airbyte-cdk 6.12.4__tar.gz → 6.12.4.dev0__tar.gz

{airbyte_cdk-6.12.4 → airbyte_cdk-6.12.4.dev0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: airbyte-cdk
-Version: 6.12.4
+Version: 6.12.4.dev0
 Summary: A framework for writing Airbyte Connectors.
 Home-page: https://airbyte.com
 License: MIT
@@ -57,7 +57,7 @@ Requires-Dist: python-calamine (==0.2.3) ; extra == "file-based"
 Requires-Dist: python-dateutil
 Requires-Dist: python-snappy (==0.7.3) ; extra == "file-based"
 Requires-Dist: python-ulid (>=3.0.0,<4.0.0)
-Requires-Dist: pytz (==2024.2)
+Requires-Dist: pytz (==2024.1)
 Requires-Dist: rapidfuzz (>=3.10.1,<4.0.0)
 Requires-Dist: requests
 Requires-Dist: requests_cache

{airbyte_cdk-6.12.4 → airbyte_cdk-6.12.4.dev0}/airbyte_cdk/sources/declarative/auth/oauth.py

@@ -43,11 +43,11 @@ class DeclarativeOauth2Authenticator(AbstractOauth2Authenticator, DeclarativeAut
         message_repository (MessageRepository): the message repository used to emit logs on HTTP requests
     """
 
+    token_refresh_endpoint: Union[InterpolatedString, str]
     client_id: Union[InterpolatedString, str]
     client_secret: Union[InterpolatedString, str]
     config: Mapping[str, Any]
     parameters: InitVar[Mapping[str, Any]]
-    token_refresh_endpoint: Optional[Union[InterpolatedString, str]] = None
     refresh_token: Optional[Union[InterpolatedString, str]] = None
     scopes: Optional[List[str]] = None
     token_expiry_date: Optional[Union[InterpolatedString, str]] = None
@@ -55,7 +55,6 @@ class DeclarativeOauth2Authenticator(AbstractOauth2Authenticator, DeclarativeAut
     token_expiry_date_format: Optional[str] = None
     token_expiry_is_time_of_expiration: bool = False
     access_token_name: Union[InterpolatedString, str] = "access_token"
-    access_token_value: Optional[Union[InterpolatedString, str]] = None
     expires_in_name: Union[InterpolatedString, str] = "expires_in"
     refresh_request_body: Optional[Mapping[str, Any]] = None
     grant_type: Union[InterpolatedString, str] = "refresh_token"
@@ -63,12 +62,9 @@ class DeclarativeOauth2Authenticator(AbstractOauth2Authenticator, DeclarativeAut
 
     def __post_init__(self, parameters: Mapping[str, Any]) -> None:
         super().__init__()
-        if self.token_refresh_endpoint is not None:
-            self._token_refresh_endpoint: Optional[InterpolatedString] = InterpolatedString.create(
-                self.token_refresh_endpoint, parameters=parameters
-            )
-        else:
-            self._token_refresh_endpoint = None
+        self._token_refresh_endpoint = InterpolatedString.create(
+            self.token_refresh_endpoint, parameters=parameters
+        )
         self._client_id = InterpolatedString.create(self.client_id, parameters=parameters)
         self._client_secret = InterpolatedString.create(self.client_secret, parameters=parameters)
         if self.refresh_token is not None:
@@ -96,31 +92,20 @@ class DeclarativeOauth2Authenticator(AbstractOauth2Authenticator, DeclarativeAut
             if self.token_expiry_date
             else pendulum.now().subtract(days=1)  # type: ignore # substract does not have type hints
         )
-        if self.access_token_value is not None:
-            self._access_token_value = InterpolatedString.create(
-                self.access_token_value, parameters=parameters
-            ).eval(self.config)
-        else:
-            self._access_token_value = None
-
-        self._access_token: Optional[str] = (
-            self._access_token_value if self.access_token_value else None
-        )
+        self._access_token: Optional[str] = None  # access_token is initialized by a setter
 
         if self.get_grant_type() == "refresh_token" and self._refresh_token is None:
             raise ValueError(
                 "OAuthAuthenticator needs a refresh_token parameter if grant_type is set to `refresh_token`"
             )
 
