airbyte-cdk 6.9.1rc2__tar.gz → 6.9.2__tar.gz

{airbyte_cdk-6.9.1rc2 → airbyte_cdk-6.9.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: airbyte-cdk
-Version: 6.9.1rc2
+Version: 6.9.2
 Summary: A framework for writing Airbyte Connectors.
 Home-page: https://airbyte.com
 License: MIT

{airbyte_cdk-6.9.1rc2 → airbyte_cdk-6.9.2}/airbyte_cdk/sources/declarative/concurrent_declarative_source.py

@@ -56,8 +56,9 @@ from airbyte_cdk.sources.types import Config, StreamState
 
 
 class ConcurrentDeclarativeSource(ManifestDeclarativeSource, Generic[TState]):
-    # By default, we defer to a value of 1 which represents running a connector using the Concurrent CDK engine on only one thread.
-    SINGLE_THREADED_CONCURRENCY_LEVEL = 1
+    # By default, we defer to a value of 2. A value lower than than could cause a PartitionEnqueuer to be stuck in a state of deadlock
+    # because it has hit the limit of futures but not partition reader is consuming them.
+    _LOWEST_SAFE_CONCURRENCY_LEVEL = 2
 
     def __init__(
         self,
@@ -107,8 +108,8 @@ class ConcurrentDeclarativeSource(ManifestDeclarativeSource, Generic[TState]):
                 concurrency_level // 2, 1
             )  # Partition_generation iterates using range based on this value. If this is floored to zero we end up in a dead lock during start up
         else:
-            concurrency_level = self.SINGLE_THREADED_CONCURRENCY_LEVEL
-            initial_number_of_partitions_to_generate = self.SINGLE_THREADED_CONCURRENCY_LEVEL
+            concurrency_level = self._LOWEST_SAFE_CONCURRENCY_LEVEL
+            initial_number_of_partitions_to_generate = self._LOWEST_SAFE_CONCURRENCY_LEVEL // 2
 
         self._concurrent_source = ConcurrentSource.create(
             num_workers=concurrency_level,

{airbyte_cdk-6.9.1rc2 → airbyte_cdk-6.9.2}/airbyte_cdk/sources/declarative/datetime/datetime_parser.py

@@ -37,17 +37,17 @@ class DatetimeParser:
             return parsed_datetime.replace(tzinfo=datetime.timezone.utc)
         return parsed_datetime
 
-    def format(self, dt: datetime.datetime, format: str) -> Union[str, int]:
+    def format(self, dt: datetime.datetime, format: str) -> str:
         # strftime("%s") is unreliable because it ignores the time zone information and assumes the time zone of the system it's running on
         # It's safer to use the timestamp() method than the %s directive
         # See https://stackoverflow.com/a/4974930
         if format == "%s":
-            return int(dt.timestamp())
+            return str(int(dt.timestamp()))
         if format == "%s_as_float":
-            return float(dt.timestamp())
+            return str(float(dt.timestamp()))
         if format == "%ms":
             # timstamp() returns a float representing the number of seconds since the unix epoch
-            return int(dt.timestamp() * 1000)
+            return str(int(dt.timestamp() * 1000))
         else:
             return dt.strftime(format)
 

{airbyte_cdk-6.9.1rc2 → airbyte_cdk-6.9.2}/airbyte_cdk/sources/declarative/declarative_component_schema.yaml

@@ -327,7 +327,7 @@ definitions:
         additionalProperties: true
   ConcurrencyLevel:
     title: Concurrency Level
-    description: Defines the amount of parallelization for the streams that are being synced. The factor of parallelization is how many partitions or streams are synced at the same time. For example, with a concurrency_level of 10, ten streams or partitions of data will processed at the same time.
+    description: Defines the amount of parallelization for the streams that are being synced. The factor of parallelization is how many partitions or streams are synced at the same time. For example, with a concurrency_level of 10, ten streams or partitions of data will processed at the same time. Note that a value of 1 could create deadlock if a stream has a very high number of partitions.
     type: object
     required:
       - default_concurrency

{airbyte_cdk-6.9.1rc2 → airbyte_cdk-6.9.2}/airbyte_cdk/sources/declarative/incremental/datetime_based_cursor.py

@@ -248,7 +248,7 @@ class DatetimeBasedCursor(DeclarativeCursor):
         return datetime.datetime.min.replace(tzinfo=datetime.timezone.utc)
 
     def _format_datetime(self, dt: datetime.datetime) -> str:
-        return str(self._parser.format(dt, self.datetime_format))
+        return self._parser.format(dt, self.datetime_format)
 
     def _partition_daterange(
         self,

{airbyte_cdk-6.9.1rc2 → airbyte_cdk-6.9.2}/airbyte_cdk/sources/declarative/interpolation/jinja.py

