fix(import-pipeline): dict and bytes mixup (#28525)

Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>

posthog/temporal/data_imports/pipelines/sql_database/__init__.py

@@ -370,7 +370,7 @@ def sql_table(
     metadata: Optional[MetaData] = None,
     incremental: Optional[dlt.sources.incremental[Any]] = None,
     chunk_size: int = DEFAULT_CHUNK_SIZE,
-    backend: TableBackend = "sqlalchemy",
+    backend: TableBackend = "pyarrow",
     detect_precision_hints: Optional[bool] = None,
     reflection_level: Optional[ReflectionLevel] = "full",
     defer_table_reflect: Optional[bool] = None,

posthog/temporal/data_imports/pipelines/sql_database/arrow_helpers.py

@@ -121,6 +121,15 @@ def row_tuples_to_arrow(rows: Sequence[RowAny], columns: TTableSchemaColumns, tz
                 [None if x is not None and math.isnan(x) else x for x in columnar_known_types[field.name]]
             )
 
+        if issubclass(py_type, bytes) or issubclass(py_type, str):
+            # For bytes/str columns, ensure any dict values are serialized to JSON strings
+            # Convert to numpy array after processing
+            processed_values = [
+                None if x is None else json_dumps(x) if isinstance(x, dict | list) else x
+                for x in columnar_known_types[field.name]
+            ]
+            columnar_known_types[field.name] = np.array(processed_values, dtype=object)
+
     # If there are unknown type columns, first create a table to infer their types
     if columnar_unknown_types:
         new_schema_fields = []

posthog/temporal/data_imports/pipelines/sql_database/test/test_arrow_helpers.py

@@ -1,6 +1,6 @@
 import pytest
 import pyarrow as pa
-from posthog.temporal.data_imports.pipelines.sql_database.arrow_helpers import json_dumps
+from posthog.temporal.data_imports.pipelines.sql_database.arrow_helpers import json_dumps, row_tuples_to_arrow
 from dlt.common.json import json
 
 
@@ -20,3 +20,18 @@ def test_handle_large_integers():
     json_str_array = pa.array([None if s is None else json_dumps(s) for s in [{"a": -(2**64)}]])
     loaded = json.loads(json_str_array[0].as_py())
     assert loaded["a"] == float(-(2**64))
+
+
+def test_row_tuples_to_arrow_string_column_with_dict():
+    # Test that row_tuples_to_arrow properly serializes dictionaries in string columns
+    test_dict = {"key": "value"}
+    rows = [("",), (test_dict,)]
+    columns = {"string_col": {"name": "string_col", "data_type": "text", "nullable": True}}
+
+    # This should now succeed and serialize the dictionary to JSON
+    table = row_tuples_to_arrow(rows, columns, "UTC")  # type: ignore
+
+    # Verify the results
+    assert table.column("string_col")[0].as_py() == ""
+    # The dictionary should be serialized to a JSON string
+    assert json.loads(table.column("string_col")[1].as_py()) == test_dict