-    def get_token_refresh_endpoint(self) -> Optional[str]:
-        if self._token_refresh_endpoint is not None:
-            refresh_token_endpoint: str = self._token_refresh_endpoint.eval(self.config)
-            if not refresh_token_endpoint:
-                raise ValueError(
-                    "OAuthAuthenticator was unable to evaluate token_refresh_endpoint parameter"
-                )
-            return refresh_token_endpoint
-        return None
+    def get_token_refresh_endpoint(self) -> str:
+        refresh_token: str = self._token_refresh_endpoint.eval(self.config)
+        if not refresh_token:
+            raise ValueError(
+                "OAuthAuthenticator was unable to evaluate token_refresh_endpoint parameter"
+            )
+        return refresh_token
 
     def get_client_id(self) -> str:
         client_id: str = self._client_id.eval(self.config)

{airbyte_cdk-6.12.4 → airbyte_cdk-6.12.4.dev0}/airbyte_cdk/sources/declarative/concurrent_declarative_source.py

@@ -20,6 +20,9 @@ from airbyte_cdk.sources.declarative.extractors.record_filter import (
     ClientSideIncrementalRecordFilterDecorator,
 )
 from airbyte_cdk.sources.declarative.incremental.datetime_based_cursor import DatetimeBasedCursor
+from airbyte_cdk.sources.declarative.incremental.per_partition_with_global import (
+    PerPartitionWithGlobalCursor,
+)
 from airbyte_cdk.sources.declarative.interpolation import InterpolatedString
 from airbyte_cdk.sources.declarative.manifest_declarative_source import ManifestDeclarativeSource
 from airbyte_cdk.sources.declarative.models.declarative_component_schema import (
@@ -300,6 +303,60 @@ class ConcurrentDeclarativeSource(ManifestDeclarativeSource, Generic[TState]):
                             cursor=final_state_cursor,
                         )
                     )
+                elif (
+                    incremental_sync_component_definition
+                    and incremental_sync_component_definition.get("type", "")
+                    == DatetimeBasedCursorModel.__name__
+                    and self._stream_supports_concurrent_partition_processing(
+                        declarative_stream=declarative_stream
+                    )
+                    and hasattr(declarative_stream.retriever, "stream_slicer")
+                    and isinstance(
+                        declarative_stream.retriever.stream_slicer, PerPartitionWithGlobalCursor
+                    )
+                ):
+                    stream_state = state_manager.get_stream_state(
+                        stream_name=declarative_stream.name, namespace=declarative_stream.namespace
+                    )
+                    partition_router = declarative_stream.retriever.stream_slicer._partition_router
+
+                    cursor = self._constructor.create_concurrent_cursor_from_perpartition_cursor(
+                        state_manager=state_manager,
+                        model_type=DatetimeBasedCursorModel,
+                        component_definition=incremental_sync_component_definition,
+                        stream_name=declarative_stream.name,
+                        stream_namespace=declarative_stream.namespace,
+                        config=config or {},
+                        stream_state=stream_state,
+                        partition_router=partition_router,
+                    )
+
+                    partition_generator = StreamSlicerPartitionGenerator(
+                        DeclarativePartitionFactory(
+                            declarative_stream.name,
+                            declarative_stream.get_json_schema(),
+                            self._retriever_factory(
+                                name_to_stream_mapping[declarative_stream.name],
+                                config,
+                                stream_state,
+                            ),
+                            self.message_repository,
+                        ),
+                        cursor,
+                    )
+
+                    concurrent_streams.append(
+                        DefaultStream(
+                            partition_generator=partition_generator,
+                            name=declarative_stream.name,
+                            json_schema=declarative_stream.get_json_schema(),
+                            availability_strategy=AlwaysAvailableAvailabilityStrategy(),
+                            primary_key=get_primary_key_from_stream(declarative_stream.primary_key),
+                            cursor_field=cursor.cursor_field.cursor_field_key,
+                            logger=self.logger,
+                            cursor=cursor,
+                        )
+                    )
                 else:
                     synchronous_streams.append(declarative_stream)
             else:

