fix(gfql): fixes found by the route sweeps: source dtypes on every route, polars prune decline, polars-gpu fused lane (#2058, #2053, #2064)

[lmeyerov]

The three product fixes the route flow found, folded into one PR at the owner's request (formerly #2062 + #2065 + #2066). Stacked on #2054; top of the stack.

1. Every route keeps source dtypes (gfql: the general chain path upcasts integer node/property columns to float where every fast lane keeps int (route-dependent dtypes) #2058): one stub-row widening artifact (id-only endpoint rows concatenated then deduplicated) mirrored on purpose at three other seams. The endpoint closure appends only missing endpoints; the bindings row frame restores an alias property's node-table dtype once its rows hold no null; the 0-row connected-join path no longer mirrors the widening; the Cypher seeded lanes keep source dtypes on every route. Contract change: pandas RETURN projections and 0-row aggregates of int/bool properties carry the source dtypes; four pins restated. The cuDF 26.02 prune_to_endpoints pins pass on the real GPU at this tree (gfql cuDF 26.02 (real GPU): chain default route and policy-forced full path disagree on prune_to_endpoints hops and duplicate node ids; tests use the removed cudf.DataFrame.from_pandas #2043 prune half; marker dropped).
2. Polars declines prune_to_endpoints on single-hop edges (gfql polars: the plain single-hop chain branch ignores prune_to_endpoints (pandas keeps only the arrival side) #2053): typed NotImplementedError and predicate decline; variable-length hops keep their native pruning.
3. The fused grouped-aggregate lane keeps the group_by plan on the GPU target (gfql polars-gpu: the fused grouped-aggregate lane and the polars-gpu strictness tests fail on the RAPIDS 26.02 image (plan not GPU-executable on its cudf-polars) #2064): its value_counts alternative lowers to unnest, which cudf-polars has no implementation for; GPU receipt 607/5 vs master 520/91 on the two files.

Receipts on this head below; earlier receipts on the closed #2065 and #2066.

🤖 Generated with Claude Code

https://claude.ai/code/session_01QztW7jYsDd66e8rb8pJNQA

[lmeyerov]

GPU receipt (dgx-spark, graphistry/test-rapids-official:26.02-gfql-polars, cudf 26.02.01 / cupy 13.6.0 / polars 1.35.2, safe_run, TEST_CUDF=1; suites test_chain.py, test_hop.py, test_chain_alias_column_collision.py, tests/compute/gfql/**) on the stack head 676232a (#2062, which contains #2055, #2056, #2054 and #2061): 13383 passed / 93 failed / 55 skipped / 44 xfailed / 4 xpassed. The 93 failures are exactly master 86de0f5's set in the same image (13227 passed / 93 failed: the polars-gpu fused grouped-aggregate lane and strictness tests, #2064): set difference empty, zero branch-specific failures.

[lmeyerov]

Head moved 676232a → 640167c (squash, tree identical) → 203845b (the row-frame helper's annotations changed from Any to DataFrameT for the type-hygiene guard; no behavior change). The GPU receipt above stands for this tree.

[lmeyerov]

CI receipt at 203845b: 80 check-runs success, 2 path-skipped (tck-gfql, test-polars ×6 incl. coverage audit, gfql-benchmarks, lint/type/comment guards all green). Local broad CPU 12687 / cuDF 1351 / native-fast-off 549 with 0 failures; GPU receipt on the stack head above. Marking ready.

[lmeyerov]

GPU receipt (dgx-spark, graphistry/test-rapids-official:26.02-gfql-polars, cudf 26.02.01 / polars 1.35.2, safe_run, TEST_CUDF=1; suites test_chain.py, test_hop.py, test_chain_alias_column_collision.py, tests/compute/gfql/**) on the restacked stack head a54ac4b (#2065, containing #2055 rework, #2056, #2054, #2061, #2062): 13386 passed / 95 failed / 38 xfailed / 4 xpassed. 93 of the failures are master 86de0f5's polars-gpu fused-lane set (#2064, fixed separately in #2066). The other 2 are strict XPASS of the #2043 prune_to_endpoints markers on cuDF 26.02: the prune shapes now agree fast-vs-full on this stack (the #2062 closure fix removed the divergence), so #2062 drops that marker in a follow-up commit (tree otherwise identical; #2065 restacked).

[lmeyerov]

CI receipt at 7d3c7dc (restacked on the #2055 rework, #2043 prune marker dropped): 80 check-runs success, 2 path-skipped. GPU receipt on the stack head above (the two prune pins pass on cuDF 26.02 at this tree).

[lmeyerov]

GPU receipt (dgx-spark, graphistry/test-rapids-official:26.02-gfql-polars, cudf 26.02.01 / polars 1.35.2, safe_run, TEST_CUDF=1; suites test_chain.py, test_hop.py, test_chain_alias_column_collision.py, tests/compute/gfql/**) on the folded stack top 500265e (#2062, containing #2056 and #2054): 13532 passed / 7 failed / 37 xfailed / 4 xpassed, versus master 86de0f5's 13227 / 93 in the same image. Every one of the 7 is inside master's set (5 polars-gpu strictness cases and the routing-contract / indexed-bypass pins tracked on #2064 and #1803); zero branch-specific failures, and the fused-lane fix folded from #2066 clears the other 86.

[lmeyerov]

CI receipt at 500265e (folded: #2058 dtypes + #2053 polars prune decline + #2064 fused lane): 80 check-runs success, 2 path-skipped. Local on this head: broad CPU + cuDF 12939 / 0; tck-gfql 4143 / 689 xfail; GPU receipt above (13532 / 7, all inside master's set).

[lmeyerov]

Rebased onto the new #2054 head b5ba132 → c2443da (clean rebase; own delta identical before and after, 244/244 lines). Gates on the new head: routes harness, specialization packages, collision matrix, #2058 dtype pins, fused polars lane, lane completeness, chain (788 pass / 10 xfail incl. cuDF). CI 80/80 (+2 filter-skips run locally in that set).

[lmeyerov]

CI on ad6dc05: 80/80 green (+2 filter-skips run locally). Head is final behind #2054.

[lmeyerov]

Rebased onto master c2d3a07 (#2054 merged) → 6910594; clean rebase, own delta identical (0 drift lines). Gates on the new head: routes harness incl. frame variants, both specialization packages, #2058 dtype pins, fused polars lane, lane completeness: 745 pass / 37 xfail incl. cuDF; lint clean. CI re-fired.

[lmeyerov]

CI on 6910594: 82/82 green (+2 filter-skips run locally). Ready to merge.