databricks-sql-connector 3.5.0__tar.gz → 3.7.0__tar.gz

{databricks_sql_connector-3.5.0 → databricks_sql_connector-3.7.0}/CHANGELOG.md

@@ -1,5 +1,17 @@
 # Release History
 
+# 3.7.0 (2024-12-23)
+
+- Fix: Incorrect number of rows fetched in inline results when fetching results with FETCH_NEXT orientation (databricks/databricks-sql-python#479 by @jprakash-db)
+- Updated the doc to specify native parameters are not supported in PUT operation (databricks/databricks-sql-python#477 by @jprakash-db)
+- Relax `pyarrow` and `numpy` pin (databricks/databricks-sql-python#452 by @arredond)
+- Feature: Support for async execute has been added (databricks/databricks-sql-python#463 by @jprakash-db)
+- Updated the HTTP retry logic to be similar to the other Databricks drivers (databricks/databricks-sql-python#467 by @jprakash-db)
+
+# 3.6.0 (2024-10-25)
+
+- Support encryption headers in the cloud fetch request (https://github.com/databricks/databricks-sql-python/pull/460 by @jackyhu-db)
+
 # 3.5.0 (2024-10-18)
 
 - Create a non pyarrow flow to handle small results for the column set (databricks/databricks-sql-python#440 by @jprakash-db)

{databricks_sql_connector-3.5.0 → databricks_sql_connector-3.7.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: databricks-sql-connector
-Version: 3.5.0
+Version: 3.7.0
 Summary: Databricks SQL Connector for Python
 License: Apache-2.0
 Author: Databricks
@@ -13,17 +13,16 @@ Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
-Classifier: Programming Language :: Python :: 3.13
 Provides-Extra: alembic
 Provides-Extra: sqlalchemy
 Requires-Dist: alembic (>=1.0.11,<2.0.0) ; extra == "alembic"
 Requires-Dist: lz4 (>=4.0.2,<5.0.0)
-Requires-Dist: numpy (>=1.16.6,<2.0.0) ; python_version >= "3.8" and python_version < "3.11"
-Requires-Dist: numpy (>=1.23.4,<2.0.0) ; python_version >= "3.11"
+Requires-Dist: numpy (>=1.16.6) ; python_version >= "3.8" and python_version < "3.11"
+Requires-Dist: numpy (>=1.23.4) ; python_version >= "3.11"
 Requires-Dist: oauthlib (>=3.1.0,<4.0.0)
 Requires-Dist: openpyxl (>=3.0.10,<4.0.0)
 Requires-Dist: pandas (>=1.2.5,<2.3.0) ; python_version >= "3.8"
-Requires-Dist: pyarrow (>=14.0.1,<17)
+Requires-Dist: pyarrow (>=14.0.1)
 Requires-Dist: requests (>=2.18.1,<3.0.0)
 Requires-Dist: sqlalchemy (>=2.0.21) ; extra == "sqlalchemy" or extra == "alembic"
 Requires-Dist: thrift (>=0.16.0,<0.21.0)

{databricks_sql_connector-3.5.0 → databricks_sql_connector-3.7.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "databricks-sql-connector"
-version = "3.5.0"
+version = "3.7.0"
 description = "Databricks SQL Connector for Python"
 authors = ["Databricks <databricks-sql-connector-maintainers@databricks.com>"]
 license = "Apache-2.0"
@@ -14,14 +14,14 @@ thrift = ">=0.16.0,<0.21.0"
 pandas = [
     { version = ">=1.2.5,<2.3.0", python = ">=3.8" }
 ]
-pyarrow = ">=14.0.1,<17"
+pyarrow = ">=14.0.1"
 
 lz4 = "^4.0.2"
 requests = "^2.18.1"
 oauthlib = "^3.1.0"
 numpy = [
-    { version = "^1.16.6", python = ">=3.8,<3.11" },
-    { version = "^1.23.4", python = ">=3.11" },
+    { version = ">=1.16.6", python = ">=3.8,<3.11" },
+    { version = ">=1.23.4", python = ">=3.11" },
 ]
 sqlalchemy = { version = ">=2.0.21", optional = true }
 openpyxl = "^3.0.10"