{airbyte_cdk-6.12.4 → airbyte_cdk-6.12.4.dev0}/airbyte_cdk/sources/declarative/declarative_component_schema.yaml

@@ -1021,6 +1021,7 @@ definitions:
       - type
       - client_id
       - client_secret
+      - token_refresh_endpoint
     properties:
       type:
         type: string
@@ -1059,12 +1060,6 @@ definitions:
         default: "access_token"
         examples:
           - access_token
-      access_token_value:
-        title: Access Token Value
-        description: The value of the access_token to bypass the token refreshing using `refresh_token`.
-        type: string
-        examples:
-          - secret_access_token_value
       expires_in_name:
         title: Token Expiry Property Name
         description: The name of the property which contains the expiry date in the response from the token refresh endpoint.
@@ -1240,7 +1235,6 @@ definitions:
             - "$ref": "#/definitions/RemoveFields"
             - "$ref": "#/definitions/KeysToLower"
             - "$ref": "#/definitions/KeysToSnakeCase"
-            - "$ref": "#/definitions/FlattenFields"
       state_migrations:
         title: State Migrations
         description: Array of state migrations to be applied on the input state
@@ -1773,18 +1767,6 @@ definitions:
           - "$ref": "#/definitions/AsyncRetriever"
           - "$ref": "#/definitions/CustomRetriever"
           - "$ref": "#/definitions/SimpleRetriever"
-      schema_transformations:
-        title: Schema Transformations
-        description: A list of transformations to be applied to the schema.
-        type: array
-        items:
-          anyOf:
-            - "$ref": "#/definitions/AddFields"
-            - "$ref": "#/definitions/CustomTransformation"
-            - "$ref": "#/definitions/RemoveFields"
-            - "$ref": "#/definitions/KeysToLower"
-            - "$ref": "#/definitions/KeysToSnakeCase"
-            - "$ref": "#/definitions/FlattenFields"
       schema_type_identifier:
         "$ref": "#/definitions/SchemaTypeIdentifier"
       $parameters:
@@ -3054,7 +3036,6 @@ definitions:
         interpolation_context:
           - config
           - components_values
-          - stream_slice
           - stream_template_config
         examples:
           - ["data"]
@@ -3071,13 +3052,10 @@ definitions:
           - config
           - stream_template_config
           - components_values
-          - stream_slice
         examples:
           - "{{ components_values['updates'] }}"
           - "{{ components_values['MetaData']['LastUpdatedTime'] }}"
           - "{{ config['segment_id'] }}"
-          - "{{ stream_slice['parent_id'] }}"
-          - "{{ stream_slice['extra_fields']['name'] }}"
       value_type:
         title: Value Type
         description: The expected data type of the value. If omitted, the type will be inferred from the value provided.

{airbyte_cdk-6.12.4 → airbyte_cdk-6.12.4.dev0}/airbyte_cdk/sources/declarative/extractors/record_filter.py

@@ -59,13 +59,11 @@ class ClientSideIncrementalRecordFilterDecorator(RecordFilter):
 
     def __init__(
         self,
-        date_time_based_cursor: DatetimeBasedCursor,
-        substream_cursor: Optional[Union[PerPartitionWithGlobalCursor, GlobalSubstreamCursor]],
+        cursor: Union[DatetimeBasedCursor, PerPartitionWithGlobalCursor, GlobalSubstreamCursor],
         **kwargs: Any,
     ):
         super().__init__(**kwargs)
-        self._date_time_based_cursor = date_time_based_cursor
-        self._substream_cursor = substream_cursor
+        self._cursor = cursor
 
     def filter_records(
         self,
@@ -77,7 +75,7 @@ class ClientSideIncrementalRecordFilterDecorator(RecordFilter):
         records = (
             record
             for record in records
-            if (self._substream_cursor or self._date_time_based_cursor).should_be_synced(
+            if self._cursor.should_be_synced(
                 # Record is created on the fly to align with cursors interface; stream name is ignored as we don't need it here
                 # Record stream name is empty cause it is not used durig the filtering
                 Record(data=record, associated_slice=stream_slice, stream_name="")

