Repository navigation

feat(gfql): WHERE clause with df_executor - #886

Closed

lmeyerov wants to merge 195 commits into

masterfrom

feat/where-clause-executor

Closed

feat(gfql): WHERE clause with df_executor#886
lmeyerov wants to merge 195 commits into
masterfrom
feat/where-clause-executor

Commits on Jan 27, 2026
- feat(gfql): add WHERE clause and df_executor (stacked PR)
  
  lmeyerov
  and
  claude
  committed
  c22e0f9
- fix(tests): skip oracle tests for multi-hop + WHERE limitations
  
  lmeyerov
  and
  claude
  committed
  5f1e9d9
- docs(changelog): add WHERE feature and bugfix entries for PR 886
  lmeyerov
  committed
  d68496c
- test(gfql): restore df_executor profiling scripts
  lmeyerov
  committed
  9afda91
- fix(enumerator): restore source/destination_node_match filter support
  lmeyerov
  committed
  0c9739c
- fix(enumerator): restore full pre-split functionality and remove test skips
  lmeyerov
  committed
  528783b
- docs(changelog): restore missing cuDF same-path and test entries
  lmeyerov
  committed
  472d672
- docs(changelog): restore from_json where validation fix entry
  lmeyerov
  committed
  a0b00bb
- docs(changelog): move WHERE entries to Development section
  
  lmeyerov
  and
  claude
  committed
  2f69034
- fix(df_executor): fix off-by-one in _bfs_reachability max_hops
  
  lmeyerov
  and
  claude
  committed
  17765cd
- test(gfql): add requires_gpu decorator for proper GPU test skipping
  
  lmeyerov
  and
  claude
  committed
  b234063
- refactor(gfql): extract ChainMeta for O(1) chain lookups
  
  lmeyerov
  and
  claude
  committed
  67b6aae
- refactor(gfql): extract EdgeSemantics for direction handling
  
  lmeyerov
  and
  claude
  committed
  3688ddc
- refactor(gfql): extract df_utils for DataFrame operations
  
  lmeyerov
  and
  claude
  committed
  7c121f8
- refactor(gfql): use EdgeSemantics in multihop methods, remove deprecated _build_edge_pairs
  
  lmeyerov
  and
  claude
  committed
  e7d0924
- refactor(gfql): use EdgeSemantics in _filter_edges_by_clauses
  
  lmeyerov
  and
  claude
  committed
  bdcb667
- refactor(gfql): extract BFS functions to same_path/bfs.py
  
  lmeyerov
  and
  claude
  committed
  b4d594c
- refactor(gfql): remove redundant _is_single_hop, use EdgeSemantics.is_multihop
  
  lmeyerov
  and
  claude
  committed
  68315ac
- refactor(gfql): extract post-prune methods to same_path/post_prune.py
  
  lmeyerov
  and
  claude
  committed
  c825918
- refactor(gfql): extract multihop methods to same_path/multihop.py
  
  lmeyerov
  and
  claude
  committed
  4b64cc8
- refactor(gfql): extract _re_propagate_backward to post_prune module
  
  lmeyerov
  and
  claude
  committed
  abb3a45
- refactor(gfql): extract WHERE edge filtering to same_path/where_filter.py
  
  lmeyerov
  and
  claude
  committed
  b1b63b9
- refactor(gfql): delete dead code and unused optimizations
  
  lmeyerov
  and
  claude
  committed
  9b9bbcf
- fix(gfql): add forward WHERE pruning to df_executor
  
  lmeyerov
  and
  claude
  committed
  70275ca
- refactor(gfql): remove unused imports from df_executor
  
  lmeyerov
  and
  claude
  committed
  25b280f
- refactor(gfql): unify NULL semantics in evaluate_clause()
  
  lmeyerov
  and
  claude
  committed
  d7218bc
- refactor(gfql): move re_propagate_backward to executor method
  
  lmeyerov
  and
  claude
  committed
  99334ce
- fix(cudf): comprehensive cuDF compatibility fixes for GFQL executor
  
  lmeyerov
  and
  claude
  committed
  3018a40
- refactor(cudf): consolidate DataFrame construction helpers in df_utils
  