{databricks_sql_connector-3.5.0 → databricks_sql_connector-3.7.0}/src/databricks/sql/__init__.py

@@ -68,7 +68,7 @@ DATETIME = DBAPITypeObject("timestamp")
 DATE = DBAPITypeObject("date")
 ROWID = DBAPITypeObject()
 
-__version__ = "3.5.0"
+__version__ = "3.7.0"
 USER_AGENT_NAME = "PyDatabricksSqlConnector"
 
 # These two functions are pyhive legacy

{databricks_sql_connector-3.5.0 → databricks_sql_connector-3.7.0}/src/databricks/sql/auth/retry.py

@@ -1,4 +1,5 @@
 import logging
+import random
 import time
 import typing
 from enum import Enum
@@ -285,25 +286,30 @@ class DatabricksRetryPolicy(Retry):
         """
         retry_after = self.get_retry_after(response)
         if retry_after:
-            backoff = self.get_backoff_time()
-            proposed_wait = max(backoff, retry_after)
-            self.check_proposed_wait(proposed_wait)
-            time.sleep(proposed_wait)
-            return True
+            proposed_wait = retry_after
+        else:
+            proposed_wait = self.get_backoff_time()
 
-        return False
+        proposed_wait = min(proposed_wait, self.delay_max)
+        self.check_proposed_wait(proposed_wait)
+        time.sleep(proposed_wait)
+        return True
 
     def get_backoff_time(self) -> float:
-        """Calls urllib3's built-in get_backoff_time.
+        """
+        This method implements the exponential backoff algorithm to calculate the delay between retries.
 
         Never returns a value larger than self.delay_max
         A MaxRetryDurationError will be raised if the calculated backoff would exceed self.max_attempts_duration
 
-        Note: within urllib3, a backoff is only calculated in cases where a Retry-After header is not present
-            in the previous unsuccessful request and `self.respect_retry_after_header` is True (which is always true)
+        :return:
         """
 
-        proposed_backoff = super().get_backoff_time()
+        current_attempt = self.stop_after_attempts_count - int(self.total or 0)
+        proposed_backoff = (2**current_attempt) * self.delay_min
+        if self.backoff_jitter != 0.0:
+            proposed_backoff += random.random() * self.backoff_jitter
+
         proposed_backoff = min(proposed_backoff, self.delay_max)
         self.check_proposed_wait(proposed_backoff)
 

{databricks_sql_connector-3.5.0 → databricks_sql_connector-3.7.0}/src/databricks/sql/client.py

@@ -1,3 +1,4 @@
+import time
 from typing import Dict, Tuple, List, Optional, Any, Union, Sequence
 
 import pandas
@@ -47,6 +48,7 @@ from databricks.sql.experimental.oauth_persistence import OAuthPersistence
 
 from databricks.sql.thrift_api.TCLIService.ttypes import (
     TSparkParameter,
+    TOperationState,
 )
 
 
@@ -430,6 +432,8 @@ class Cursor:
         self.escaper = ParamEscaper()
         self.lastrowid = None
 
+        self.ASYNC_DEFAULT_POLLING_INTERVAL = 2
+
     # The ideal return type for this method is perhaps Self, but that was not added until 3.11, and we support pre-3.11 pythons, currently.
     def __enter__(self) -> "Cursor":
         return self
@@ -796,6 +800,7 @@ class Cursor:
             cursor=self,
             use_cloud_fetch=self.connection.use_cloud_fetch,
             parameters=prepared_params,
+            async_op=False,
         )
         self.active_result_set = ResultSet(
             self.connection,
@@ -803,6 +808,7 @@ class Cursor:
             self.thrift_backend,
             self.buffer_size_bytes,
             self.arraysize,
+            self.connection.use_cloud_fetch,
         )
 