{airbyte_cdk-6.12.4 → airbyte_cdk-6.12.4.dev0}/airbyte_cdk/sources/declarative/incremental/__init__.py

@@ -2,6 +2,7 @@
 # Copyright (c) 2022 Airbyte, Inc., all rights reserved.
 #
 
+from airbyte_cdk.sources.declarative.incremental.concurrent_partition_cursor import ConcurrentCursorFactory, ConcurrentPerPartitionCursor
 from airbyte_cdk.sources.declarative.incremental.datetime_based_cursor import DatetimeBasedCursor
 from airbyte_cdk.sources.declarative.incremental.declarative_cursor import DeclarativeCursor
 from airbyte_cdk.sources.declarative.incremental.global_substream_cursor import GlobalSubstreamCursor
@@ -14,6 +15,8 @@ from airbyte_cdk.sources.declarative.incremental.resumable_full_refresh_cursor i
 
 __all__ = [
     "CursorFactory",
+    "ConcurrentCursorFactory"
+    "ConcurrentPerPartitionCursor",
     "DatetimeBasedCursor",
     "DeclarativeCursor",
     "GlobalSubstreamCursor",

airbyte_cdk-6.12.4.dev0/airbyte_cdk/sources/declarative/incremental/concurrent_partition_cursor.py

