fix: Also load input from a file with a .json extension in file system storage (#617)

- fixes bug reported in
https://apify.slack.com/archives/C0L33UM7Z/p1759754068120539

---------

Co-authored-by: Vlada Dusek <[email protected]>

Author: janbuchar

pyproject.toml

@@ -36,7 +36,7 @@ keywords = [
 dependencies = [
     "apify-client>=2.0.0,<3.0.0",
     "apify-shared>=2.0.0,<3.0.0",
-    "crawlee>=1.0.0,<2.0.0",
+    "crawlee>=1.0.2,<2.0.0",
     "cachetools>=5.5.0",
     "cryptography>=42.0.0",
     "impit>=0.6.1",

src/apify/_configuration.py

@@ -3,6 +3,7 @@
 from datetime import datetime, timedelta
 from decimal import Decimal
 from logging import getLogger
+from pathlib import Path
 from typing import Annotated, Any
 
 from pydantic import AliasChoices, BeforeValidator, Field, model_validator
@@ -421,6 +422,14 @@ def disable_browser_sandbox_on_platform(self) -> Self:
             logger.warning('Actor is running on the Apify platform, `disable_browser_sandbox` was changed to True.')
         return self
 
+    @property
+    def canonical_input_key(self) -> str:
+        return str(Path(self.input_key).with_suffix('.json'))
+
+    @property
+    def input_key_candidates(self) -> set[str]:
+        return {self.input_key, self.canonical_input_key, Path(self.canonical_input_key).stem}
+
     @classmethod
     def get_global_configuration(cls) -> Configuration:
         """Retrieve the global instance of the configuration.

src/apify/storage_clients/_file_system/_key_value_store_client.py

@@ -1,14 +1,19 @@
 import asyncio
 import json
-from pathlib import Path
+import logging
 
-from typing_extensions import override
+from more_itertools import flatten
+from typing_extensions import Self, override
 
 from crawlee._consts import METADATA_FILENAME
+from crawlee.configuration import Configuration as CrawleeConfiguration
 from crawlee.storage_clients._file_system import FileSystemKeyValueStoreClient
+from crawlee.storage_clients.models import KeyValueStoreRecord
 
 from apify._configuration import Configuration
 
+logger = logging.getLogger(__name__)
+
 
 class ApifyFileSystemKeyValueStoreClient(FileSystemKeyValueStoreClient):
     """Apify-specific implementation of the `FileSystemKeyValueStoreClient`.
@@ -17,23 +22,39 @@ class ApifyFileSystemKeyValueStoreClient(FileSystemKeyValueStoreClient):
     directory, except for the metadata file and the `INPUT.json` file.
     """
 
+    @override
+    @classmethod
+    async def open(
+        cls,
+        *,
+        id: str | None,
+        name: str | None,
+        alias: str | None,
+        configuration: CrawleeConfiguration,
+    ) -> Self:
+        client = await super().open(id=id, name=name, alias=alias, configuration=configuration)
+
+        await client._sanitize_input_json_files()  # noqa: SLF001 - it's okay, this is a factory method
+
+        return client
+
     @override
     async def purge(self) -> None:
         """Purges the key-value store by deleting all its contents.
 
         It deletes all files in the key-value store directory, except for the metadata file and
         the `INPUT.json` file. It also updates the metadata to reflect that the store has been purged.
         """
-        kvs_input_key = Configuration.get_global_configuration().input_key
-
-        # First try to find the alternative format of the input file and process it if it exists.
-        for file_path in self.path_to_kvs.glob('*'):
-            if file_path.name == f'{kvs_input_key}.json':
-                await self._process_input_json(file_path)
+        configuration = Configuration.get_global_configuration()
 
         async with self._lock:
+            files_to_keep = set(
+                flatten([key, f'{key}.{METADATA_FILENAME}'] for key in configuration.input_key_candidates)
+            )
+            files_to_keep.add(METADATA_FILENAME)
+
             for file_path in self.path_to_kvs.glob('*'):
-                if file_path.name in {METADATA_FILENAME, kvs_input_key, f'{kvs_input_key}.{METADATA_FILENAME}'}:
+                if file_path.name in files_to_keep:
                     continue
                 if file_path.is_file():
                     await asyncio.to_thread(file_path.unlink, missing_ok=True)
@@ -43,15 +64,40 @@ async def purge(self) -> None:
                 update_modified_at=True,
             )
 
