gabotorresruiz commented on code in PR #44009:
URL: https://github.com/apache/superset/pull/44009#discussion_r4030310558
##########
superset/versioning/changes/listener.py:
##########
@@ -395,7 +408,115 @@ def _persist_buffered_records(
incr_capture_error("bulk_insert")
-def register_change_record_listener() -> None: # noqa: C901
+def finalize_change_records(session: Session) -> None:
+ """Build and persist the transaction's change records at commit time.
+
+ Module-level (rather than a closure inside the registration function)
+ so the capture write path can be exercised directly by unit tests
+ against an isolated session; it depends only on the session and the
+ module helpers, never on the registered entity classes.
+ """
+ if session.in_nested_transaction() or session.info.get(_FINALIZING_KEY):
+ return
+
+ session.info[_FINALIZING_KEY] = True
+ # Measures the FINALIZE stage only: the timer starts after the flush,
+ # which excludes the transaction's own write cost but also excludes
+ # capture_initial_states' per-entity pre-state SELECTs (those are timed
+ # as their own ``capture_initial_states`` stage in before_flush) — and
+ # runs through every capture step and early return. Every commit on the
+ # session emits a sample, including commits touching no versioned
+ # entity, because the whole-listener overhead is exactly what the
+ # kill-switch removes; a flush that raises emits nothing.
+ start: float | None = None
+ try:
+ session.flush()
+ start = perf_counter()
+ initial_states: dict[tuple[str, int], tuple[Any, dict[str, Any]]] = (
+ session.info.get(_INITIAL_STATES_KEY, {})
+ )
+ buffer = _build_scalar_buffer(initial_states)
+
+ try:
+ tx_id = _current_transaction_id(session)
+ except Exception: # pylint: disable=broad-except
+ logger.exception("version_changes: transaction lookup failed")
+ incr_capture_error("transaction_lookup")
+ return
+ if tx_id is None:
+ return
+
+ _stamp_action_kind_on_transaction(session, tx_id)
+ _append_child_records_to_buffer(session, tx_id, buffer)
+ _inject_action_meta_record(session, buffer)
+
+ if buffer:
+ _persist_buffered_records(session, tx_id, buffer)
+ finally:
+ session.info.pop(_FINALIZING_KEY, None)
+ if start is not None:
+ emit_capture_timing("finalize", (perf_counter() - start) * 1000.0)
+
+
+def _capture_initial_states(
+ session: Session, versioned_classes: tuple[type, ...]
+) -> None:
+ """Retain dirty versioned entities' pre-flush states for the final diff.
+
+ Timed as its own metric stage. The per-entity pre-state SELECTs issued
+ here are the capture cost that scales with the number of dirty versioned
+ entities — on a bulk edit plausibly the dominant cost the kill-switch
+ removes — and they run before the flush, outside ``finalize``'s timer. A
+ sample is emitted only when at least one pre-state read was attempted,
+ including failed reads that return no state,
+ so the many unrelated autoflushes do not flood the series with empty
+ samples; together with ``finalize`` the two stages cover the whole
+ listener. Module-level (not the registered closure) so it is
+ unit-testable without ``db.session``.
+ """
+ initial_states: dict[tuple[str, int], tuple[Any, dict[str, Any]]] = (
+ session.info.setdefault(_INITIAL_STATES_KEY, {})
+ )
+ start: float = perf_counter()
+ attempted: bool = False
+ obj: Any
+ try:
+ for obj in list(session.dirty):
+ if isinstance(obj, versioned_classes):
+ key: tuple[str, int] | None = _uncaptured_identity(obj,
initial_states)
+ if key is None:
+ continue
Review Comment:
Re-checked at `d1af084d`: `superset/versioning/changes/listener.py` is byte
identical to `e0aafeae3a` (same blob `8574895887`), so the only push since my
last pass is the master merge and this one is still open in code.
Same probe at the new head, three dirty entities with a `RuntimeError`
raised from the `id` property on the second: attempted `[1]`, retained `[1]`,
entity 3's pre-state gone. So yes to @aminghadersohi's guard as written.
One note on the test: `test_initial_state_capture_isolates_each_entity`
raises inside `capture_initial_state`, which the inner `try` already covers, so
it passes with or without the guard. The new case needs the raise to come from
the identity lookup itself for it to pin anything.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