@@ -0,0 +1,344 @@
+import copy
+import logging
+
+#
+# Copyright (c) 2023 Airbyte, Inc., all rights reserved.
+#
+import threading
+from collections import OrderedDict
+from copy import deepcopy
+from typing import Any, Callable, Iterable, Mapping, MutableMapping, Optional
+
+from airbyte_cdk.sources.connector_state_manager import ConnectorStateManager
+from airbyte_cdk.sources.declarative.incremental.declarative_cursor import DeclarativeCursor
+from airbyte_cdk.sources.declarative.incremental.global_substream_cursor import (
+    Timer,
+    iterate_with_last_flag_and_state,
+)
+from airbyte_cdk.sources.declarative.partition_routers.partition_router import PartitionRouter
+from airbyte_cdk.sources.message import MessageRepository
+from airbyte_cdk.sources.streams.checkpoint.per_partition_key_serializer import (
+    PerPartitionKeySerializer,
+)
+from airbyte_cdk.sources.streams.concurrent.cursor import Cursor, CursorField
+from airbyte_cdk.sources.streams.concurrent.partitions.partition import Partition
+from airbyte_cdk.sources.types import Record, StreamSlice, StreamState
+
+logger = logging.getLogger("airbyte")
+
+
+class ConcurrentCursorFactory:
+    def __init__(self, create_function: Callable[..., Cursor]):
+        self._create_function = create_function
+
+    def create(self, stream_state: Mapping[str, Any]) -> Cursor:
+        return self._create_function(stream_state=stream_state)
+
+
+class ConcurrentPerPartitionCursor(Cursor):
+    """
+    Manages state per partition when a stream has many partitions, to prevent data loss or duplication.
+
+    **Partition Limitation and Limit Reached Logic**
+
+    - **DEFAULT_MAX_PARTITIONS_NUMBER**: The maximum number of partitions to keep in memory (default is 10,000).
+    - **_cursor_per_partition**: An ordered dictionary that stores cursors for each partition.
+    - **_over_limit**: A counter that increments each time an oldest partition is removed when the limit is exceeded.
+
+    The class ensures that the number of partitions tracked does not exceed the `DEFAULT_MAX_PARTITIONS_NUMBER` to prevent excessive memory usage.
+
+    - When the number of partitions exceeds the limit, the oldest partitions are removed from `_cursor_per_partition`, and `_over_limit` is incremented accordingly.
+    - The `limit_reached` method returns `True` when `_over_limit` exceeds `DEFAULT_MAX_PARTITIONS_NUMBER`, indicating that the global cursor should be used instead of per-partition cursors.
+
+    This approach avoids unnecessary switching to a global cursor due to temporary spikes in partition counts, ensuring that switching is only done when a sustained high number of partitions is observed.
+    """
+
+    DEFAULT_MAX_PARTITIONS_NUMBER = 10000
+    _NO_STATE: Mapping[str, Any] = {}
+    _NO_CURSOR_STATE: Mapping[str, Any] = {}
+    _KEY = 0
+    _VALUE = 1
+
+    def __init__(
+        self,
+        cursor_factory: ConcurrentCursorFactory,
+        partition_router: PartitionRouter,
+        stream_name: str,
+        stream_namespace: Optional[str],
+        stream_state: Any,
+        message_repository: MessageRepository,
+        connector_state_manager: ConnectorStateManager,
+        cursor_field: CursorField,
+    ) -> None:
+        self._global_cursor: Mapping[str, Any] = {}
+        self._stream_name = stream_name
+        self._stream_namespace = stream_namespace
+        self._message_repository = message_repository
+        self._connector_state_manager = connector_state_manager
+        self._cursor_field = cursor_field
+
+        self._cursor_factory = cursor_factory
+        self._partition_router = partition_router
+
+        # The dict is ordered to ensure that once the maximum number of partitions is reached,
+        # the oldest partitions can be efficiently removed, maintaining the most recent partitions.
+        self._cursor_per_partition: OrderedDict[str, Cursor] = OrderedDict()
+        self._state = {"states": []}
+        self._semaphore_per_partition = OrderedDict()
+        self._finished_partitions = set()
+        self._lock = threading.Lock()
+        self._timer = Timer()
+        self._new_global_cursor = None
+        self._lookback_window = 0
+        self._parent_state = None
+        self._over_limit = 0
+        self._partition_serializer = PerPartitionKeySerializer()
+
+        self._set_initial_state(stream_state)
+
+    @property
+    def cursor_field(self) -> CursorField:
+        return self._cursor_field
+
+    @property
+    def state(self) -> MutableMapping[str, Any]:
+        states = []
+        for partition_tuple, cursor in self._cursor_per_partition.items():
+            cursor_state = cursor._connector_state_converter.convert_to_state_message(
+                self.cursor_field, cursor.state
+            )
+            if cursor_state:
+                states.append(
+                    {
+                        "partition": self._to_dict(partition_tuple),
+                        "cursor": copy.deepcopy(cursor_state),
+                    }
+                )
+        state: dict[str, Any] = {"states": states}
+
+        if self._global_cursor:
+            state["state"] = self._global_cursor
+        if self._lookback_window is not None:
+            state["lookback_window"] = self._lookback_window
+        if self._parent_state is not None:
+            state["parent_state"] = self._parent_state
+        return state
+
+    def close_partition(self, partition: Partition) -> None:
+        print(f"Closing partition {self._to_partition_key(partition._stream_slice.partition)}")
