scrapely-python 0.1.9__py3-none-any.whl → 0.1.11__py3-none-any.whl

scrapely/__init__.py

@@ -132,7 +132,7 @@ from .errors import (
     raise_for_status,
 )
 
-__version__ = '0.1.9'
+__version__ = '0.1.11'
 __all__ = [
     # Core
     'Actor',

scrapely/actor.py

@@ -24,7 +24,12 @@ from crawlee.configuration import Configuration as CrawleeConfiguration
 from crawlee.storage_clients._base import StorageClient
 from crawlee.storages import Dataset, KeyValueStore, RequestQueue
 
-from .charging import ChargingManager, is_default_dataset_item_event, push_data_and_charge
+from .charging import (
+    ChargingManager,
+    is_default_dataset_item_event,
+    push_data_and_charge,
+    set_active_charging_manager,
+)
 from .configuration import Configuration
 from .crypto import decrypt_secret_inputs
 from .proxy_configuration import with_proxy_key_aliases
@@ -515,6 +520,7 @@ class Actor(metaclass=_ActorMeta):
             # Clear charging + patched client so a second Actor lifecycle cannot
             # post charges to the previous run_id / credentials.
             cls._charging_manager = None
+            set_active_charging_manager(None)
             cls._apify_client_instance = None
 
             cls._default_dataset = None
@@ -1222,6 +1228,8 @@ class Actor(metaclass=_ActorMeta):
             # Prefer the storage factory client (same URL/token/run as API storage).
             client = cls._client if cls._client is not None else cls.new_client()
             cls._charging_manager = ChargingManager(config, client)
+            # Storage clients refuse default-dataset writes past this run's limit.
+            set_active_charging_manager(cls._charging_manager)
         return cls._charging_manager
 
 

scrapely/charging.py

@@ -9,7 +9,9 @@ never treated as retries of a previous operation.
 from __future__ import annotations
 
 import asyncio
+import contextvars
 import json
+import os
 import logging
 import time
 import uuid
@@ -63,6 +65,50 @@ _RETRY_BUDGET_SECS = 60.0
 LOCAL_CHARGING_LOG_DATASET_NAME = 'charging_log'
 
 
+# Charge-limit safety net for the default dataset.
+#
+# Once a pay-per-event run has used up its spending limit (maxTotalChargeUsd),
+# rows written to the run's default dataset would go out unbilled. The SDK then
+# refuses such writes: nothing is stored, one warning is logged and
+# ``ChargingManager.is_charge_limit_reached()`` turns True. Pushes made by
+# ``push_data_and_charge`` (``Actor.push_data``) are decided there and marked
+# with this context variable, so the storage client does not check them twice.
+_charge_gated_push: contextvars.ContextVar[bool] = contextvars.ContextVar(
+    'scrapely_charge_gated_push', default=False
+)
+
+# The charging manager of the running Actor (set by ``Actor``); storage clients
+# consult it before writing to the default dataset.
+_active_charging_manager: Optional['ChargingManager'] = None
+
+
+def set_active_charging_manager(manager: Optional['ChargingManager']) -> None:
+    """Register the charging manager whose limit guards default-dataset writes."""
+    global _active_charging_manager
+    _active_charging_manager = manager
+
+
+def default_dataset_id_from_env() -> Optional[str]:
+    """The run's default dataset ID, as the platform passes it."""
+    return os.environ.get('ACTOR_DEFAULT_DATASET_ID') or os.environ.get('APIFY_DEFAULT_DATASET_ID')
+
+
+def refuse_default_dataset_push(item_count: int) -> bool:
+    """True when a write of ``item_count`` rows to the default dataset must be dropped.
+
+    Called by storage clients for writes that did not go through
+    ``push_data_and_charge`` (``Actor.open_dataset().push_data()``, crawler
+    ``context.push_data``, ...). Non-PPE runs and runs below their limit are
+    never refused.
+    """
+    if _charge_gated_push.get():
+        return False
+    manager = _active_charging_manager
+    if manager is None:
+        return False
+    return manager._refuse_default_dataset_push(item_count)
+
+
 def is_default_dataset_item_event(event_name: str) -> bool:
     """Check if event name is a default dataset item event."""
     return event_name in (APIFY_DEFAULT_DATASET_ITEM_EVENT, SCRAPELY_DEFAULT_DATASET_ITEM_EVENT)
@@ -209,10 +255,69 @@ class ChargingManager:
         self._last_accepted_operation_id: Optional[str] = None
         # Voids not yet confirmed by the platform; retried before the next push and at exit.
         self._pending_voids: List[tuple[str, str, int, bool]] = []
+        # Sticky: the platform refused or capped a charge (HTTP 402 / fewer
+        # events accepted) or the local budget gate trimmed one.
+        self._charge_limit_hit = False
+        # Default-dataset rows refused after the limit (one warning, then a count).
+        self._refused_default_dataset_items = 0
 