         if execute_response.is_staging_operation:
@@ -812,6 +818,106 @@ class Cursor:
 
         return self
 
+    def execute_async(
+        self,
+        operation: str,
+        parameters: Optional[TParameterCollection] = None,
+    ) -> "Cursor":
+        """
+
+        Execute a query and do not wait for it to complete and just move ahead
+
+        :param operation:
+        :param parameters:
+        :return:
+        """
+        param_approach = self._determine_parameter_approach(parameters)
+        if param_approach == ParameterApproach.NONE:
+            prepared_params = NO_NATIVE_PARAMS
+            prepared_operation = operation
+
+        elif param_approach == ParameterApproach.INLINE:
+            prepared_operation, prepared_params = self._prepare_inline_parameters(
+                operation, parameters
+            )
+        elif param_approach == ParameterApproach.NATIVE:
+            normalized_parameters = self._normalize_tparametercollection(parameters)
+            param_structure = self._determine_parameter_structure(normalized_parameters)
+            transformed_operation = transform_paramstyle(
+                operation, normalized_parameters, param_structure
+            )
+            prepared_operation, prepared_params = self._prepare_native_parameters(
+                transformed_operation, normalized_parameters, param_structure
+            )
+
+        self._check_not_closed()
+        self._close_and_clear_active_result_set()
+        self.thrift_backend.execute_command(
+            operation=prepared_operation,
+            session_handle=self.connection._session_handle,
+            max_rows=self.arraysize,
+            max_bytes=self.buffer_size_bytes,
+            lz4_compression=self.connection.lz4_compression,
+            cursor=self,
+            use_cloud_fetch=self.connection.use_cloud_fetch,
+            parameters=prepared_params,
+            async_op=True,
+        )
+
+        return self
+
+    def get_query_state(self) -> "TOperationState":
+        """
+        Get the state of the async executing query or basically poll the status of the query
+
+        :return:
+        """
+        self._check_not_closed()
+        return self.thrift_backend.get_query_state(self.active_op_handle)
+
+    def get_async_execution_result(self):
+        """
+
+        Checks for the status of the async executing query and fetches the result if the query is finished
+        Otherwise it will keep polling the status of the query till there is a Not pending state
+        :return:
+        """
+        self._check_not_closed()
+
+        def is_executing(operation_state) -> "bool":
+            return not operation_state or operation_state in [
+                ttypes.TOperationState.RUNNING_STATE,
+                ttypes.TOperationState.PENDING_STATE,
+            ]
+
+        while is_executing(self.get_query_state()):
+            # Poll after some default time
+            time.sleep(self.ASYNC_DEFAULT_POLLING_INTERVAL)
+
+        operation_state = self.get_query_state()
+        if operation_state == ttypes.TOperationState.FINISHED_STATE:
+            execute_response = self.thrift_backend.get_execution_result(
+                self.active_op_handle, self
+            )
+            self.active_result_set = ResultSet(
+                self.connection,
+                execute_response,
+                self.thrift_backend,
+                self.buffer_size_bytes,
+                self.arraysize,
+            )
+
+            if execute_response.is_staging_operation:
+                self._handle_staging_operation(
+                    staging_allowed_local_path=self.thrift_backend.staging_allowed_local_path
+                )
+
+            return self
+        else:
+            raise Error(
+                f"get_execution_result failed with Operation status {operation_state}"
+            )
+
     def executemany(self, operation, seq_of_parameters):
         """
         Execute the operation once for every set of passed in parameters.
@@ -1097,6 +1203,7 @@ class ResultSet:
         thrift_backend: ThriftBackend,
         result_buffer_size_bytes: int = DEFAULT_RESULT_BUFFER_SIZE_BYTES,
         arraysize: int = 10000,
+        use_cloud_fetch: bool = True,
     ):
         """
         A ResultSet manages the results of a single command.
@@ -1118,6 +1225,7 @@ class ResultSet:
         self.description = execute_response.description
         self._arrow_schema_bytes = execute_response.arrow_schema_bytes
         self._next_row_index = 0
+        self._use_cloud_fetch = use_cloud_fetch
 
         if execute_response.arrow_queue:
             # In this case the server has taken the fast path and returned an initial batch of
@@ -1145,6 +1253,7 @@ class ResultSet:
             lz4_compressed=self.lz4_compressed,
             arrow_schema_bytes=self._arrow_schema_bytes,
             description=self.description,
+            use_cloud_fetch=self._use_cloud_fetch,
         )
         self.results = results
         self.has_more_rows = has_more_rows

