GFQLSchemaError filtering on boolean label__ columns

[lmeyerov]

Summary

Filtering on boolean label columns in a GFQL chain raises GFQLSchemaError claiming the label column is a string, even though the dataframe shows a bool dtype. This blocks label predicates on relationship endpoints (e.g., MATCH (:A)-[r]->(:B)).

Repro (minimal)

import pandas as pd
from graphistry.compute import n, e_forward
from graphistry.tests.test_compute import CGFull
from tests.cypher_tck.parse_cypher import graph_fixture_from_create

fixture = graph_fixture_from_create(
    """
    CREATE (:A)-[:T1]->(:B),
           (:B)-[:T2]->(:A),
           (:B)-[:T3]->(:B),
           (:A)-[:T4]->(:A)
    """
)

g = CGFull()

nodes_df = pd.DataFrame(fixture.nodes)
labels = nodes_df.get("labels", pd.Series([], dtype=object)).tolist()
all_labels = sorted({label for labs in labels for label in (labs or [])})
for label in all_labels:
    nodes_df[f"label__{label}"] = [label in (labs or []) for labs in labels]

edges_df = pd.DataFrame(fixture.edges)

g = g.nodes(nodes_df, "id").edges(edges_df, "src", "dst", edge="edge_id")

g.gfql([n({"label__A": True}), e_forward(), n({"label__B": True})], engine="pandas")

Actual

GFQLSchemaError: [incompatible-column-type] Type mismatch: column "label__B" is string but filter value is numeric

Note: nodes_df.dtypes shows label__B as bool before the call.

Expected

The boolean label columns should remain boolean through the chain, and filtering on label__B=True should succeed.

Environment

• commit: 7b721c9
• engine: pandas
• branch: test/cypher-conformance

Context

This surfaced while translating openCypher TCK scenario Match2 [2]: MATCH (:A)-[r]->(:B) RETURN r.