     def close(self) -> None:
         """Mark the manager closed; subsequent charge/push operations are rejected."""
         self._closed = True
+        if self._refused_default_dataset_items:
+            logger.warning(
+                '%s result(s) pushed to the default dataset after the spending limit was '
+                'reached were not saved',
+                self._refused_default_dataset_items,
+            )
+
+    def is_charge_limit_reached(self) -> bool:
+        """Whether this pay-per-event run has used up its spending limit.
+
+        True once a charge was refused or capped by the limit (the platform's
+        HTTP 402 or a partial acceptance, or the local budget gate), or once the
+        remaining budget cannot pay for even one event of the cheapest priced
+        type. From then on, pushes to the default dataset that are not paid
+        for are refused. Always False for actors without pay-per-event
+        pricing, without a limit, or with only free events.
+        """
+        if self._pricing_info is None:
+            return False
+        if self._charge_limit_hit:
+            return True
+        if self._max_total_charge_usd is None:
+            return False
+        prices = [price for price in self._per_event_prices.values() if price > 0]
+        if not prices:
+            return False
+        return self._max_total_charge_usd - self._total_charged_usd < min(prices)
+
+    def _mark_charge_limit_hit(self) -> None:
+        if self._pricing_info is not None:
+            self._charge_limit_hit = True
+
+    def _note_refused_default_dataset_items(self, count: int) -> None:
+        """Count rows dropped because of the limit; warn on the first one only."""
+        if count <= 0:
+            return
+        if not self._refused_default_dataset_items:
+            limit = (
+                f'${self._max_total_charge_usd}' if self._max_total_charge_usd is not None else 'the limit'
+            )
+            logger.warning(
+                'The spending limit of this run (maxTotalChargeUsd = %s) has been reached: '
+                'results pushed to the default dataset are no longer saved. The Actor should '
+                'stop now (check ChargeResult.event_charge_limit_reached or '
+                'Actor.get_charging_manager().is_charge_limit_reached()).',
+                limit,
+            )
+        self._refused_default_dataset_items += count
+
+    def _refuse_default_dataset_push(self, item_count: int) -> bool:
+        if item_count <= 0 or not self.is_charge_limit_reached():
+            return False
+        self._note_refused_default_dataset_items(item_count)
+        return True
 
     @property
     def charge_lock(self) -> RLock:
@@ -688,6 +793,8 @@ class ChargingManager:
                             deadline=deadline,
                         )
                     except ChargeRejectedError as err:
+                        if _is_charge_limit_rejection(err):
+                            self._mark_charge_limit_hit()
                         self._unwind_reservation(event_name, count)
                         self._drop_pending(idempotency_key)
                         logger.warning(
@@ -699,6 +806,7 @@ class ChargingManager:
                         continue
                     if accepted is not None:
                         if accepted < count:
+                            self._mark_charge_limit_hit()
                             self._unwind_reservation(event_name, count - accepted)
                         self._drop_pending(idempotency_key)
                         confirmed += 1
@@ -784,6 +892,9 @@ class ChargingManager:
                         else 0
                     )
 
+            if chargeable_count < count:
+                self._mark_charge_limit_hit()
+
             if chargeable_count <= 0:
                 return ChargeResult(
                     chargeable_within_limit=self._compute_chargeable(),
@@ -835,7 +946,9 @@ class ChargingManager:
                     count=chargeable_count,
                     idempotency_key=idempotency_key,
                 )
-            except ChargeRejectedError:
+            except ChargeRejectedError as rejected:
+                if _is_charge_limit_rejection(rejected):
+                    self._mark_charge_limit_hit()
                 # Server definitively rejected: unwind local reservation.
                 self._drop_pending(idempotency_key)
                 self._unwind_reservation(event_name, chargeable_count)
@@ -849,6 +962,8 @@ class ChargingManager:
                 if accepted > 0:
                     self._last_accepted_operation_id = idempotency_key
                 if accepted < chargeable_count:
+                    # The platform capped the charge at the run's limit.
+                    self._mark_charge_limit_hit()
                     self._unwind_reservation(event_name, chargeable_count - accepted)
                     chargeable_count = accepted
                     result = ChargeResult(
@@ -1066,6 +1181,15 @@ class ChargingManager:
 