  lmeyerov
  and
  claude
  committed
  8db9f11
- fix(cudf): fix cuDF compatibility in chain backward pass and cross-engine coercion
  
  lmeyerov
  and
  claude
  committed
  d8d40b1
- docs(changelog): add WHERE clause feature entries
  
  lmeyerov
  and
  claude
  committed
  e5df51a
- fix(tests): cuDF compatibility for tolist() calls in chain optimization tests
  
  lmeyerov
  and
  claude
  committed
  2750990
- fix(cudf): use module string checks for cross-engine coercion
  
  lmeyerov
  and
  claude
  committed
  841e329
- refactor(gfql): remove dead SamePathPlan code (~80 LOC)
  
  lmeyerov
  and
  claude
  committed
  62f5d28
- refactor(gfql): remove unused same_path/__init__.py (~35 LOC)
  
  lmeyerov
  and
  claude
  committed
  64b0e56
- refactor(gfql): add immutable PathState type (Phase 1)
  
  lmeyerov
  and
  claude
  committed
  400a5bc
- refactor(gfql): _backward_prune tracks pruned edges separately (Phase 2a)
  
  lmeyerov
  and
  claude
  committed
  c1e42d7
- refactor(gfql): backward_propagate_constraints uses local state (Phase 2b)
  
  lmeyerov
  and
  claude
  committed
  dc3a85c
- refactor(gfql): post_prune.py uses local state copies (Phase 2c)
  
  lmeyerov
  and
  claude
  committed
  74fe617
- refactor(gfql): add edges_df_for_step accessor (Phase 2d)
  
  lmeyerov
  and
  claude
  committed
  5d95f13
- test(gfql): add PathState immutability unit tests (Phase 3)
  
  lmeyerov
  and
  claude
  committed
  5a4f509
- refactor(gfql): Phase 4 - convert to pure PathState API
  
  lmeyerov
  and
  claude
  committed
  bd47ba2
- refactor(gfql): Phase 5 - remove old _PathState class and update docstrings
  
  lmeyerov
  and
  claude
  committed
  2441e6c
- test(gfql): Phase 6 - add PathState immutability contract tests
  
  lmeyerov
  and
  claude
  committed
  2aa801c
- fix(cudf): use series_values helper instead of .tolist() in backward_propagate_constraints
  
  lmeyerov
  and
  claude
  committed
  c72bede
- chore: remove redundant comments from post_prune.py
  
  lmeyerov
  and
  claude
  committed
  2251716
- perf(gfql): replace Set with pd.Index for 7x faster ID operations
  
  lmeyerov
  and
  claude
  committed
  bdfa72c
- perf(chain): use .isin() instead of merge for endpoint filtering
  
  lmeyerov
  and
  claude
  committed
  af81ab6
- revert: remove cuDF dtype coercion from safe_merge
  
  lmeyerov
  and
  claude
  committed
  41fccdc
- perf(hop): use .isin() instead of merge for wavefront->edges join (8x faster)
  
  lmeyerov
  and
  claude
  committed
  607fa71
- refactor(hop): remove unused prepare_merge_dataframe function
  
