sadpandajoe commented on code in PR #45681:
URL: https://github.com/apache/superset/pull/45681#discussion_r4383273338
##########
superset/models/helpers.py:
##########
@@ -3994,6 +4019,201 @@ def get_rendered_sql(
def text(self, clause: str) -> TextClause:
return self.db_engine_spec.get_text_clause(clause)
+ def _get_rls_schema(self) -> str:
+ """Resolve RLS tables with the same schema policy as SQL
expressions."""
+ return get_rls_schema(
+ self.database,
+ self.catalog,
+ self.schema,
+ self.sql or "",
+ self.template_params_dict,
+ )
+
+ def _get_filter_column(
+ self,
+ name: str,
+ columns_by_name: dict[str, "TableColumn"],
+ ) -> Optional["TableColumn"]:
+ """Resolve real names before the first matching verbose filter
label."""
+ col = columns_by_name.get(name)
+ if col is not None:
+ return col
+ return next((col for col in self.columns if col.verbose_name == name),
None)
+
+ def _get_query_sql_expressions( # noqa: C901
+ self,
+ query_obj: QueryObjectDict,
+ template_processor: BaseTemplateProcessor,
+ granularity: str | None,
+ ) -> list[str]:
+ """Find SQL-bearing chart fields without building a chart or its
prequeries."""
+ columns_by_name = {col.column_name: col for col in self.columns}
+ columns = {name: col.expression for name, col in
columns_by_name.items()}
+ metrics = {metric.metric_name: metric.expression for metric in
self.metrics}
+ selected_columns = query_obj.get("columns") or []
+ selected_metrics = query_obj.get("metrics") or []
+ adhoc_columns_by_label = {
+ col.get("label"): col
+ for col in selected_columns
+ if utils.is_adhoc_column(col)
+ }
+ metric_labels = {utils.get_metric_name(metric) for metric in
selected_metrics}
+ expressions: list[str] = []
+
+ def add(value: Any, metric: bool = False) -> None:
+ """Resolve stored references in the namespace used by the query
sink."""
+ if isinstance(value, str):
+ references = metrics if metric else columns
+ expressions.append(references.get(value) or value)
+ elif isinstance(value, dict):
+ if sql := value.get("sqlExpression"):
+ sql = template_processor.process_template(sql)
+ # Adhoc columns can name a saved calculated column after
+ # rendering. SQL metrics use their SQL literally instead.
+ expressions.append(
+ sql if metric else columns.get(sql.strip()) or sql
+ )
+ add(value.get("column"))
+ add(value.get("column_name"))
+
+ def add_orderby(value: Any) -> None:
+ """ORDER BY prefers selected labels before stored metric/column
names."""
+ if isinstance(value, str):
+ if value in metric_labels:
+ # Selected metrics have already been collected.
+ return
+ if value in adhoc_columns_by_label:
+ add(adhoc_columns_by_label[value])
+ return
+ add(
+ value,
+ metric=utils.is_adhoc_metric(value)
+ or (isinstance(value, str) and value in metrics),
+ )
+
+ def add_filter(value: Any) -> None:
+ """Filters accept verbose column names before stored metric
names."""
+ if not isinstance(value, str):
+ add(value)
+ elif value == utils.DTTM_ALIAS and query_obj.get("is_timeseries",
True):
+ add(granularity)
+ elif (col := self._get_filter_column(value, columns_by_name)) is
not None:
+ expressions.append(col.expression or value)
+ elif value in metrics:
+ add(value, metric=True)
+ elif value in adhoc_columns_by_label:
+ add(adhoc_columns_by_label[value])
+
+ for value in query_obj.get("groupby") or selected_columns:
+ add(value)
+ for metric_value in selected_metrics:
+ add(metric_value, metric=True)
+ for orderby_value, _ascending in query_obj.get("orderby") or []:
+ add_orderby(orderby_value)
+ for flt in query_obj.get("filter") or []:
+ add_filter(flt.get("col"))
+ series_limit = query_obj.get("series_limit") or (
+ query_obj.get("timeseries_limit")
+ if query_obj.get("is_timeseries", True)
+ else 0
+ )
+ add(
+ (
+ query_obj.get("series_limit_metric")
+ or query_obj.get("timeseries_limit_metric")
+ )
+ if series_limit
+ else None,
+ metric=True,
+ )
+ add(granularity)
+ if (
+ granularity
+ and self.always_filter_main_dttm
+ and self.main_dttm_col in self.dttm_cols
+ and self.main_dttm_col != granularity
+ ):
+ add(self.main_dttm_col)
+ extras = query_obj.get("extras") or {}
+ expressions.extend(
+ extras[key] for key in ("where", "having") if extras.get(key)
+ )
+ if (
+ query_obj.get("apply_fetch_values_predicate")
+ and self.fetch_values_predicate
+ ):
+ expressions.append(self.fetch_values_predicate)
+ return expressions
+
+ def _get_query_rls_cache_keys(
+ self,
+ query_obj: QueryObjectDict,
+ default_schema: str | None = None,
+ resolve_catalog: Callable[[], str | None] | None = None,
+ ) -> list[str]:
+ """Collect expression-only RLS rules without building or executing SQL.
+
+ Only subquery expressions need table discovery. Reuse the
+ datasource SQL's resolved schema, or resolve it once for physical
datasets.
+ Do not exclude the dataset's own id: an independent expression
subquery is
+ not constrained by the outer query's RLS.
+ """
+ from superset.utils.rls import collect_rls_predicates_for_sql
+
+ if not query_obj or not is_feature_enabled("ALLOW_ADHOC_SUBQUERY"):
+ return []
+ resolve_catalog = resolve_catalog or
cache(self.database.get_default_catalog)
+ extras = query_obj.get("extras") or {}
+ granularity = query_obj.get("granularity")
+ if granularity is not None and granularity not in self.dttm_cols:
+ granularity = self.main_dttm_col
+ template_kwargs = {
+ key: query_obj.get(key)
+ for key in (
+ "columns",
+ "groupby",
+ "metrics",
+ "row_limit",
+ "row_offset",
+ "filter",
+ )
+ }
+ template_kwargs.update(
+ time_column=granularity,
+ time_grain=extras.get("time_grain_sqla"),
+ table_columns=[col.column_name for col in self.columns],
+ )
+ template_kwargs.update(self.template_params_dict)
+ template_processor = self.get_template_processor(**template_kwargs)
+ predicates: set[str] = set()
+ for expression in self._get_query_sql_expressions(
+ query_obj, template_processor, granularity
+ ):
+ expression = template_processor.process_template(expression)
Review Comment:
Adhoc SQL gets rendered twice on this path: `add()` already calls
`process_template` on `sqlExpression` (line 4070), and this loop renders the
result again, whereas execution renders an adhoc metric once. Two consequences:
a valid metric like `MAX('{% raw %}{{ x }}{% endraw %}')` (or a template param
that expands to `{{ 1 / 0 }}`) runs fine but raises here, so the chart request
fails before the cache lookup. Worse, a second pass can drop SQL that execution
keeps, e.g. `MAX(0 /*{% raw %}{% if false %}{% endraw %}*/ + (SELECT
MAX(secret) FROM protected) /*{% raw %}{% endif %}{% endraw %}*/)` executes the
protected subquery with RLS but collects as `MAX(0 /**/)`, so its predicate
never reaches the cache key and users with different RLS rules can share a
cached result. Should the adhoc expression be rendered only once, in `add()` or
here, matching what execution does?
--
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]