-    async def _process_input_json(self, path: Path) -> None:
-        """Process simple input json file to format expected by the FileSystemKeyValueStoreClient.
+    async def _sanitize_input_json_files(self) -> None:
+        """Handle missing metadata for input files."""
+        configuration = Configuration.get_global_configuration()
+        alternative_keys = configuration.input_key_candidates - {configuration.canonical_input_key}
 
-        For example: INPUT.json -> INPUT, INPUT.json.metadata
-        """
-        try:
-            f = await asyncio.to_thread(path.open)
-            input_data = json.load(f)
-        finally:
-            f.close()
-        await asyncio.to_thread(path.unlink, missing_ok=True)
-        await self.set_value(key=path.stem, value=input_data)
+        if (self.path_to_kvs / configuration.canonical_input_key).exists():
+            # Refresh metadata to prevent inconsistencies
+            input_data = await asyncio.to_thread(
+                lambda: json.loads((self.path_to_kvs / configuration.canonical_input_key).read_text())
+            )
+            await self.set_value(key=configuration.canonical_input_key, value=input_data)
+
+            for alternative_key in alternative_keys:
+                if (alternative_input_file := self.path_to_kvs / alternative_key).exists():
+                    logger.warning(f'Redundant input file found: {alternative_input_file}')
+        else:
+            for alternative_key in alternative_keys:
+                alternative_input_file = self.path_to_kvs / alternative_key
+
+                # Only process files that actually exist
+                if alternative_input_file.exists():
+                    # Refresh metadata to prevent inconsistencies
+                    with alternative_input_file.open() as f:
+                        input_data = await asyncio.to_thread(lambda: json.load(f))
+                    await self.set_value(key=alternative_key, value=input_data)
+
+    @override
+    async def get_value(self, *, key: str) -> KeyValueStoreRecord | None:
+        configuration = Configuration.get_global_configuration()
+
+        if key in configuration.input_key_candidates:
+            for candidate in configuration.input_key_candidates:
+                value = await super().get_value(key=candidate)
+                if value is not None:
+                    return value
+
+        return await super().get_value(key=key)

tests/unit/storage_clients/test_file_system.py

@@ -2,16 +2,16 @@
 
 import asyncio
 import json
-from typing import TYPE_CHECKING
+from pathlib import Path
 
+import pytest
+
+from crawlee import service_locator
 from crawlee._consts import METADATA_FILENAME
 
 from apify import Actor, Configuration
 from apify.storage_clients._file_system import ApifyFileSystemKeyValueStoreClient
 
-if TYPE_CHECKING:
-    from pathlib import Path
-
 
 async def test_purge_preserves_input_file_and_metadata() -> None:
     """Test that purge() preserves INPUT.json and metadata files but removes other files."""
@@ -61,19 +61,32 @@ async def test_purge_preserves_input_file_and_metadata() -> None:
 
     # Verify INPUT.json content is unchanged
     input_content = await asyncio.to_thread(input_file.read_text)
-    assert input_content == '{"test": "input"}'
+    assert json.loads(input_content) == json.loads('{"test": "input"}')
+
 
+@pytest.mark.parametrize('input_file_name', ['INPUT', 'INPUT.json'])
+async def test_pre_existing_input_used_by_actor(input_file_name: str) -> None:
+    configuration = Configuration()
+    service_locator.set_configuration(configuration)
+
+    # Create key-value store directory and make sure that it is empty
+    path_to_input = Path(configuration.storage_dir) / 'key_value_stores' / 'default'
+    path_to_input.mkdir(parents=True)
+    assert list(path_to_input.glob('*')) == []
 
-async def test_pre_existing_input_used_by_actor(tmp_path: Path) -> None:
     pre_existing_input = {
         'foo': 'bar',
     }
 
-    configuration = Configuration.get_global_configuration()
     # Create pre-existing INPUT.json file
-    path_to_input = tmp_path / 'key_value_stores' / 'default'
-    path_to_input.mkdir(parents=True)
-    (path_to_input / f'{configuration.input_key}.json').write_text(json.dumps(pre_existing_input))
+    (path_to_input / input_file_name).write_text(json.dumps(pre_existing_input))
 
     async with Actor():
         assert pre_existing_input == await Actor.get_input()
+
+    # Make sure that the input file doesn't get renamed in the process and metadata are added
+    assert set(path_to_input.glob('*')) == {
+        path_to_input / '__metadata__.json',
+        path_to_input / input_file_name,
+        path_to_input / f'{input_file_name}.__metadata__.json',
+    }