Parenthesized and repeated paths

Parenthesized subpaths let you constrain a segment before an enclosing selector. Repeating a group joins each iteration’s endpoint to the next start. A repeated body that is a fixed chain of nodes and edges extends only the paths reached so far. A quantifier, alternative, or optional subpath inside the body is still matched as one step and then joined to that frontier. That one-step relation is shared across input rows when the body does not read an outer variable; dependent bodies keep a step for each row.

MATCH p = (a {name: 'Alice'})((x)-[edges:Knows]->(y)){0,2}(b)
RETURN b.name AS destination, PATH_LENGTH(p) AS hops ORDER BY hops;

New element declarations inside a repeated group become ordered lists outside it. Zero repetitions expose empty lists. Nested groups concatenate their lists. Variables must be fresh relative to the outer bindings; declaration capture is rejected. A MATCH WHERE sees those lists, so WHERE x.name = 'Alice' is rejected. Put that comparison on the node inside the group to test each repetition.

MATCH p = ALL SHORTEST
  ((a {name: 'Alice'})-[:Knows]->(via)-[:Knows]->(b)
   WHERE via.name = 'Bob')
RETURN b.name AS destination, PATH_LENGTH(p) AS hops;

This prefilters the two-edge segment before selecting paths. A repeated group must have positive minimum edge length; otherwise repeated zero-length matches would not make progress. Adjacent node patterns constrain the same node, and consecutive edges introduce an implicit intermediate node.

The implementation composes DataFusion relations and recursive queries. General repeatable unbounded groups and full history-dependent selection are not implemented; provide a finite bound.