{databricks_sql_connector-3.5.0 → databricks_sql_connector-3.7.0}/src/databricks/sql/cloudfetch/downloader.py

@@ -100,6 +100,7 @@ class ResultSetDownloadHandler:
                 self.link.fileLink,
                 timeout=self.settings.download_timeout,
                 verify=self._ssl_options.tls_verify,
+                headers=self.link.httpHeaders
                 # TODO: Pass cert from `self._ssl_options`
             )
             response.raise_for_status()

{databricks_sql_connector-3.5.0 → databricks_sql_connector-3.7.0}/src/databricks/sql/thrift_backend.py

@@ -7,6 +7,8 @@ import uuid
 import threading
 from typing import List, Union
 
+from databricks.sql.thrift_api.TCLIService.ttypes import TOperationState
+
 try:
     import pyarrow
 except ImportError:
@@ -64,8 +66,8 @@ DEFAULT_SOCKET_TIMEOUT = float(900)
 # - 900s attempts-duration lines up w ODBC/JDBC drivers (for cluster startup > 10 mins)
 _retry_policy = {  # (type, default, min, max)
     "_retry_delay_min": (float, 1, 0.1, 60),
-    "_retry_delay_max": (float, 60, 5, 3600),
-    "_retry_stop_after_attempts_count": (int, 30, 1, 60),
+    "_retry_delay_max": (float, 30, 5, 3600),
+    "_retry_stop_after_attempts_count": (int, 5, 1, 60),
     "_retry_stop_after_attempts_duration": (float, 900, 1, 86400),
     "_retry_delay_default": (float, 5, 1, 60),
 }
@@ -319,7 +321,7 @@ class ThriftBackend:
 
     # FUTURE: Consider moving to https://github.com/litl/backoff or
     # https://github.com/jd/tenacity for retry logic.
-    def make_request(self, method, request):
+    def make_request(self, method, request, retryable=True):
         """Execute given request, attempting retries when
             1. Receiving HTTP 429/503 from server
             2. OSError is raised during a GetOperationStatus
@@ -458,7 +460,7 @@ class ThriftBackend:
         #       return on success
         #       if available: bounded delay and retry
         #       if not: raise error
-        max_attempts = self._retry_stop_after_attempts_count
+        max_attempts = self._retry_stop_after_attempts_count if retryable else 1
 
         # use index-1 counting for logging/human consistency
         for attempt in range(1, max_attempts + 1):
@@ -769,6 +771,63 @@ class ThriftBackend:
             arrow_schema_bytes=schema_bytes,
         )
 