+        self._cursor_per_partition[
+            self._to_partition_key(partition._stream_slice.partition)
+        ].close_partition(partition=partition)
+        with self._lock:
+            self._semaphore_per_partition[
+                self._to_partition_key(partition._stream_slice.partition)
+            ].acquire()
+            cursor = self._cursor_per_partition[
+                self._to_partition_key(partition._stream_slice.partition)
+            ]
+            cursor_state = cursor._connector_state_converter.convert_to_state_message(
+                cursor._cursor_field, cursor.state
+            )
+            print(f"State {cursor_state} {cursor.state}")
+            if (
+                self._to_partition_key(partition._stream_slice.partition)
+                in self._finished_partitions
+                and self._semaphore_per_partition[
+                    self._to_partition_key(partition._stream_slice.partition)
+                ]._value
+                == 0
+            ):
+                if (
+                    self._new_global_cursor is None
+                    or self._new_global_cursor[self.cursor_field.cursor_field_key]
+                    < cursor_state[self.cursor_field.cursor_field_key]
+                ):
+                    self._new_global_cursor = copy.deepcopy(cursor_state)
+
+    def ensure_at_least_one_state_emitted(self) -> None:
+        """
+        The platform expect to have at least one state message on successful syncs. Hence, whatever happens, we expect this method to be
+        called.
+        """
+        if not any(
+            semaphore_item[1]._value for semaphore_item in self._semaphore_per_partition.items()
+        ):
+            self._global_cursor = self._new_global_cursor
+            self._lookback_window = self._timer.finish()
+            self._parent_state = self._partition_router.get_stream_state()
+        self._emit_state_message()
+
+    def _emit_state_message(self) -> None:
+        self._connector_state_manager.update_state_for_stream(
+            self._stream_name,
+            self._stream_namespace,
+            self.state,
+        )
+        state_message = self._connector_state_manager.create_state_message(
+            self._stream_name, self._stream_namespace
+        )
+        self._message_repository.emit_message(state_message)
+
+    def stream_slices(self) -> Iterable[StreamSlice]:
+        slices = self._partition_router.stream_slices()
+        self._timer.start()
+        for partition in slices:
+            yield from self.generate_slices_from_partition(partition)
+
+    def generate_slices_from_partition(self, partition: StreamSlice) -> Iterable[StreamSlice]:
+        # Ensure the maximum number of partitions is not exceeded
+        self._ensure_partition_limit()
+
+        cursor = self._cursor_per_partition.get(self._to_partition_key(partition.partition))
+        if not cursor:
+            partition_state = (
+                self._global_cursor
+                if self._global_cursor
+                else self._NO_CURSOR_STATE
+            )
+            cursor = self._create_cursor(partition_state)
+            self._cursor_per_partition[self._to_partition_key(partition.partition)] = cursor
+            self._semaphore_per_partition[self._to_partition_key(partition.partition)] = (
+                threading.Semaphore(0)
+            )
+
+        for cursor_slice, is_last_slice, _ in iterate_with_last_flag_and_state(
+            cursor.stream_slices(),
+            lambda: None,
+        ):
+            self._semaphore_per_partition[self._to_partition_key(partition.partition)].release()
+            if is_last_slice:
+                self._finished_partitions.add(self._to_partition_key(partition.partition))
+            yield StreamSlice(
+                partition=partition, cursor_slice=cursor_slice, extra_fields=partition.extra_fields
+            )
+
+    def _ensure_partition_limit(self) -> None:
+        """
+        Ensure the maximum number of partitions is not exceeded. If so, the oldest added partition will be dropped.
+        """
+        while len(self._cursor_per_partition) > self.DEFAULT_MAX_PARTITIONS_NUMBER - 1:
+            self._over_limit += 1
+            oldest_partition = self._cursor_per_partition.popitem(last=False)[
+                0
+            ]  # Remove the oldest partition
+            logger.warning(
+                f"The maximum number of partitions has been reached. Dropping the oldest partition: {oldest_partition}. Over limit: {self._over_limit}."
+            )
+
+    def limit_reached(self) -> bool:
+        return self._over_limit > self.DEFAULT_MAX_PARTITIONS_NUMBER
+
+    def _set_initial_state(self, stream_state: StreamState) -> None:
+        """
+        Set the initial state for the cursors.
+
+        This method initializes the state for each partition cursor using the provided stream state.
+        If a partition state is provided in the stream state, it will update the corresponding partition cursor with this state.
+
+        Additionally, it sets the parent state for partition routers that are based on parent streams. If a partition router
+        does not have parent streams, this step will be skipped due to the default PartitionRouter implementation.
+
+        Args:
+            stream_state (StreamState): The state of the streams to be set. The format of the stream state should be:
+                {
+                    "states": [