 
 
+def _is_charge_limit_rejection(err: ChargeRejectedError) -> bool:
+    """Whether a rejected charge was refused because the run's limit is used up."""
+    return (
+        err.status_code == 402
+        or err.reason == 'charge-limit-exceeded'
+        or 'charge-limit-exceeded' in str(err).lower()
+    )
+
+
 def _read_accepted_count(body: Any, fallback: int) -> int:
     if not isinstance(body, dict):
         return fallback
@@ -1166,6 +1290,24 @@ async def push_data_and_charge(
             if fallback_event:
                 resolved_event = fallback_event
 
+        # Charge-limit safety net: once the run's spending limit is used up, a
+        # default-dataset push that does not pay per row (no event, or only
+        # free events) stores nothing. Paid pushes go on below, where the
+        # budget cap stores only the rows that are actually charged.
+        if (
+            is_ppe
+            and is_default_dataset
+            and items
+            and charging_manager.is_charge_limit_reached()
+            and not _push_is_paid_per_item(charging_manager, event_name, resolved_event)
+        ):
+            charging_manager._note_refused_default_dataset_items(len(items))
+            return ChargeResult(
+                chargeable_within_limit=charging_manager.compute_chargeable(),
+                charged_count=0,
+                event_charge_limit_reached=True,
+            )
+
         limited_count = len(items)
         if is_ppe and resolved_event:
             # Keep the explicit event separate from the implicit default-dataset
@@ -1177,6 +1319,10 @@ async def push_data_and_charge(
             )
 
         limited_items = items[:limited_count] if limited_count < len(items) else items
+        if is_ppe and is_default_dataset and limited_count < len(items):
+            # Rows beyond what the budget can pay for are dropped.
+            charging_manager._mark_charge_limit_hit()
+            charging_manager._note_refused_default_dataset_items(len(items) - limited_count)
 
         if not limited_items:
             return ChargeResult(
@@ -1302,9 +1448,13 @@ async def push_data_and_charge(
                             await charging_manager._cancel_uncertain_charge(name, op_id, reserved)
                         else:
                             await charging_manager._void_charge(name, op_id, reserved)
+                if is_default_dataset and accepted_for_items < len(limited_items):
+                    charging_manager._note_refused_default_dataset_items(
+                        len(limited_items) - accepted_for_items
+                    )
                 if accepted_for_items > 0:
                     try:
-                        await push_fn(limited_items[:accepted_for_items])
+                        await _gated_push(push_fn, limited_items[:accepted_for_items])
                     except Exception as push_err:
                         # A result is billed only if it was saved: void this
                         # operation's platform charges and unwind local
@@ -1333,7 +1483,9 @@ async def push_data_and_charge(
                         raise
                 return ChargeResult(
                     chargeable_within_limit=charging_manager.compute_chargeable(),
-                    charged_count=accepted_for_items * len(results) if results else 0,
+                    # Rows charged, as Apify reports it for the pushed event; the
+                    # implicit default-dataset event is not counted on top.
+                    charged_count=accepted_for_items,
                     event_charge_limit_reached=any(
                         result.event_charge_limit_reached for result in results
                     )
@@ -1342,7 +1494,7 @@ async def push_data_and_charge(
 
 
         # No billable events — still persist the budget-limited items.
-        await push_fn(limited_items)
+        await _gated_push(push_fn, limited_items)
         return ChargeResult(
             chargeable_within_limit=charging_manager.compute_chargeable(),
             charged_count=0,
@@ -1350,6 +1502,23 @@ async def push_data_and_charge(
         )
 
