GFQL/polars: connected comma-pattern (multi-alias join) crashes with AttributeError instead of running natively or clean NIE

[lmeyerov]

Summary

On engine='polars' / 'polars-gpu', a connected comma-pattern that shares a node alias — MATCH (p)-[R1]->(i), (p)-[R2]->(c) WHERE ... (graph-benchmark q5/q6/q7 shape) — crashes with AttributeError: 'DataFrame' object has no attribute 'merge'. The _compile_connected_match_join lowering emits a chain that runs the join via the pandas row pipeline (.merge), which a polars frame does not support.

Impact

• These queries run correctly on pandas/cuDF but crash on polars/polars-gpu (surfaced by the benchmarks/gfql: q1-q9 runner uses dataframe shortcuts (bypasses GFQL engine) + untimed lowercase precompute — unfair vs competitor DBs #1710 real-GFQL benchmark run — q5/q6/q7).
• NO-CHEATING: even if not natively supported, it should raise a clean NotImplementedError pointing at engine='pandas', not an AttributeError from deep in the pipeline.

Fix options

1. Minimum: make the polars engine detect the connected-join / non-native row-pipeline op and raise a clean NotImplementedError (honest NIE).
2. Full: a native polars connected-join lowering (the binding-ops path GFQL/polars: native rows(binding_ops) — traversal queries NIE (blocks polars for edge/multi-hop Cypher) #1709 already does multi-alias joins natively; route the comma-pattern through it).

Related

• GFQL/cypher: multi-part MATCH...WITH <alias> MATCH (alias)... does NOT carry the filtered binding (silent wrong answer) #1712 (the connected-join correctness fix that made these run on pandas/cuDF), benchmarks/gfql: q1-q9 runner uses dataframe shortcuts (bypasses GFQL engine) + untimed lowercase precompute — unfair vs competitor DBs #1710 (benchmark).