+    def get_execution_result(self, op_handle, cursor):
+
+        assert op_handle is not None
+
+        req = ttypes.TFetchResultsReq(
+            operationHandle=ttypes.TOperationHandle(
+                op_handle.operationId,
+                op_handle.operationType,
+                False,
+                op_handle.modifiedRowCount,
+            ),
+            maxRows=cursor.arraysize,
+            maxBytes=cursor.buffer_size_bytes,
+            orientation=ttypes.TFetchOrientation.FETCH_NEXT,
+            includeResultSetMetadata=True,
+        )
+
+        resp = self.make_request(self._client.FetchResults, req)
+
+        t_result_set_metadata_resp = resp.resultSetMetadata
+
+        lz4_compressed = t_result_set_metadata_resp.lz4Compressed
+        is_staging_operation = t_result_set_metadata_resp.isStagingOperation
+        has_more_rows = resp.hasMoreRows
+        description = self._hive_schema_to_description(
+            t_result_set_metadata_resp.schema
+        )
+
+        schema_bytes = (
+            t_result_set_metadata_resp.arrowSchema
+            or self._hive_schema_to_arrow_schema(t_result_set_metadata_resp.schema)
+            .serialize()
+            .to_pybytes()
+        )
+
+        queue = ResultSetQueueFactory.build_queue(
+            row_set_type=resp.resultSetMetadata.resultFormat,
+            t_row_set=resp.results,
+            arrow_schema_bytes=schema_bytes,
+            max_download_threads=self.max_download_threads,
+            lz4_compressed=lz4_compressed,
+            description=description,
+            ssl_options=self._ssl_options,
+        )
+
+        return ExecuteResponse(
+            arrow_queue=queue,
+            status=resp.status,
+            has_been_closed_server_side=False,
+            has_more_rows=has_more_rows,
+            lz4_compressed=lz4_compressed,
+            is_staging_operation=is_staging_operation,
+            command_handle=op_handle,
+            description=description,
+            arrow_schema_bytes=schema_bytes,
+        )
+
     def _wait_until_command_done(self, op_handle, initial_operation_status_resp):
         if initial_operation_status_resp:
             self._check_command_not_in_error_or_closed_state(
@@ -787,6 +846,12 @@ class ThriftBackend:
             self._check_command_not_in_error_or_closed_state(op_handle, poll_resp)
         return operation_state
 
+    def get_query_state(self, op_handle) -> "TOperationState":
+        poll_resp = self._poll_for_status(op_handle)
+        operation_state = poll_resp.operationState
+        self._check_command_not_in_error_or_closed_state(op_handle, poll_resp)
+        return operation_state
+
     @staticmethod
     def _check_direct_results_for_error(t_spark_direct_results):
         if t_spark_direct_results:
@@ -817,6 +882,7 @@ class ThriftBackend:
         cursor,
         use_cloud_fetch=True,
         parameters=[],
+        async_op=False,
     ):
         assert session_handle is not None
 
@@ -846,7 +912,11 @@ class ThriftBackend:
             parameters=parameters,
         )
         resp = self.make_request(self._client.ExecuteStatement, req)
-        return self._handle_execute_response(resp, cursor)
+
+        if async_op:
+            self._handle_execute_response_async(resp, cursor)
+        else:
+            return self._handle_execute_response(resp, cursor)
 
     def get_catalogs(self, session_handle, max_rows, max_bytes, cursor):
         assert session_handle is not None
@@ -945,6 +1015,10 @@ class ThriftBackend:
 
         return self._results_message_to_execute_response(resp, final_operation_state)
 
+    def _handle_execute_response_async(self, resp, cursor):
+        cursor.active_op_handle = resp.operationHandle
+        self._check_direct_results_for_error(resp.directResults)
+
     def fetch_results(
         self,
         op_handle,
@@ -954,6 +1028,7 @@ class ThriftBackend:
         lz4_compressed,
         arrow_schema_bytes,
         description,
+        use_cloud_fetch=True,
     ):
         assert op_handle is not None
 
@@ -970,10 +1045,11 @@ class ThriftBackend:
             includeResultSetMetadata=True,
         )
 
-        resp = self.make_request(self._client.FetchResults, req)
+        # Fetch results in Inline mode with FETCH_NEXT orientation are not idempotent and hence not retried
+        resp = self.make_request(self._client.FetchResults, req, use_cloud_fetch)
         if resp.results.startRowOffset > expected_row_start_offset:
-            logger.warning(
-                "Expected results to start from {} but they instead start at {}".format(
+            raise DataError(
+                "fetch_results failed due to inconsistency in the state between the client and the server. Expected results to start from {} but they instead start at {}, some result batches must have been skipped".format(
                     expected_row_start_offset, resp.results.startRowOffset
                 )
             )