  lmeyerov
  and
  claude
  committed
  3001d54
- perf(hop): precompute node predicate domains
  lmeyerov
  committed
  8f32b95
- perf(hop): unify direction pairs; modest CPU gains
  lmeyerov
  committed
  251f83e
- perf(hop): mask target/dest filters with isin
  lmeyerov
  committed
  b6ec416
- perf(hop): precompute target wavefront domains
  lmeyerov
  committed
  f201041
- perf(hop): use merge for EDGE_ID joins
  lmeyerov
  committed
  75c9180
- perf(df_executor): DF-native cuDF forward prune
  lmeyerov
  committed
  515ad7e
- perf(hop): undirected single-pass expansion
  lmeyerov
  committed
  1c80ef9
- perf(hop): domain-based fast path traversal
  lmeyerov
  committed
  e1c5347
- fix(hop): undirected single-pass frontier
  lmeyerov
  committed
  dc6125e
- fix(gfql): use domain helpers for same-path ids
  lmeyerov
  committed
  073f9a4
- fix(docs): include gfql same_path package in build
  lmeyerov
  committed
  6883f84
- fix(lint): clean unused imports and f-string
  lmeyerov
  committed
  174d600
- fix(mypy): narrow optional frames in hop/gfql
  lmeyerov
  committed
  0a80346
- fix(mypy): drop dataclass slots for py3.9
  lmeyerov
  committed
  79afdbf
- fix(mypy): avoid optional node cols in hop/bfs
  lmeyerov
  committed
  cb92e68
- Fix hop pair typing without asserts
  lmeyerov
  committed
  c82f901
- Re-export col/compare in reference enumerator
  lmeyerov
  committed
  519bc40
- Expose col/compare without flake8 shadowing
  lmeyerov
  committed
  9cd16e7
- DRY hop edge expansion helpers
  lmeyerov
  committed
  a71213c
- Avoid re-deduping matches in hop
  lmeyerov
  committed
  c3e95a4
- Revert hop undirected fast path and match anti-join
  lmeyerov
  committed
  acf3568
- Add tracked benchmark scripts and docs
  lmeyerov
  committed
  8b30023
- Expand real-data benchmark coverage
  lmeyerov
  committed
  5dabfd0
- Add benchmark results log
  lmeyerov
  committed
  fada9ea
- Add real-data WHERE benchmark scenarios
  lmeyerov
  committed
  7f581dd
- Log real-data WHERE benchmark run
  lmeyerov
  committed
  c4d4290
- Add scores to real-data benchmark output
  lmeyerov
  committed
  984442d
- Log real-data benchmark scores
  lmeyerov
  committed
  3876abf
- Log redteam benchmark rerun
  lmeyerov
  committed
  1e38bbd
- Add redteam categorical benchmark option
  lmeyerov
  committed
  4c9e908
- Add optional df_executor OTel spans
  lmeyerov
  committed
  d3af7d1
- Add OTel detail stats for df_executor
  lmeyerov
  committed
  c74bb9c
- Add non-adjacent OTel detail stats
  lmeyerov
  committed
  a64ddda
- benchmarks: add optional otel setup
  lmeyerov
  committed
  7330896
- otel: core helper, spans, trace headers
  lmeyerov
  committed
  ed0bdfe
- tests: assert traceparent headers
  lmeyerov
  committed
  2b1ba36
- tests: behavior-level trace headers
  lmeyerov
  committed
  d757293
- benchmarks: log fast-path A/B; hop: clarify toggle
  lmeyerov
  committed
  7aff2cc
- experiments: add non-adjacent WHERE modes
  lmeyerov
  committed
  a6f12ce
- experiments: add non-adj ordering/bounds
  lmeyerov
  committed
  8aac3b8
- benchmarks: log non-adj experiment
  lmeyerov
  committed
  1e65099
- gfql: limit value-mode to equality; log phase-18 runs
  lmeyerov
  committed
  eb31425
- fix(gfql): honor node-id WHERE in adjacent filters
  lmeyerov
  committed
  cf99bc8
- bench: log phase-19 where opt results
  lmeyerov
  committed
  2788bc0
- bench: add low-card nonadj stress cases
  lmeyerov
  committed
  e5eebf4
- bench: log low-card nonadj stress results
  lmeyerov
  committed
  7dd85db
- bench: add realdata where stress cases and timeouts
  lmeyerov
  committed
  8e5076a