+                        {
+                            "partition": {
+                                "partition_key": "value"
+                            },
+                            "cursor": {
+                                "last_updated": "2023-05-27T00:00:00Z"
+                            }
+                        }
+                    ],
+                    "parent_state": {
+                        "parent_stream_name": {
+                            "last_updated": "2023-05-27T00:00:00Z"
+                        }
+                    }
+                }
+        """
+        if not stream_state:
+            return
+
+        if "states" not in stream_state:
+            # We assume that `stream_state` is in a global format that can be applied to all partitions.
+            # Example: {"global_state_format_key": "global_state_format_value"}
+            self._global_cursor = deepcopy(stream_state)
+            self._new_global_cursor = deepcopy(stream_state)
+
+        else:
+            for state in stream_state["states"]:
+                self._cursor_per_partition[self._to_partition_key(state["partition"])] = (
+                    self._create_cursor(state["cursor"])
+                )
+                self._semaphore_per_partition[self._to_partition_key(state["partition"])] = (
+                    threading.Semaphore(0)
+                )
+
+            # set default state for missing partitions if it is per partition with fallback to global
+            if "state" in stream_state:
+                self._global_cursor = deepcopy(stream_state["state"])
+                self._new_global_cursor = deepcopy(stream_state["state"])
+
+        # Set parent state for partition routers based on parent streams
+        self._partition_router.set_initial_state(stream_state)
+
+    def observe(self, record: Record) -> None:
+        print("Observing record in concirrent perpartition ", self._to_partition_key(record.associated_slice.partition), record, self._cursor_per_partition[
+            self._to_partition_key(record.associated_slice.partition)
+        ].state)
+        self._cursor_per_partition[
+            self._to_partition_key(record.associated_slice.partition)
+        ].observe(record)
+
+    def _to_partition_key(self, partition: Mapping[str, Any]) -> str:
+        return self._partition_serializer.to_partition_key(partition)
+
+    def _to_dict(self, partition_key: str) -> Mapping[str, Any]:
+        return self._partition_serializer.to_partition(partition_key)
+
+    def _create_cursor(self, cursor_state: Any) -> Cursor:
+        cursor = self._cursor_factory.create(stream_state=deepcopy(cursor_state))
+        return cursor
+
+    def should_be_synced(self, record: Record) -> bool:
+        return self._get_cursor(record).should_be_synced(record)
+
+    def is_greater_than_or_equal(self, first: Record, second: Record) -> bool:
+        if not first.associated_slice or not second.associated_slice:
+            raise ValueError(
+                f"Both records should have an associated slice but got {first.associated_slice} and {second.associated_slice}"
+            )
+        if first.associated_slice.partition != second.associated_slice.partition:
+            raise ValueError(
+                f"To compare records, partition should be the same but got {first.associated_slice.partition} and {second.associated_slice.partition}"
+            )
+
+        return self._get_cursor(first).is_greater_than_or_equal(
+            self._convert_record_to_cursor_record(first),
+            self._convert_record_to_cursor_record(second),
+        )
+
+    @staticmethod
+    def _convert_record_to_cursor_record(record: Record) -> Record:
+        return Record(
+            record.data,
+            StreamSlice(partition={}, cursor_slice=record.associated_slice.cursor_slice)
+            if record.associated_slice
+            else None,
+        )
+
+    def _get_cursor(self, record: Record) -> Cursor:
+        if not record.associated_slice:
+            raise ValueError(
+                "Invalid state as stream slices that are emitted should refer to an existing cursor"
+            )
+        partition_key = self._to_partition_key(record.associated_slice.partition)
+        if partition_key not in self._cursor_per_partition:
+            raise ValueError(
+                "Invalid state as stream slices that are emitted should refer to an existing cursor"
+            )
+        cursor = self._cursor_per_partition[partition_key]
+        return cursor

{airbyte_cdk-6.12.4 → airbyte_cdk-6.12.4.dev0}/airbyte_cdk/sources/declarative/incremental/per_partition_cursor.py

@@ -303,6 +303,20 @@ class PerPartitionCursor(DeclarativeCursor):
             raise ValueError("A partition needs to be provided in order to get request body json")
 
     def should_be_synced(self, record: Record) -> bool:
+        if (
+            self._to_partition_key(record.associated_slice.partition)
+            not in self._cursor_per_partition
+        ):
+            partition_state = (
+                self._state_to_migrate_from
+                if self._state_to_migrate_from
+                else self._NO_CURSOR_STATE
+            )
+            cursor = self._create_cursor(partition_state)
+
+            self._cursor_per_partition[
+                self._to_partition_key(record.associated_slice.partition)
+            ] = cursor
         return self._get_cursor(record).should_be_synced(
             self._convert_record_to_cursor_record(record)
         )