@@ -4,7 +4,7 @@
 
 import ast
 from functools import cache
-from typing import Any, Mapping, Optional, Tuple, Type
+from typing import Any, Mapping, Optional, Set, Tuple, Type
 
 from jinja2 import meta
 from jinja2.environment import Template
@@ -27,7 +27,35 @@ class StreamPartitionAccessEnvironment(SandboxedEnvironment):
     def is_safe_attribute(self, obj: Any, attr: str, value: Any) -> bool:
         if attr in ["_partition"]:
             return True
-        return super().is_safe_attribute(obj, attr, value)
+        return super().is_safe_attribute(obj, attr, value)  # type: ignore  # for some reason, mypy says 'Returning Any from function declared to return "bool"'
+
+
+# These aliases are used to deprecate existing keywords without breaking all existing connectors.
+_ALIASES = {
+    "stream_interval": "stream_slice",  # Use stream_interval to access incremental_sync values
+    "stream_partition": "stream_slice",  # Use stream_partition to access partition router's values
+}
+
+# These extensions are not installed so they're not currently a problem,
+# but we're still explicitly removing them from the jinja context.
+# At worst, this is documentation that we do NOT want to include these extensions because of the potential security risks
+_RESTRICTED_EXTENSIONS = ["jinja2.ext.loopcontrols"]  # Adds support for break continue in loops
+
+# By default, these Python builtin functions are available in the Jinja context.
+# We explicitly remove them because of the potential security risk.
+# Please add a unit test to test_jinja.py when adding a restriction.
+_RESTRICTED_BUILTIN_FUNCTIONS = [
+    "range"
+]  # The range function can cause very expensive computations
+
+_ENVIRONMENT = StreamPartitionAccessEnvironment()
+_ENVIRONMENT.filters.update(**filters)
+_ENVIRONMENT.globals.update(**macros)
+
+for extension in _RESTRICTED_EXTENSIONS:
+    _ENVIRONMENT.extensions.pop(extension, None)
+for builtin in _RESTRICTED_BUILTIN_FUNCTIONS:
+    _ENVIRONMENT.globals.pop(builtin, None)
 
 
 class JinjaInterpolation(Interpolation):
@@ -48,34 +76,6 @@ class JinjaInterpolation(Interpolation):
     Additional information on jinja templating can be found at https://jinja.palletsprojects.com/en/3.1.x/templates/#
     """
 
-    # These aliases are used to deprecate existing keywords without breaking all existing connectors.
-    ALIASES = {
-        "stream_interval": "stream_slice",  # Use stream_interval to access incremental_sync values
-        "stream_partition": "stream_slice",  # Use stream_partition to access partition router's values
-    }
-
-    # These extensions are not installed so they're not currently a problem,
-    # but we're still explicitely removing them from the jinja context.
-    # At worst, this is documentation that we do NOT want to include these extensions because of the potential security risks
-    RESTRICTED_EXTENSIONS = ["jinja2.ext.loopcontrols"]  # Adds support for break continue in loops
-
-    # By default, these Python builtin functions are available in the Jinja context.
-    # We explicitely remove them because of the potential security risk.
-    # Please add a unit test to test_jinja.py when adding a restriction.
-    RESTRICTED_BUILTIN_FUNCTIONS = [
-        "range"
-    ]  # The range function can cause very expensive computations
-
-    def __init__(self) -> None:
-        self._environment = StreamPartitionAccessEnvironment()
-        self._environment.filters.update(**filters)
-        self._environment.globals.update(**macros)
-
-        for extension in self.RESTRICTED_EXTENSIONS:
-            self._environment.extensions.pop(extension, None)
-        for builtin in self.RESTRICTED_BUILTIN_FUNCTIONS:
-            self._environment.globals.pop(builtin, None)
-
     def eval(
         self,
         input_str: str,
@@ -86,7 +86,7 @@ class JinjaInterpolation(Interpolation):
     ) -> Any:
         context = {"config": config, **additional_parameters}
 
-        for alias, equivalent in self.ALIASES.items():
+        for alias, equivalent in _ALIASES.items():
             if alias in context:
                 # This is unexpected. We could ignore or log a warning, but failing loudly should result in fewer surprises
                 raise ValueError(
@@ -105,6 +105,7 @@ class JinjaInterpolation(Interpolation):
                 raise Exception(f"Expected a string, got {input_str}")
         except UndefinedError:
             pass
+
         # If result is empty or resulted in an undefined error, evaluate and return the default string
         return self._literal_eval(self._eval(default, context), valid_types)
 
@@ -132,16 +133,16 @@ class JinjaInterpolation(Interpolation):
             return s
 
     @cache
-    def _find_undeclared_variables(self, s: Optional[str]) -> set[str]:
+    def _find_undeclared_variables(self, s: Optional[str]) -> Set[str]:
         """
         Find undeclared variables and cache them
         """
-        ast = self._environment.parse(s)  # type: ignore # parse is able to handle None
+        ast = _ENVIRONMENT.parse(s)  # type: ignore # parse is able to handle None
         return meta.find_undeclared_variables(ast)
 