 
+def _push_is_paid_per_item(
+    charging_manager: ChargingManager, event_name: Optional[str], resolved_event: Optional[str]
+) -> bool:
+    """Whether every row of this push is charged a positive price."""
+    names = {name for name in (event_name, resolved_event) if name}
+    return any((charging_manager._calculate_event_price(name) or Decimal('0')) > 0 for name in names)
+
+
+async def _gated_push(push_fn: Any, items: List[Dict[str, Any]]) -> Any:
+    """Run a push already decided by push_data_and_charge (storage skips its own check)."""
+    token = _charge_gated_push.set(True)
+    try:
+        return await push_fn(items)
+    finally:
+        _charge_gated_push.reset(token)
+
+
 def _http_status_of(err: BaseException) -> Optional[int]:
     status = getattr(err, 'status_code', None)
     if status is None:

scrapely/charging_client.py

@@ -43,7 +43,11 @@ class _ChargingManagerHooks:
         from .charging import has_default_dataset_item_event_in_pricing
 
         pricing_info = self._charging_manager.get_pricing_info()
-        return has_default_dataset_item_event_in_pricing(pricing_info.per_event_prices or {})
+        # Past the spending limit, route default-dataset pushes through
+        # push_data_and_charge too, which refuses rows nobody pays for.
+        return has_default_dataset_item_event_in_pricing(
+            pricing_info.per_event_prices or {}
+        ) or bool(getattr(self._charging_manager, 'is_charge_limit_reached', lambda: False)())
 
     async def push_and_charge(
         self,

scrapely/storage/scrapely_dataset_client.py

@@ -99,6 +99,12 @@ class ScrapelyDatasetClient(DatasetClient):
             items = [data]
         else:
             items = list(data)
+        # Charge-limit safety net: a pay-per-event run past its spending limit
+        # must not store unbilled rows in its default dataset.
+        from ..charging import default_dataset_id_from_env, refuse_default_dataset_push
+
+        if self._metadata.id == default_dataset_id_from_env() and refuse_default_dataset_push(len(items)):
+            return
         await self._api.dataset_push_items(self._metadata.id, items)
 
     @override

{scrapely_python-0.1.9.dist-info → scrapely_python-0.1.11.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: scrapely-python
-Version: 0.1.9
+Version: 0.1.11
 Summary: Scrapely Python SDK - A toolkit for building Actors on the Scrapely platform
 Author-email: Scrapely <support@scrape.ly>
 License-Expression: MIT AND Apache-2.0

