crawlee 1.0.5b20__py3-none-any.whl → 1.0.5b22__py3-none-any.whl

crawlee/crawlers/_basic/_basic_crawler.py

@@ -56,7 +56,7 @@ from crawlee.errors import (
     SessionError,
     UserDefinedErrorHandlerError,
 )
-from crawlee.events._types import Event, EventCrawlerStatusData, EventPersistStateData
+from crawlee.events._types import Event, EventCrawlerStatusData
 from crawlee.http_clients import ImpitHttpClient
 from crawlee.router import Router
 from crawlee.sessions import SessionPool
@@ -751,9 +751,6 @@ class BasicCrawler(Generic[TCrawlingContext, TStatisticsState]):
 
             await self._autoscaled_pool.run()
 
-            # Emit PERSIST_STATE event when crawler is finishing to allow listeners to persist their state if needed
-            event_manager.emit(event=Event.PERSIST_STATE, event_data=EventPersistStateData(is_migrating=False))
-
     async def add_requests(
         self,
         requests: Sequence[str | Request],

crawlee/events/_event_manager.py

@@ -130,11 +130,13 @@ class EventManager:
         if not self._active:
             raise RuntimeError(f'The {self.__class__.__name__} is not active.')
 
+        # Stop persist state event periodic emission and manually emit last one to ensure latest state is saved.
+        await self._emit_persist_state_event_rec_task.stop()
+        await self._emit_persist_state_event()
         await self.wait_for_all_listeners_to_complete(timeout=self._close_timeout)
         self._event_emitter.remove_all_listeners()
         self._listener_tasks.clear()
         self._listeners_to_wrappers.clear()
-        await self._emit_persist_state_event_rec_task.stop()
         self._active = False
 
     @overload

crawlee/request_loaders/_sitemap_request_loader.py

@@ -9,7 +9,7 @@ from typing import TYPE_CHECKING, Annotated, Any
 from pydantic import BaseModel, ConfigDict, Field
 from typing_extensions import override
 
-from crawlee import Request
+from crawlee import Request, RequestOptions
 from crawlee._utils.docs import docs_group
 from crawlee._utils.globs import Glob
 from crawlee._utils.recoverable_state import RecoverableState
@@ -18,9 +18,10 @@ from crawlee.request_loaders._request_loader import RequestLoader
 
 if TYPE_CHECKING:
     import re
-    from collections.abc import Sequence
+    from collections.abc import Callable, Sequence
     from types import TracebackType
 
+    from crawlee import RequestTransformAction
     from crawlee.http_clients import HttpClient
     from crawlee.proxy_configuration import ProxyInfo
     from crawlee.storage_clients.models import ProcessedRequest
@@ -112,6 +113,7 @@ class SitemapRequestLoader(RequestLoader):
         exclude: list[re.Pattern[Any] | Glob] | None = None,
         max_buffer_size: int = 200,
         persist_state_key: str | None = None,
+        transform_request_function: Callable[[RequestOptions], RequestOptions | RequestTransformAction] | None = None,
     ) -> None:
         """Initialize the sitemap request loader.
 
@@ -125,6 +127,9 @@ class SitemapRequestLoader(RequestLoader):
             persist_state_key: A key for persisting the loader's state in the KeyValueStore.
                 When provided, allows resuming from where it left off after interruption.
                 If None, no state persistence occurs.
+            transform_request_function: An optional function to transform requests
+                generated by the loader. It receives `RequestOptions` with `url` and should return either
+                modified `RequestOptions` or a `RequestTransformAction`.
         """
         self._http_client = http_client
         self._sitemap_urls = sitemap_urls
@@ -132,6 +137,7 @@ class SitemapRequestLoader(RequestLoader):
         self._exclude = exclude
         self._proxy_info = proxy_info
         self._max_buffer_size = max_buffer_size
+        self._transform_request_function = transform_request_function
 
         # Synchronization for queue operations
         self._queue_has_capacity = asyncio.Event()
@@ -313,8 +319,15 @@ class SitemapRequestLoader(RequestLoader):
 
             async with self._queue_lock:
                 url = state.url_queue.popleft()
-
-                request = Request.from_url(url)
+                request_option = RequestOptions(url=url)
+                if self._transform_request_function:
+                    transform_request_option = self._transform_request_function(request_option)
+                    if transform_request_option == 'skip':
+                        state.total_count -= 1
+                        continue
+                    if transform_request_option != 'unchanged':
+                        request_option = transform_request_option
+                request = Request.from_url(**request_option)
                 state.in_progress.add(request.url)
                 if len(state.url_queue) < self._max_buffer_size:
                     self._queue_has_capacity.set()

