Summary

DataStore.explain() was labeling GroupBy / groupby().apply() ops as 🐼 [Pandas] even when they were clearly pushed down to chDB — the segment header said [chDB] and the generated SQL contained GROUP BY, but the per-op line lied.

Root cause

The base LazyOp.execution_engine() contract (lazy_ops.py:113-126) requires returning the canonical literals "chDB" or "Pandas". DataStore.explain() (core.py:801-816) only branches on "chDB"; anything else falls into the Pandas branch.

Two implementations returned the non-canonical literal "SQL":

• LazyGroupByAgg.execution_engine() always returned "SQL".
• LazyApply.execution_engine() returned "SQL" in its pushable branch.

So pushed-down GroupBy and apply ops silently rendered as Pandas. Display-only bug — execution path was unaffected; no consumer was matching on "SQL".

Reproduction (before fix)

ds = DataStore.from_file('amazon_sample.parquet')
res = (
    ds[ds['verified_purchase'] == True]
    .groupby('product_category')
    .agg({'star_rating': ['mean', 'count']})
)
res.explain()

Before:

    Segment 1 [chDB] (from source): Operations 2-5
 [2] 🚀 [chDB] WHERE: "verified_purchase" = TRUE
 [3] 🐼 [Pandas] GroupBy(['product_category']).{'star_rating': ['mean', 'count']}()   <-- wrong
 ...
SELECT "product_category", avg("star_rating") AS "mean", ...
  FROM file('amazon_sample.parquet', 'Parquet') WHERE ... GROUP BY ...                <-- chDB!

After:

    Segment 1 [chDB] (from source): Operations 2-3
 [2] 🚀 [chDB] WHERE: "verified_purchase" = TRUE
 [3] 🚀 [chDB] GroupBy(['product_category']).{'star_rating': ['mean', 'count']}()

Changes

• datastore/lazy_ops.py: LazyGroupByAgg.execution_engine() and LazyApply.execution_engine() now return "chDB" / "Pandas" based on can_push_to_sql().
• datastore/tests/test_groupby_apply_sql_pushdown.py: two existing assertions had baked in the buggy "SQL" literal — updated to "chDB".
• datastore/tests/test_explain_method.py: new TestExplainEngineLabelingForPushedOps class with 5 regression tests covering the original user scenario (agg dict-of-list), simple .mean() agg, the non-pushable case (must stay Pandas), groupby().apply(sum) pushdown, and a direct unit test on the canonical-literal contract.

Test plan

• pytest datastore/tests/test_explain_method.py — 22 passed
• pytest datastore/tests/test_groupby_apply_sql_pushdown.py — 28 passed
• pytest test_explain_segmented_execution.py test_comprehensive_segmented_execution.py test_sql_vs_pandas_filter.py test_lazy_chain_engine_verification.py test_lazy_execution.py test_case_when.py test_nullable_sql_pushdown.py — 203 passed, 1 xfailed (pre-existing, unrelated)
• Manually verified the original user scenario now labels GroupBy as [chDB].