{scrapely_python-0.1.9.dist-info → scrapely_python-0.1.11.dist-info}/RECORD

@@ -1,10 +1,10 @@
-scrapely/__init__.py,sha256=3LP91mpUIJywB0w8vwL1kKhrMnuYE3eHZP_Bj9QjylI,5745
+scrapely/__init__.py,sha256=RxtELQHc6mWoQhDgFu_EK_jFjuPOMlKEcDEZizk1vxQ,5746
 scrapely/_browser_locations.py,sha256=8QVVnkms2bDn3_LDYgYDR59MvFx8757l_JYnZ7xb3oI,145269
-scrapely/actor.py,sha256=cQl9RM1MUZWWR8uNF7q6Pjd8vpB1MCoDp6KM5ITJq5I,77479
+scrapely/actor.py,sha256=2nWeaEIo11-XO0w3QAzA1lbn4Gm1MvjStsiNdOvk_tc,77721
 scrapely/browser_configuration.py,sha256=jAZapNSHLFW05B96df4wNBtcyiiU9DA3ml0yTL1FwfA,5770
 scrapely/cdp_retry.py,sha256=hdTJEAIKhApM7IOk1DrQEl8NoflSJSmqd4RbNJ2IsgI,8540
-scrapely/charging.py,sha256=bwJ6VPJOdxpo0ZHOTrMAe8NCTXeRBuxHN0EDyXbGXss,59138
-scrapely/charging_client.py,sha256=tDtlPwLSVFvoRa2NBgNjyfHkAUM77SSIyYT2ACgpDm8,3710
+scrapely/charging.py,sha256=41KG60D-YM9oi-1qbSKRfZbw92pFWi1yvvc2c2L6V5k,66862
+scrapely/charging_client.py,sha256=-4q-Ww7gu9EZXpM9QIn1Vd4vwAXmgL_WEKqaQIx3ViA,3961
 scrapely/configuration.py,sha256=GjMr3YsxM16HYjAOTUzWZKG2E8FuMzAoFXmkn5dbDPM,16396
 scrapely/errors.py,sha256=DCAlVvE3jToZuPEG2HVDL7i8D8hLboomQ5JpSOcsXoQ,4107
 scrapely/log.py,sha256=iVKoQdMLQXEWz1STzbSXrNUxsS_FNwiIdqoEg7CzeqU,1440
@@ -26,7 +26,7 @@ scrapely/storage/_metadata.py,sha256=WWawBnVN0b1iK2K-SlxZXUc6czb5wXWf8iC_0tzkgAQ
 scrapely/storage/api_client.py,sha256=jqfBOOHo97DeTzJQlh0bBTn0EjjK184VtnxFIs20rbg,16649
 scrapely/storage/client.py,sha256=Ln_NA8nJz62nXldkcj54PFZopX5P3Lp8HGot_8friNM,22015
 scrapely/storage/patch_storage.py,sha256=o3Fv0BByeZ1T9HhkzhY6bjolOFsIQcnqBoHq1EdG8Fw,8758
-scrapely/storage/scrapely_dataset_client.py,sha256=gRvlCDw9c0yO5GE5LQOa1xAyXc9pnb-E37Eq7fHqHCM,10338
+scrapely/storage/scrapely_dataset_client.py,sha256=u5dnXxe1BAsq7rh-haR5W1ChE7UZ_k7B9ShR8YtwFkg,10695
 scrapely/storage/scrapely_kvs_client.py,sha256=ItQhe2rcSl20NkSmegfqHiRyKncH4Qg9BjQVpF6jHfk,7403
 scrapely/storage/scrapely_rq_client.py,sha256=gA4nMDuWyrh6QfbnMv1y7ZgcwFGWIPXGE5Y4OsWA1qI,18920
 scrapely/storage/scrapely_storage_client.py,sha256=N-p2ISICuKQ8RNzmN-JiXcj_lTttULLOGl9lpLl62qI,6567
@@ -38,10 +38,10 @@ scrapely/storage/_file_system/_key_value_store_client.py,sha256=4Ipgbh4qEqAkwQJ9
 scrapely/storage/_file_system/_storage_client.py,sha256=aaGJ4P0QAtIgEMl6yHpuobeWFD5M4rMC5HwhOOvuGuo,1794
 scrapely/storage_clients/__init__.py,sha256=cs1YX7JV2_GKBXeiiJgkQFali08b5_y_jH6Yw7iROAA,656
 scrapely/storages/__init__.py,sha256=4PAELhJNZ3exJ7Uouicn-bCQ58NafKjhLBixQ8ATQc4,173
-scrapely_python-0.1.9.dist-info/licenses/LICENSE,sha256=pQGY0XwjIYh-a2HfawxHq1en53bXkFkDj2Q70jKWHr4,1088
-scrapely_python-0.1.9.dist-info/licenses/LICENSE-APACHE,sha256=N8m0FzMq8i4UbTgXZvKEZJSD-27OF8pWZYS_atz4kTU,11355
-scrapely_python-0.1.9.dist-info/licenses/NOTICE,sha256=DEsNrw6UdSIqdYYqMWVFgJnyt6jUivTD7iFU6YM_e1A,1122
-scrapely_python-0.1.9.dist-info/METADATA,sha256=E_rdzHW3bC3asqFegdYfxL27jTuGhIm1uXvVpGFm_-o,3486
-scrapely_python-0.1.9.dist-info/WHEEL,sha256=YVMoNqKzERt-wjUZwJ33xBGAwnFl-4cqbYkTtWa4itE,91
-scrapely_python-0.1.9.dist-info/top_level.txt,sha256=T4B9UMrs4q99QSAFyD1zdXhl40KnLhBKqrDkPhOTsmY,9
-scrapely_python-0.1.9.dist-info/RECORD,,
+scrapely_python-0.1.11.dist-info/licenses/LICENSE,sha256=pQGY0XwjIYh-a2HfawxHq1en53bXkFkDj2Q70jKWHr4,1088
+scrapely_python-0.1.11.dist-info/licenses/LICENSE-APACHE,sha256=N8m0FzMq8i4UbTgXZvKEZJSD-27OF8pWZYS_atz4kTU,11355
+scrapely_python-0.1.11.dist-info/licenses/NOTICE,sha256=DEsNrw6UdSIqdYYqMWVFgJnyt6jUivTD7iFU6YM_e1A,1122
+scrapely_python-0.1.11.dist-info/METADATA,sha256=TjBOSZzW9n2wrL0IQQQ6Qxp-T_SCsjPsOAP1I_lCTEc,3487
+scrapely_python-0.1.11.dist-info/WHEEL,sha256=YVMoNqKzERt-wjUZwJ33xBGAwnFl-4cqbYkTtWa4itE,91
+scrapely_python-0.1.11.dist-info/top_level.txt,sha256=T4B9UMrs4q99QSAFyD1zdXhl40KnLhBKqrDkPhOTsmY,9
+scrapely_python-0.1.11.dist-info/RECORD,,