crawlee/statistics/_models.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import json
+import warnings
 from dataclasses import asdict, dataclass
 from datetime import datetime, timedelta, timezone
 from typing import Annotated, Any
@@ -76,7 +77,6 @@ class StatisticsState(BaseModel):
     crawler_started_at: Annotated[datetime | None, Field(alias='crawlerStartedAt')] = None
     crawler_last_started_at: Annotated[datetime | None, Field(alias='crawlerLastStartTimestamp')] = None
     crawler_finished_at: Annotated[datetime | None, Field(alias='crawlerFinishedAt')] = None
-    crawler_runtime: Annotated[timedelta_ms, Field(alias='crawlerRuntimeMillis')] = timedelta()
     errors: dict[str, Any] = Field(default_factory=dict)
     retry_errors: dict[str, Any] = Field(alias='retryErrors', default_factory=dict)
     requests_with_status_code: dict[str, int] = Field(alias='requestsWithStatusCode', default_factory=dict)
@@ -93,6 +93,37 @@ class StatisticsState(BaseModel):
         ),
     ] = {}
 
+    # Used to track the crawler runtime, that had already been persisted. This is the runtime from previous runs.
+    _runtime_offset: Annotated[timedelta, Field(exclude=True)] = timedelta()
+
+    def model_post_init(self, /, __context: Any) -> None:
+        self._runtime_offset = self.crawler_runtime or self._runtime_offset
+
+    @property
+    def crawler_runtime(self) -> timedelta:
+        if self.crawler_last_started_at:
+            finished_at = self.crawler_finished_at or datetime.now(timezone.utc)
+            return self._runtime_offset + finished_at - self.crawler_last_started_at
+        return self._runtime_offset
+
+    @crawler_runtime.setter
+    def crawler_runtime(self, value: timedelta) -> None:
+        # Setter for backwards compatibility only, the crawler_runtime is now computed_field, and cant be set manually.
+        # To be removed in v2 release https://github.com/apify/crawlee-python/issues/1567
+        warnings.warn(
+            f"Setting 'crawler_runtime' is deprecated and will be removed in a future version."
+            f' Value {value} will not be used.',
+            DeprecationWarning,
+            stacklevel=2,
+        )
+
+    @computed_field(alias='crawlerRuntimeMillis')
+    def crawler_runtime_for_serialization(self) -> timedelta:
+        if self.crawler_last_started_at:
+            finished_at = self.crawler_finished_at or datetime.now(timezone.utc)
+            return self._runtime_offset + finished_at - self.crawler_last_started_at
+        return self._runtime_offset
+
     @computed_field(alias='requestTotalDurationMillis', return_type=timedelta_ms)  # type: ignore[prop-decorator]
     @property
     def request_total_duration(self) -> timedelta:

crawlee/statistics/_statistics.py

@@ -110,9 +110,6 @@ class Statistics(Generic[TStatisticsState]):
         # Flag to indicate the context state.
         self._active = False
 
-        # Pre-existing runtime offset, that can be non-zero when restoring serialized state from KVS.
-        self._runtime_offset = timedelta(seconds=0)
-
     def replace_state_model(self, state_model: type[TNewStatisticsState]) -> Statistics[TNewStatisticsState]:
         """Create near copy of the `Statistics` with replaced `state_model`."""
         new_statistics: Statistics[TNewStatisticsState] = Statistics(
@@ -168,8 +165,8 @@ class Statistics(Generic[TStatisticsState]):
             raise RuntimeError(f'The {self.__class__.__name__} is already active.')
 
         await self._state.initialize()
-
-        self._runtime_offset = self.state.crawler_runtime
+        # Reset `crawler_finished_at` to indicate a new run in progress.
+        self.state.crawler_finished_at = None
 
         # Start periodic logging and let it print initial state before activation.
         self._periodic_logger.start()
@@ -200,10 +197,6 @@ class Statistics(Generic[TStatisticsState]):
         # Stop logging and deactivate the statistics to prevent further changes to crawler_runtime
         await self._periodic_logger.stop()
         self.state.crawler_finished_at = datetime.now(timezone.utc)
-        self.state.crawler_runtime = (
-            self._runtime_offset + self.state.crawler_finished_at - self.state.crawler_last_started_at
-        )
-
         self._active = False
         await self._state.teardown()
 
@@ -262,20 +255,8 @@ class Statistics(Generic[TStatisticsState]):
 
         del self._requests_in_progress[request_id_or_key]
 
