revert: "chore: switch data warehouse describe table to cluster" (#42244)

Author: orian

mypy-baseline.txt

@@ -923,6 +923,7 @@ products/data_warehouse/backend/models/modeling.py:0: error: No overload variant
 products/data_warehouse/backend/models/modeling.py:0: error: Value of type "Any | None" is not indexable  [index]
 products/data_warehouse/backend/models/modeling.py:0: error: Value of type "Any | None" is not indexable  [index]
 products/data_warehouse/backend/models/ssh_tunnel.py:0: error: Incompatible types in assignment (expression has type "NoEncryption", variable has type "BestAvailableEncryption")  [assignment]
+products/data_warehouse/backend/models/table.py:0: error: Incompatible return value type (got "dict[str, dict[str, object]]", expected "dict[str, dict[str, str | bool]] | dict[str, str]")  [return-value]
 products/data_warehouse/backend/models/table.py:0: error: Item "None" of "Any | None" has no attribute "get"  [union-attr]
 products/data_warehouse/backend/models/table.py:0: error: Item "None" of "Any | None" has no attribute "keys"  [union-attr]
 products/data_warehouse/backend/models/table.py:0: error: Item "None" of "DataWarehouseCredential | None" has no attribute "access_key"  [union-attr]
@@ -933,6 +934,7 @@ products/data_warehouse/backend/models/table.py:0: error: Item "None" of "DataWa
 products/data_warehouse/backend/models/table.py:0: error: Item "None" of "DataWarehouseCredential | None" has no attribute "access_secret"  [union-attr]
 products/data_warehouse/backend/models/table.py:0: error: Item "None" of "DataWarehouseCredential | None" has no attribute "access_secret"  [union-attr]
 products/data_warehouse/backend/models/table.py:0: error: Item "None" of "DataWarehouseCredential | None" has no attribute "access_secret"  [union-attr]
+products/data_warehouse/backend/models/table.py:0: error: Subclass of "list[tuple[str, ...]]" and "int" cannot exist: have distinct disjoint bases  [unreachable]
 products/data_warehouse/backend/models/table.py:0: error: Value of type "Any | None" is not indexable  [index]
 products/data_warehouse/backend/models/table.py:0: error: Value of type "Any | None" is not indexable  [index]
 products/data_warehouse/backend/test/utils.py:0: error: Dict entry 1 has incompatible type "str": "str | Mapping[str, str | bool]"; expected "str": "str | bool"  [dict-item]

products/data_warehouse/backend/models/table.py

@@ -9,6 +9,7 @@
 from django.db.models import Q
 
 import chdb
+import structlog
 
 from posthog.schema import DatabaseSerializedFieldType, HogQLQueryModifiers
 
@@ -184,30 +185,53 @@ def get_columns(
             context=placeholder_context,
             table_size_mib=self.size_in_s3_mib,
         )
+        logger = structlog.get_logger(__name__)
+        try:
+            # chdb hangs in CI during tests
+            if TEST:
+                raise Exception()
 
-        tag_queries(team_id=self.team.pk, table_id=self.id, warehouse_query=True, name="describe_wh_table")
+            quoted_placeholders = {k: f"'{v}'" for k, v in placeholder_context.values.items()}
+            # chdb doesn't support parameterized queries
+            chdb_query = f"DESCRIBE TABLE (SELECT * FROM {s3_table_func} LIMIT 1)" % quoted_placeholders
 
-        # The cluster is a little broken right now, and so this can intermittently fail.
-        # See https://posthog.slack.com/archives/C076R4753Q8/p1756901693184169 for context
-        attempts = 5
-        result = None
-        for i in range(attempts):
-            try:
-                result = sync_execute(
-                    f"""DESCRIBE TABLE {s3_table_func}""",
-                    args=placeholder_context.values,
-                )
-                break
-            except Exception as err:
-                if i >= attempts - 1:
-                    capture_exception(err)
-                    if safe_expose_ch_error:
-                        self._safe_expose_ch_error(err)
-                    else:
-                        raise
-
-                # Pause execution slightly to not overload clickhouse
-                time.sleep(2**i)
+            # TODO: upgrade chdb once https://github.com/chdb-io/chdb/issues/342 is actually resolved
+            # See https://github.com/chdb-io/chdb/pull/374 for the fix
+            chdb_result = chdb.query(chdb_query, output_format="CSV")
+            reader = csv.reader(StringIO(str(chdb_result)))
+            result = [tuple(row) for row in reader]
+        except Exception as chdb_error:
+            if self._is_suppressed_chdb_error(chdb_error):
+                logger.debug(chdb_error)
+            else:
+                capture_exception(chdb_error)
+
+            tag_queries(team_id=self.team.pk, table_id=self.id, warehouse_query=True)
+
+            # The cluster is a little broken right now, and so this can intermittently fail.
+            # See https://posthog.slack.com/archives/C076R4753Q8/p1756901693184169 for context
+            attempts = 5
+            for i in range(attempts):
+                try:
+                    result = sync_execute(
+                        f"""DESCRIBE TABLE (
+                            SELECT *
+                            FROM {s3_table_func}
+                            LIMIT 1
+                        )""",
+                        args=placeholder_context.values,
+                    )
+                    break
+                except Exception as err:
+                    if i >= attempts - 1:
+                        capture_exception(err)
+                        if safe_expose_ch_error:
+                            self._safe_expose_ch_error(err)
+                        else:
+                            raise
+
+                    # Pause execution slightly to not overload clickhouse
+                    time.sleep(2**i)
 
         if result is None or isinstance(result, int):
             raise Exception("No columns types provided by clickhouse in get_columns")