Skip to content

Conversation

@lmeyerov
Copy link
Contributor

@lmeyerov lmeyerov commented Jan 16, 2026

Summary

  • replace merge-based seen-marker tracking in hop label paths with engine-native domains (Series/Index + isin)
  • avoids pandas Index conversions on cuDF while preserving label semantics
  • add hop label regression coverage and changelog entry

Before / After

  • Before (cuDF): hop labels could error or sync to host because label tracking used pandas Index conversions.
  • After: label tracking stays engine-native.

Example (GFQL)

from graphistry.compute.ast import n, e

# Labeling hop paths triggers the code path fixed here
result = g.chain([
    n(),
    e(hops=2, label_node_hops='nh', label_edge_hops='eh', label_seeds=True),
])

Scope

Affects only hop label tracking (when label_node_hops, label_edge_hops, label_seeds, output_min_hops/output_max_hops, or min_hops > 1 are used).

Testing

  • uv run env PYTHONPATH=. pytest graphistry/tests/compute/test_hop.py -q

@lmeyerov lmeyerov merged commit 12729d2 into master Jan 16, 2026
61 checks passed
@lmeyerov lmeyerov deleted the fix/hop-label-seen-ids-cudf branch January 16, 2026 06:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants