fix(cypher): support multi-row scalar prefix in MATCH re-entry (#1047) - #1068

…entry (#1071)

Pre-rewrite the prefix WITH stage in _compile_bounded_reentry_query so
secondary whole-row aliases are demoted to scalar property carries
(S.X AS __cypher_reentry_<S>_<X>__) and downstream S.X references compose
with the existing single-whole-row machinery (#1047/#1068). RETURN of a
secondary whole-row alias, and re-binding a secondary alias as a node
variable in the trailing MATCH, remain unsupported with precise errors.

Closes #1071.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

…entry (#1071)

Pre-rewrite the prefix WITH stage in _compile_bounded_reentry_query so
secondary whole-row aliases are demoted to scalar property carries
(S.X AS __cypher_reentry_<S>_<X>__) and downstream S.X references compose
with the existing single-whole-row machinery (#1047/#1068). RETURN of a
secondary whole-row alias, and re-binding a secondary alias as a node
variable in the trailing MATCH, remain unsupported with precise errors.

Closes #1071.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

…entry (graphistry#1071)

Pre-rewrite the prefix WITH stage in _compile_bounded_reentry_query so
secondary whole-row aliases are demoted to scalar property carries
(S.X AS __cypher_reentry_<S>_<X>__) and downstream S.X references compose
with the existing single-whole-row machinery (graphistry#1047/graphistry#1068). RETURN of a
secondary whole-row alias, and re-binding a secondary alias as a node
variable in the trailing MATCH, remain unsupported with precise errors.

Closes graphistry#1071.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

…entry (#1071)

Pre-rewrite the prefix WITH stage in _compile_bounded_reentry_query so
secondary whole-row aliases are demoted to scalar property carries
(S.X AS __cypher_reentry_<S>_<X>__) and downstream S.X references compose
with the existing single-whole-row machinery (#1047/#1068). RETURN of a
secondary whole-row alias, and re-binding a secondary alias as a node
variable in the trailing MATCH, remain unsupported with precise errors.

Closes #1071.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

) (#1249)

* fix(gfql/row): resolve bare relationship alias in row expressions (#1072)

Bare edge aliases in select / where_rows / order_by row expressions raised
"unsupported token in row expression" because _gfql_resolve_token only had a
node-id fallback for the bare-alias case. Add an edge fallback: prefer the
edge-id column, otherwise render the relationship as a Cypher-style
[:TYPE {props}] string, mirroring what cypher RETURN <relAlias> already
produces via result_postprocess. Direct GFQL chains and cypher-lowered chains
now see the same column shape.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

* feat(cypher): support multi-alias carry through WITH before MATCH re-entry (#1071)

Pre-rewrite the prefix WITH stage in _compile_bounded_reentry_query so
secondary whole-row aliases are demoted to scalar property carries
(S.X AS __cypher_reentry_<S>_<X>__) and downstream S.X references compose
with the existing single-whole-row machinery (#1047/#1068). RETURN of a
secondary whole-row alias, and re-binding a secondary alias as a node
variable in the trailing MATCH, remain unsupported with precise errors.

Closes #1071.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

* feat(gfql): unify relationship alias rendering and harden row resolution (#1072)

* fix(gfql/cypher): close wave-3 alias resolution and reentry where gaps (#1072)

* docs(changelog): add #1072 entry for relationship alias in row expressions

Records the user-facing fix and amplifications for #1072 under
## [Development] → ### Fixed: bare edge-alias resolution in row
expressions, shared vectorized renderer in row/entity_props.py,
edge-alias gating via _gfql_rows_edge_aliases, string-escape /
float / empty-type rendering normalization, and re-entry WHERE
secondary-alias predicate rewriting via synthesized text.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

* test(gfql/row): add cuDF smoke test for bare relationship alias rendering

Wave 4 finding T2 (IMPORTANT): TestRelationshipAliasInRowExpression
had 14 new tests in PR #1249 but zero cuDF coverage on the new
edge-alias rendering path, despite Wave 2 vectorization being
specifically motivated by cuDF support (Wave 1 finding #4).

Adds test_select_bare_relationship_alias_renders_on_cudf_when_available
following the sibling pattern at test_row_pipeline_vectorized_cudf_when_available
(pytest.importorskip("cudf")). Asserts: (a) result keeps cuDF backend,
(b) bare-alias rendering produces "[:WORKS_AT {workFrom: 2010}]",
(c) select/return cell-for-cell parity holds on cuDF.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

* refactor(gfql/row): collapse duplicate ValueError in _gfql_resolve_token

Wave 4 finding D1 (SUGGESTION): the conditional at pipeline.py:2310 and
the raise at pipeline.py:2312 produce identical errors. After the
node-id branch (lines 2306-2309) returns, the only remaining outcome
is "unsupported token", regardless of edge_aliases membership — both
branches raised the same f-string. Collapsed to a single raise.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>