- bench: log realdata timeout stress results
  lmeyerov
  committed
  4c31419
- fix(gfql): null-safe where + singleton prefilter
  lmeyerov
  committed
  8b8e520
- bench: log phase-22 timeout results
  lmeyerov
  committed
  d808476
- feat(gfql): allow value-mode on selected ops
  lmeyerov
  committed
  7686293
- docs(bench): log phase-23 value-mode ops results
  lmeyerov
  committed
  f825bba
- docs(bench): log phase-24 realdata baseline
  lmeyerov
  committed
  13c29ee
- feat(gfql): dynamic non-adj clause ordering
  lmeyerov
  committed
  f0cee3b
- docs(bench): log phase-25 ordering results
  lmeyerov
  committed
  a28942f
- revert(gfql): drop dynamic non-adj ordering
  lmeyerov
  committed
  112447e
- feat(gfql): group non-adj clauses by endpoints
  lmeyerov
  committed
  eca3028
- docs(bench): log phase-26 grouping results
  lmeyerov
  committed
  3b71dd8
- feat(gfql): add pair-gated non-adj clause option
  lmeyerov
  committed
  90dc129
- docs(bench): log phase 27 pair-gate results
  lmeyerov
  committed
  f917ac2
- Revert "feat(gfql): add pair-gated non-adj clause option"
  lmeyerov
  committed
  25ae226
- feat(gfql): prefilter multi-eq non-adj clauses
  lmeyerov
  committed
  5169804
- docs(bench): log phase 28 multi-eq prefilter results
  lmeyerov
  committed
  9289fd8
- Revert "feat(gfql): prefilter multi-eq non-adj clauses"
  lmeyerov
  committed
  b80e329
- feat(gfql): composite value-mode for multi-eq non-adj
  lmeyerov
  committed
  38fc2b8
- docs(bench): log phase 29 composite value-mode
  lmeyerov
  committed
  30a0acf
- feat(gfql): add capped vector non-adj strategy
  lmeyerov
  committed
  28ff887
- chore(bench): log phase 30 vector results
  lmeyerov
  committed
  d07c983
- feat(gfql): intersect vector clauses by adjacency
  lmeyerov
  committed
  9035c60
- chore(bench): log phase 31 vector results
  lmeyerov
  committed
  b0412c7
- test(gfql): add vector parity coverage
  lmeyerov
  committed
  cf6790a
- feat(gfql): gate vector paths on mid intersection
  lmeyerov
  committed
  9a0d06c
- feat(gfql): add value-aware vector path join
  lmeyerov
  committed
  294e5ea
- chore(bench): log phase 33.1 vector results
  lmeyerov
  committed
  5b16b21
- Revert "feat(gfql): add value-aware vector path join"
  lmeyerov
  committed
  d6ac7fa
- checkpoint: auto mode + edge fast path benchmarks
  lmeyerov
  committed
  aa05e39
- benchmarks: add optional kuzu comparisons
  lmeyerov
  committed
  8c9c259
- benchmarks: add graph/scenario filters
  lmeyerov
  committed
  4c1d47b
- benchmarks: handle kuzu db path variants
  lmeyerov
  committed
  4f30a6e
- benchmarks: add WHERE opt matrix runner
  lmeyerov
  committed
  7ab8652
- benchmarks: seed support + multi-eq semijoin flag
  lmeyerov
  committed
  620d9a1
- feat(gfql): default non-adj WHERE auto mode
  lmeyerov
  committed
  55454b4
- bench: add timeout repro harness
  lmeyerov
  committed
  7393994
- feat(gfql): guard auto value mode on multi-clause
  lmeyerov
  committed
  cdca910
- feat(gfql): add opt-in inequality aggregation
  lmeyerov
  committed
  def5ad2
- feat(gfql): gate ineq aggregation by label
  lmeyerov
  committed
  32fe264
- feat(gfql): gate ineq aggregation on label eq
  lmeyerov
  committed
  2f1d51f
- perf(gfql): avoid semijoin pair build when inactive
  lmeyerov
  committed
  2e40f30
- chore(gfql): add otel size counters
  lmeyerov
  committed
  4d4e9ba
- fix(gfql): init vector guard domains
  lmeyerov
  committed
  bea8067
- perf(gfql): reduce semijoin dedup overhead
  lmeyerov
  committed
  762d37c
- perf(gfql): cache edge pairs for semijoins
  lmeyerov
  committed
  c483a6c
- docs(changelog): note WHERE perf + otel updates