-    def _update_crawler_runtime(self) -> None:
-        current_run_duration = (
-            (datetime.now(timezone.utc) - self.state.crawler_last_started_at)
-            if self.state.crawler_last_started_at
-            else timedelta()
-        )
-        self.state.crawler_runtime = current_run_duration + self._runtime_offset
-
     def calculate(self) -> FinalStatistics:
         """Calculate the current statistics."""
-        if self._active:
-            # Only update state when active. If not, just report the last known runtime.
-            self._update_crawler_runtime()
-
         total_minutes = self.state.crawler_runtime.total_seconds() / 60
         state = self._state.current_value
         serialized_state = state.model_dump(by_alias=False)

{crawlee-1.0.5b20.dist-info → crawlee-1.0.5b22.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: crawlee
-Version: 1.0.5b20
+Version: 1.0.5b22
 Summary: Crawlee for Python
 Project-URL: Apify Homepage, https://apify.com
 Project-URL: Changelog, https://crawlee.dev/python/docs/changelog

{crawlee-1.0.5b20.dist-info → crawlee-1.0.5b22.dist-info}/RECORD

@@ -65,7 +65,7 @@ crawlee/crawlers/_adaptive_playwright/_rendering_type_predictor.py,sha256=TM4mkb
 crawlee/crawlers/_adaptive_playwright/_result_comparator.py,sha256=NAfw5VKzTnkvARtLr_zrZj6UGeMp05Voc6Oi8oPxU3w,1747
 crawlee/crawlers/_adaptive_playwright/_utils.py,sha256=EUYVz5i2YkLpL_gbVRp9BAD5u6w1xJ_AFzc_qB9bdDQ,1102
 crawlee/crawlers/_basic/__init__.py,sha256=LPln8SiBBXSMqrApiFUfpqz3hvqxN5HUa1cHQXMVKgU,280
-crawlee/crawlers/_basic/_basic_crawler.py,sha256=yZ_A_l9Dux9Y2eYa9XbN3c7h-3YO7MgGmJbzCMbCplg,73257
+crawlee/crawlers/_basic/_basic_crawler.py,sha256=-lo9yMjPkT8fU4ndOMaeEiwWHgu-DHw9Fny6f1kPPkk,73004
 crawlee/crawlers/_basic/_basic_crawling_context.py,sha256=fjxm2RQXMDkDlWu38dQ3xn5rrGUOhJXkXiqkgbFJFk4,155
 crawlee/crawlers/_basic/_context_pipeline.py,sha256=vM8EEvnCoguERjRV3oyrxUq2Ln2F9DzY7P5dAEiuMHo,5869
 crawlee/crawlers/_basic/_logging_utils.py,sha256=jp5mEwSq5a_BgzUhNPJ9WrIDcoIeYGbeHstcRqCcP0s,3093
@@ -92,7 +92,7 @@ crawlee/crawlers/_playwright/_playwright_pre_nav_crawling_context.py,sha256=fEI2
 crawlee/crawlers/_playwright/_types.py,sha256=hMKA9K9gjzQuwwbnmmfJsQrwR-kq235jH-WBXWeqkGo,2174
 crawlee/crawlers/_playwright/_utils.py,sha256=FQ_-LYo7DGHsNHRrTtWt3mC06VzQvQ2wkGqpA2wBzYU,3441
 crawlee/events/__init__.py,sha256=YMgOXKI0LsXfImKQy06PZ2Vdjy-uD_-acioagHft1do,577
-crawlee/events/_event_manager.py,sha256=kP5_zO2JmFReWnCZqg_uSS1kSV4reem2XN3oRY-SGcI,11428
+crawlee/events/_event_manager.py,sha256=wjZTYIKBI8daKUkOVxUrbPHuU8LnFpUtWStdkts7r3U,11588
 crawlee/events/_local_event_manager.py,sha256=CSiMJ6a_BwX0PPwtffEOtHm21dmALJz1zifo3AuMAk8,3708
 crawlee/events/_types.py,sha256=MKsI014OOKKhjPJRrvWYrezIDGoLjGGhWXrkqYw26Ns,3313
 crawlee/events/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -137,7 +137,7 @@ crawlee/request_loaders/_request_list.py,sha256=SIalHBMuFanE5GLnFocI0QCppWUiJQjr