     @cache
-    def _compile(self, s: Optional[str]) -> Template:
+    def _compile(self, s: str) -> Template:
         """
         We must cache the Jinja Template ourselves because we're using `from_string` instead of a template loader
         """
-        return self._environment.from_string(s)  # type: ignore [arg-type]  # Expected `str | Template` but passed `str | None`
+        return _ENVIRONMENT.from_string(s)

{airbyte_cdk-6.9.1rc2 → airbyte_cdk-6.9.2}/airbyte_cdk/sources/streams/concurrent/state_converters/datetime_stream_state_converter.py

@@ -4,7 +4,7 @@
 
 from abc import abstractmethod
 from datetime import datetime, timedelta, timezone
-from typing import Any, Callable, List, MutableMapping, Optional, Tuple, Union
+from typing import Any, Callable, List, MutableMapping, Optional, Tuple
 
 import pendulum
 from pendulum.datetime import DateTime
@@ -202,7 +202,7 @@ class CustomFormatConcurrentStreamStateConverter(IsoMillisConcurrentStreamStateC
         self._input_datetime_formats += [self._datetime_format]
         self._parser = DatetimeParser()
 
-    def output_format(self, timestamp: datetime) -> Union[str, int]:
+    def output_format(self, timestamp: datetime) -> str:
         return self._parser.format(timestamp, self._datetime_format)
 
     def parse_timestamp(self, timestamp: str) -> datetime:

{airbyte_cdk-6.9.1rc2 → airbyte_cdk-6.9.2}/airbyte_cdk/sources/streams/http/http_client.py

@@ -138,12 +138,22 @@ class HttpClient:
             cache_dir = os.getenv(ENV_REQUEST_CACHE_PATH)
             # Use in-memory cache if cache_dir is not set
             # This is a non-obvious interface, but it ensures we don't write sql files when running unit tests
-            if cache_dir:
-                sqlite_path = str(Path(cache_dir) / self.cache_filename)
-            else:
-                sqlite_path = "file::memory:?cache=shared"
+            # Use in-memory cache if cache_dir is not set
+            # This is a non-obvious interface, but it ensures we don't write sql files when running unit tests
+            sqlite_path = (
+                str(Path(cache_dir) / self.cache_filename)
+                if cache_dir
+                else "file::memory:?cache=shared"
+            )
+            # By using `PRAGMA synchronous=OFF` and `PRAGMA journal_mode=WAL`, we reduce the possible occurrences of `database table is locked` errors.
+            # Note that those were blindly added at the same time and one or the other might be sufficient to prevent the issues but we have seen good results with both. Feel free to revisit given more information.
+            # There are strong signals that `fast_save` might create problems but if the sync crashes, we start back from the beginning in terms of sqlite anyway so the impact should be minimal. Signals are:
+            # * https://github.com/requests-cache/requests-cache/commit/7fa89ffda300331c37d8fad7f773348a3b5b0236#diff-f43db4a5edf931647c32dec28ea7557aae4cae8444af4b26c8ecbe88d8c925aaR238
+            # * https://github.com/requests-cache/requests-cache/commit/7fa89ffda300331c37d8fad7f773348a3b5b0236#diff-2e7f95b7d7be270ff1a8118f817ea3e6663cdad273592e536a116c24e6d23c18R164-R168
+            # * `If the application running SQLite crashes, the data will be safe, but the database [might become corrupted](https://www.sqlite.org/howtocorrupt.html#cfgerr) if the operating system crashes or the computer loses power before that data has been written to the disk surface.` in [this description](https://www.sqlite.org/pragma.html#pragma_synchronous).
+            backend = requests_cache.SQLiteCache(sqlite_path, fast_save=True, wal=True)
             return CachedLimiterSession(
-                sqlite_path, backend="sqlite", api_budget=self._api_budget, match_headers=True
+                sqlite_path, backend=backend, api_budget=self._api_budget, match_headers=True
             )
         else:
             return LimiterSession(api_budget=self._api_budget)

{airbyte_cdk-6.9.1rc2 → airbyte_cdk-6.9.2}/pyproject.toml

@@ -23,7 +23,7 @@ classifiers = [
 keywords = ["airbyte", "connector-development-kit", "cdk"]
 
 # Python CDK uses dynamic versioning: https://github.com/mtkennerly/poetry-dynamic-versioning
-version = "6.9.1rc2"  # Version will be calculated dynamically.
+version = "6.9.2"  # Version will be calculated dynamically.
 
 [tool.poetry-dynamic-versioning]
 enable = false
@@ -124,6 +124,9 @@ line-length = 100
 select = ["I"]
 
 [tool.poe.tasks]
+# Installation
+install = { shell = "poetry install --all-extras" }
+
 # Build tasks
 assemble = {cmd = "bin/generate-component-manifest-dagger.sh", help = "Generate component manifest files."}
 build-package = {cmd = "poetry build", help = "Build the python package: source and wheels archives."}