  lmeyerov
  committed
  91baa3d
- style: fix flake8 spacing
  lmeyerov
  committed
  865d9f4
- chore: tidy hop/df_executor imports
  lmeyerov
  committed
  7acf5de
- chore: tighten gfql typing
  lmeyerov
  committed
  026295f
- chore: reuse cudf checks
  lmeyerov
  committed
  2411334
- chore: simplify gfql where parsing
  lmeyerov
  committed
  d29e074
- chore: drop stale hop TODOs
  lmeyerov
  committed
  d9254c1
- chore: tighten domain helpers
  lmeyerov
  committed
  f8f56c2
- style: fix test_str spacing
  lmeyerov
  committed
  889c0a0
- chore: fix post_prune typing
  lmeyerov
  committed
  e5f96ad
- benchmarks: add graph-benchmark q1-q9 harness
  lmeyerov
  committed
  bde816c
- benchmarks: fix interest edge filename
  lmeyerov
  committed
  53343e6
- benchmarks: log graph-benchmark q1-q9 baseline
  lmeyerov
  committed
  7e0d890
- benchmarks: add preindexed graph-benchmark mode
  lmeyerov
  committed
  fd112f4
- benchmarks: add presorted graph-benchmark mode
  lmeyerov
  committed
  d4b9843
- benchmarks: add preindex timing for graph-benchmark
  lmeyerov
  committed
  341f294
- benchmarks: log preindex timing results
  lmeyerov
  committed
  d1ed642
- refactor: trim same_path slop
  lmeyerov
  committed
  59a7ffd
- test: trim df_executor core slop
  lmeyerov
  committed
  31a2f2d
- test: trim df_executor pattern slop
  lmeyerov
  committed
  f1c14e0
- test: trim df_executor dimension slop
  lmeyerov
  committed
  4cf89b3
- test: trim df_executor amplify slop
  lmeyerov
  committed
  f74924c
- test: trim path_state slop
  lmeyerov
  committed
  364bff8
- test: drop df_executor core docstrings
  lmeyerov
  committed
  35b2391
- test: drop df_executor pattern docstrings
  lmeyerov
  committed
  3065324
- refactor: drop df_executor docstrings
  lmeyerov
  committed
  9c3fa05
- refactor: drop same_path docstrings
  lmeyerov
  committed
  2b24622
- Trim gfql ref test slop
  lmeyerov
  committed
  872c89b
- Trim compute test docstrings
  lmeyerov
  committed
  9a01918
- Trim gfql_unified comments
  lmeyerov
  committed
  e5ddc90
- Trim remote helper comments
  lmeyerov
  committed
  dfbc36c
- Trim chain/hop comment slop
  lmeyerov
  committed
  9d9e946
- Trim df_executor comments
  lmeyerov
  committed
  2cb7ba1
- Trim where_filter comments
  lmeyerov
  committed
  d148f79
- Trim df_utils/multihop comments
  lmeyerov
  committed
  a5d7027
- Trim redundant ArrowFileUploader comment
  lmeyerov
  committed
  05f0463
- Deduplicate edge semantics endpoint cols
  lmeyerov
  committed
  0064361
- Tighten same-path domain typing
  lmeyerov
  committed
  aea0c44
- Tighten PathState sync typing
  lmeyerov
  committed
  6ac6a3a
- Tighten multihop domain typing
  lmeyerov
  committed
  2095de2
- Type post-prune allowed_edges as DomainT
  lmeyerov
  committed
  9d44df5
- Remove unused multihop visited tracking
  lmeyerov
  committed
  fb6e129
- Reduce post-prune duplication and fix typing
  lmeyerov
  committed
  7c88702
- Simplify same-path pruning helpers
  lmeyerov
  committed
  34fa13e
- Compact post-prune tracing and helpers
  lmeyerov
  committed
  982b763
- Reduce post-prune and where-filter duplication
  lmeyerov
  committed
  2002828
- Collapse post-prune pruning patterns
  lmeyerov
  committed
  12a6249
- Consolidate post-prune allowed-node updates
  lmeyerov
  committed
  e0895c8
- Refine post-prune allowed updates
  lmeyerov
  committed
  ecaefc6
- Reduce post-prune inequality duplication
  lmeyerov
  committed
  4c14280
- Reduce post_prune duplication
  lmeyerov
  committed
  1aee0a3
- Reduce post_prune duplication further
  lmeyerov
  committed
  66337e0

Commits

Commits on Jan 27, 2026