 crawlee/request_loaders/_request_loader.py,sha256=2Bg-AWWkIV1W-Dwjqo91dPY8nmc7H3teQy7d6OSgliQ,3620
 crawlee/request_loaders/_request_manager.py,sha256=qFizyJuV2meIb9iiPfuii7ciuERMrp4SldAufiH46dc,3000
 crawlee/request_loaders/_request_manager_tandem.py,sha256=lv-s94KPsoQAqx1KaXFch96ejhO147uOflF3UK5ORTk,4058
-crawlee/request_loaders/_sitemap_request_loader.py,sha256=s65D_N0mZxeIrGJEjqUYfu1uYj2AXSOkmErSnfAHv2A,15554
+crawlee/request_loaders/_sitemap_request_loader.py,sha256=W1_k_Szrtk0iE2LJBkHrrFeDtcKReXzr3DG32EnQaQE,16565
 crawlee/sessions/__init__.py,sha256=dJdelbL-6MK5sW4SMU4QrjFbb9kRZ9uRnN-VS3R5-8Y,190
 crawlee/sessions/_cookies.py,sha256=ihYbmpXfCzClzXDT7M2wefB_3KVzcMUdIzTZo6uGk6Y,9356
 crawlee/sessions/_models.py,sha256=JMRQgDUP30XUdZ32isncHowOsXvK9jC_m9QYegbBI1E,2916
@@ -147,8 +147,8 @@ crawlee/sessions/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 crawlee/statistics/__init__.py,sha256=lXAsHNkeRZQBffW1B7rERarivXIUJveNlcKTGOXQZY0,154
 crawlee/statistics/_error_snapshotter.py,sha256=g-roZgkJ-glyStZL7gXrOhrpdZvZ686W9lR43uZjPao,3279
 crawlee/statistics/_error_tracker.py,sha256=x9Yw1TuyEptjwgPPJ4gIom-0oVjawcNReQDsHH2nZ3w,8553
-crawlee/statistics/_models.py,sha256=SFWYpT3r1c4XugU8nrm0epTpcM5_0fS1mXi9fnbhGJ8,5237
-crawlee/statistics/_statistics.py,sha256=AnxbVq6w8fuiRumUJMznhTtQmtONyF4pzqrFYgO4yjo,13076
+crawlee/statistics/_models.py,sha256=n4sT35D4dqNPYREl8Q_YXANZtxaWC0HaZizobA4qK_c,6674
+crawlee/statistics/_statistics.py,sha256=mSgnCnV7q2buJdyuXBxbUU9MQEUjxaLST_NO4ej3XRw,12341
 crawlee/storage_clients/__init__.py,sha256=X3M6Z_WAOJ3M9I8JhGhJDnrtbCOmM27DpGAzgt87R2A,874
 crawlee/storage_clients/models.py,sha256=gfW_kpSCOBuoTBIW0N7tb3FUv7BgD3keZADS7pyT4_I,6586
 crawlee/storage_clients/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -199,8 +199,8 @@ crawlee/storages/_request_queue.py,sha256=bjBOGbpMaGUsqJPVB-JD2VShziPAYMI-GvWKKp
 crawlee/storages/_storage_instance_manager.py,sha256=72n0YlPwNpSQDJSPf4TxnI2GvIK6L-ZiTmHRbFcoVU0,8164
 crawlee/storages/_utils.py,sha256=Yz-5tEBYKYCFJemYT29--uGJqoJLApLDLgPcsnbifRw,439
 crawlee/storages/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-crawlee-1.0.5b20.dist-info/METADATA,sha256=Lxk3f7BsX5OyPPnvQNVgvj-9XR6TphvJRaSg7ZriFZU,29533
-crawlee-1.0.5b20.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-crawlee-1.0.5b20.dist-info/entry_points.txt,sha256=1p65X3dA-cYvzjtlxLL6Kn1wpY-3uEDVqJLp53uNPeo,45
-crawlee-1.0.5b20.dist-info/licenses/LICENSE,sha256=AsFjHssKjj4LGd2ZCqXn6FBzMqcWdjQre1byPPSypVw,11355
-crawlee-1.0.5b20.dist-info/RECORD,,
+crawlee-1.0.5b22.dist-info/METADATA,sha256=GOYPQgm__OcaRBA8drb9Ay30Ddois3nthwX-xMME6As,29533
+crawlee-1.0.5b22.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+crawlee-1.0.5b22.dist-info/entry_points.txt,sha256=1p65X3dA-cYvzjtlxLL6Kn1wpY-3uEDVqJLp53uNPeo,45
+crawlee-1.0.5b22.dist-info/licenses/LICENSE,sha256=AsFjHssKjj4LGd2ZCqXn6FBzMqcWdjQre1byPPSypVw,11355
+crawlee-1.0.5b22.dist-info/RECORD,,