mikebridge commented on code in PR #45071:
URL: https://github.com/apache/superset/pull/45071#discussion_r4235494094
##########
superset/commands/semantic_layer/delete.py:
##########
@@ -32,11 +39,214 @@
)
from superset.commands.utils import current_user_can_modify_object
from superset.daos.semantic_layer import SemanticLayerDAO, SemanticViewDAO
+from superset.dashboards.filters import DashboardAccessFilter
from superset.exceptions import SupersetSecurityException
+from superset.models.dashboard import Dashboard, dashboard_slices
+from superset.models.slice import Slice
+from superset.reports.filters import ReportScheduleFilter
+from superset.reports.models import ReportSchedule
from superset.semantic_layers.models import SemanticLayer, SemanticView
+from superset.utils import json
from superset.utils.decorators import on_error, transaction
logger = logging.getLogger(__name__)
+_DEPENDENT_LIMIT: int = 20
+
+
+def _semantic_target_matches(target: object, view_ids: set[int]) -> bool:
+ """A missing type denotes a table, even when its integer ID collides."""
+ if not isinstance(target, dict) or target.get("datasourceType") !=
"semantic_view":
+ return False
+ raw_id: object = target.get("datasetId")
+ if isinstance(raw_id, bool) or not (
+ isinstance(raw_id, int) or (isinstance(raw_id, str) and
raw_id.isdecimal())
+ ):
+ return False
+ try:
+ return int(raw_id) in view_ids
+ except ValueError:
+ return False
+
+
+def _dashboard_targets_view(
+ metadata: str, view_ids: set[int], dashboard_id: int
+) -> bool:
+ """Match typed semantic targets in native filters or display controls."""
+ try:
+ decoded: object = json.loads(metadata)
+ except (TypeError, ValueError, RecursionError):
+ logger.debug("Skipping malformed targets on dashboard %s",
dashboard_id)
+ return False
+ if not isinstance(decoded, dict):
+ logger.debug("Skipping non-object targets on dashboard %s",
dashboard_id)
+ return False
+ for key in ("native_filter_configuration", "chart_customization_config"):
+ controls: object = decoded.get(key)
+ if not isinstance(controls, list):
+ continue
+ for control in controls:
+ if not isinstance(control, dict):
+ continue
+ targets: object = control.get("targets")
+ if isinstance(targets, list) and any(
+ _semantic_target_matches(target, view_ids) for target in
targets
+ ):
+ return True
+ return False
+
+
+def _dashboard_target_ids(view_ids: sa.Select) -> set[int]:
+ """Find live dashboards with semantic targets after a coarse SQL
prefilter."""
+ selected_ids: set[int] = set(db.session.execute(view_ids).scalars())
+ if not selected_ids:
+ return set()
+ candidates: sa.Select = sa.select(Dashboard.id,
Dashboard.json_metadata).where(
+ Dashboard.deleted_at.is_(None),
+ Dashboard.json_metadata.like("%semantic_view%"),
+ )
Review Comment:
Agree with @aminghadersohi. The scan only runs when someone deletes a
semantic source, which is a rare admin action, and native-filter targets are
only stored in unindexed `json_metadata`, so a coarse `LIKE` prefilter streamed
with `yield_per` is about the cheapest option. Resolving.
--
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]
