Source code in src/kups/core/neighborlist/fixed.py 
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
@dataclass
class FixedEdgesNeighborList[D: int]:
    """Neighbor list for a fixed topology edge set.

    Full self-graph calls return all fixed topology rows with shifts computed
    from the current particle positions. Affected self-graph calls pass
    keyword-only ``queried_keys`` after updated particle data has been written
    into ``keys``; the neighbor list returns only fixed rows touched by those
    affected ``keys`` ids. ``queries`` is reserved for true bipartite
    neighbor-list implementations and is not a fixed-edge update mechanism.

    Attributes:
        indices: Fixed edge topology. Shifts are intentionally not stored;
            they are computed from the call's current particle positions.
        avg_edges: Update-only average affected-edge capacity per affected
            ``keys`` id. Full calls ignore this field and use the stored
            topology length; affected calls default to the full edge-buffer
            size when it is not provided.
    """

    indices: Index[ParticleId]
    avg_edges: Capacity[int] | None = None

    @overload
    def __call__(
        self,
        keys: Table[ParticleId, NeighborListPoints],
        systems: Table[SystemId, NeighborListSystems],
        *,
        queries: Table[ParticleId, NeighborListPoints],
    ) -> Edges[D]: ...

    @overload
    def __call__(
        self,
        keys: Table[ParticleId, NeighborListPoints],
        systems: Table[SystemId, NeighborListSystems],
        *,
        queried_keys: Index[ParticleId] | None = None,
    ) -> Edges[D]: ...

    @jit
    def __call__(
        self,
        keys: Table[ParticleId, NeighborListPoints],
        systems: Table[SystemId, NeighborListSystems],
        *,
        queries: Table[ParticleId, NeighborListPoints] | None = None,
        queried_keys: Index[ParticleId] | None = None,
    ) -> Edges[D]:
        assert queries is None or queried_keys is None, (
            "Neighbor-list calls cannot combine queries with queried_keys."
        )
        assert queries is None, "FixedEdgesNeighborList only supports self-graph calls."
        if queried_keys is None:
            max_edges = FixedCapacity(len(self.indices))
        else:
            max_edges = (
                self.avg_edges.multiply(queried_keys.size)
                if self.avg_edges is not None
                else FixedCapacity(len(self.indices))
            )

        pipeline = Pipeline[D](
            selector=_FixedEdgesSelector(self.indices),
            masks=(TouchesQueriedKeysMask(),),
            compactor=ReduceCompactor(max_edges),
        )
        return pipeline(keys, systems, queried_keys=queried_keys)