Repository navigation
Cypher engine hardcodes id as the node-identity column #1490
Description
Summary
The GFQL Cypher lowering assumes the node-identity column is literally named
id. It does not respect the node-id column bound via g.nodes(df, node_col),
and it cannot distinguish a node's identity from a Cypher property that
happens to be named id.
Surfaced while working graphistry/tck-gfql#126 (the TCK harness wanted to model
a node with an id property distinct from its synthetic identity, per the
openCypher with2-1 setup CREATE (a:End {num: 42, id: 0}), ...).
Where id is hardcoded
graphistry/compute/gfql/cypher/lowering.py:
- L3456 — repeated node-variable dedup:
compare(col(alias, "id"), "==", col(rewritten_alias, "id")) - L1427 — connected-join alias identity:
prop = "id" if isinstance(target, ASTNode) else ... - L6549, L6623 — aggregate / alias identity helpers:
return "id"forASTNode.
When the bound node-id column is not named id, repeated-variable patterns
(e.g. MATCH (n)-[]->(n)) fail validation:
GFQLValidationError: [unsupported-cypher-query] WHERE references missing
column 'id' on alias 'n' (node, step=2). Available columns: __node_id__, ...
Impact / two facets
- Bound node-id column ignored.
g.nodes(df, "__node_id__").gfql("<cypher>")
with a repeated node variable raises E108 even though the graph is valid. - Identity vs.
idproperty conflated. Because identity is theid
column, a node that genuinely has a Cypher property namedid(distinct
from identity) cannot be represented — the renderer only surfacesidas a
property via a numeric-dtype heuristic
(row/entity_props.py::_include_numeric_id_as_property), which breaks on
mixed-dtype id columns.
Suggested fix
Thread the graph's bound node-id column name through Cypher lowering instead of
the "id" string literal (and likewise for edges, which already use
__gfql_edge_index_0__). Longer term, model node identity as an internal
column separate from user-visible properties so a Cypher id property is
representable.
Links
- graph_fixture_from_create: unresolved property references and id/identity-column conflation tck-gfql#126 — TCK-harness fixture parser issue; its identity/
id
separation half is blocked by this. - Direct-Cypher residual: reconcile with2-1 setup/join contract tck-gfql#115 — the
with2-1residual that surfaced it (worked
around harness-side with an explicit fixture).
Activity
lmeyerov commentedon May 17, 2026
tck-gfql #126 now has an implementation PR that adopts reserved harness-side node identity and records the resulting direct-Cypher fallout:
testspassed in 5m35s)4011 passed, 724 xfailedRelevant count changes from the corrected fixture:
933 (rows 794, errors 139)->926 (rows 787, errors 139)2893 / 3627 (79.8%)->2886 / 3627 (79.6%)supported 2910, xfail 717->supported 2903, xfail 724Seven scenarios returned from direct-Cypher promotion to xfail because
graph_fixture_from_createno longer exposes synthetic node identity as data columnid:match7-24,return-orderby2-6,expr-graph3-5,expr-graph7-1,usecase-countingsubgraphmatches1-6,usecase-countingsubgraphmatches1-7,usecase-countingsubgraphmatches1-9.This is the expected evidence for this pygraphistry issue: tck-gfql can now model Cypher
idseparately from fixture identity, and the remaining failures are direct-Cypher lowering/runtime assumptions about identity column name. No pygraphistry runtime changes were made in the tck-gfql PR.lmeyerov commentedon May 17, 2026
tck-gfql #126 is now merged via PR #131:
b0e1b59673eb6b3654ed1b8e45c51d861b07b2b8testspassed in 5m32sThe merged tck-gfql baseline keeps the corrected harness fixture model and records the pygraphistry fallout here:
926 (rows 787, errors 139)2886 / 3627 (79.6%)supported 2903, xfail 724match7-24,return-orderby2-6,expr-graph3-5,expr-graph7-1,usecase-countingsubgraphmatches1-6,usecase-countingsubgraphmatches1-7,usecase-countingsubgraphmatches1-9No pygraphistry runtime changes landed in that PR; this issue remains the upstream follow-up for direct-Cypher lowering honoring the bound node-id column separately from a Cypher
idproperty.lmeyerov commentedon May 17, 2026
Closeout after conflict resolution/merge:
origin/masterand resolving theCHANGELOG.mdconflict.python3 bin/ci_cypher_surface_guard.py: passed (lowering.py9045/9045).idas the node-identity column #1490 regressions:3 passed, 1 skipped, 4 warnings.mainmoved, keeping Bug: Spigo demo in Python3 #129 promotions (match5-8,with5-2) promoted and tracking only the Cypher engine hardcodesidas the node-identity column #1490 residual drift.1 passed.2 passed.722 passed.tests: passed in 5m41s.Issue is closed by #1504. Remaining direct-Cypher/TCK work should be recalculated from pygraphistry
masterplus tck-gfqlmain, not the pre-#1490 drift snapshot.perf(gfql): projection-pushdown for binding tables — skip unused prop…
perf(gfql): projection-pushdown for binding tables — skip unused prop…
perf(gfql): projection-pushdown for binding tables — skip unused prop…
perf(gfql): projection-pushdown for binding tables — skip unused prop…
perf(gfql): projection-pushdown for binding tables — skip unused prop…
perf(gfql): projection-pushdown for binding tables — skip unused prop…
perf(gfql): projection-pushdown for binding tables — skip unused prop…
perf(gfql): projection-pushdown for binding tables — skip unused prop…