Implement a request fingerprinter that accounts for dependencies

README.rst

@@ -63,6 +63,7 @@ Add the following inside Scrapy's ``settings.py`` file:
     SPIDER_MIDDLEWARES = {
         "scrapy_poet.RetryMiddleware": 275,
     }
+    REQUEST_FINGERPRINTER_CLASS = "scrapy_poet.ScrapyPoetRequestFingerprinter"
 
 Developing
 ==========

docs/intro/install.rst

@@ -32,6 +32,7 @@ of your Scrapy project:
     SPIDER_MIDDLEWARES = {
         "scrapy_poet.RetryMiddleware": 275,
     }
+    REQUEST_FINGERPRINTER_CLASS = "scrapy_poet.ScrapyPoetRequestFingerprinter"
 
 Things that are good to know
 ============================

docs/settings.rst

@@ -107,3 +107,18 @@ Sets the class, or its import path, that will be used as an adapter in the
 generated test fixtures.
 
 More info at :ref:`fixture-adapter`.
+
+
+SCRAPY_POET_REQUEST_FINGERPRINTER_BASE_CLASS
+--------------------------------------------
+
+The default value is the default value of the ``REQUEST_FINGERPRINTER_CLASS``
+setting for the version of Scrapy currently installed (e.g.
+``"scrapy.utils.request.RequestFingerprinter"``).
+
+You can assign a request fingerprinter class to this setting to configure a
+custom request fingerprinter class to use for requests.
+
+This class is used to generate a base fingerprint for a request. If that
+request uses dependency injection, that fingerprint is then modified to account
+for requested dependencies. Otherwise, the fingerprint is used as is.

pyproject.toml

@@ -4,6 +4,8 @@ line-length = 88
 [tool.isort]
 profile = "black"
 multi_line_output = 3
+# scrapy_poet/__init__.py: Automatic sorting causes circular dependencies.
+skip = ["scrapy_poet/__init__.py"]
 
 [[tool.mypy.overrides]]
 module = [

scrapy_poet/__init__.py

@@ -2,3 +2,4 @@
 from .downloadermiddlewares import InjectionMiddleware
 from .page_input_providers import HttpResponseProvider, PageObjectInputProvider
 from .spidermiddlewares import RetryMiddleware
+from ._request_fingerprinter import ScrapyPoetRequestFingerprinter

scrapy_poet/_request_fingerprinter.py

@@ -0,0 +1,102 @@
+try:
+    from scrapy.utils.request import RequestFingerprinter  # NOQA
+except ImportError:
+    from typing import TYPE_CHECKING
+
+    if not TYPE_CHECKING:
+        ScrapyPoetRequestFingerprinter = None
+else:
+    import hashlib
+    import json
+    from functools import cached_property
+    from typing import Callable, Dict, List, Optional, get_args, get_origin
+    from weakref import WeakKeyDictionary
+
+    from scrapy import Request
+    from scrapy.crawler import Crawler
+    from scrapy.settings.default_settings import REQUEST_FINGERPRINTER_CLASS
+    from scrapy.utils.misc import create_instance, load_object
+
+    from scrapy_poet import InjectionMiddleware
+    from scrapy_poet.injection import get_callback
+
+    def _serialize_dep(cls):
+        try:
+            from typing import Annotated
+        except ImportError:
+            pass
+        else:
+            if get_origin(cls) is Annotated:
+                annotated, *annotations = get_args(cls)
+                return f"{_serialize_dep(annotated)}{repr(annotations)}"
+        return f"{cls.__module__}.{cls.__qualname__}"
+
+    class ScrapyPoetRequestFingerprinter:
+        @classmethod
+        def from_crawler(cls, crawler):
+            return cls(crawler)
+
+        def __init__(self, crawler: Crawler) -> None:
+            settings = crawler.settings
+            self._base_request_fingerprinter = create_instance(
+                load_object(
+                    settings.get(
+                        "SCRAPY_POET_REQUEST_FINGERPRINTER_BASE_CLASS",
+                        REQUEST_FINGERPRINTER_CLASS,
+                    )
+                ),
+                settings=crawler.settings,
+                crawler=crawler,
+            )
+            self._callback_cache: Dict[Callable, bytes] = {}
+            self._request_cache: "WeakKeyDictionary[Request, bytes]" = (
+                WeakKeyDictionary()
+            )
+            self._crawler: Crawler = crawler
+
+        @cached_property
+        def _injector(self):
+            middlewares = self._crawler.engine.downloader.middleware.middlewares
+            for middleware in middlewares:
+                if isinstance(middleware, InjectionMiddleware):
+                    return middleware.injector
+            raise RuntimeError(
+                "scrapy_poet.InjectionMiddleware not found at run time, has it "
+                "been configured in the DOWNLOADER_MIDDLEWARES setting?"
+            )
+
+        def _get_deps(self, request: Request) -> Optional[List[str]]:
+            """Return a JSON-serializable structure that uniquely identifies the
+            dependencies requested by the request, or None if dependency injection
+            is not required."""
+            plan = self._injector.build_plan(request)
+            root_deps = plan[-1][1]
+            if not root_deps:
+                return None
+            return sorted([_serialize_dep(cls) for cls in root_deps.values()])
+
+        def fingerprint_deps(self, request: Request) -> Optional[bytes]:
+            """Return a fingerprint based on dependencies requested through
+            scrapy-poet injection, or None if no injection was requested."""
+            callback = get_callback(request, self._crawler.spider)
+            if callback in self._callback_cache:
+                return self._callback_cache[callback]
+
+            deps = self._get_deps(request)
+            if deps is None:
+                return None
+
+            deps_key = json.dumps(deps, sort_keys=True).encode()
+            self._callback_cache[callback] = hashlib.sha1(deps_key).digest()
+            return self._callback_cache[callback]
+
+        def fingerprint(self, request: Request) -> bytes:
+            if request in self._request_cache:
+                return self._request_cache[request]
+            fingerprint = self._base_request_fingerprinter.fingerprint(request)
+            deps_fingerprint = self.fingerprint_deps(request)
+            if deps_fingerprint is None:
+                return fingerprint
+            fingerprints = fingerprint + deps_fingerprint
+            self._request_cache[request] = hashlib.sha1(fingerprints).digest()
+            return self._request_cache[request]

scrapy_poet/utils/testing.py

@@ -3,11 +3,12 @@
 from typing import Dict
 from unittest import mock
 
-from scrapy import signals
+from scrapy import Spider, signals
 from scrapy.crawler import Crawler
 from scrapy.exceptions import CloseSpider
 from scrapy.settings import Settings
 from scrapy.utils.python import to_bytes
+from scrapy.utils.test import get_crawler as _get_crawler
 from twisted.internet import reactor
 from twisted.internet.defer import inlineCallbacks
 from twisted.internet.task import deferLater
@@ -151,6 +152,10 @@ def crawl_single_item(
     return item, url, crawler
 
 
+def get_download_handler(crawler, schema):
+    return crawler.engine.downloader.handlers._get_handler(schema)
+
+
 def make_crawler(spider_cls, settings):
     if not getattr(spider_cls, "name", None):
 
@@ -163,6 +168,33 @@ class Spider(spider_cls):
     return Crawler(spider_cls, settings)
 
 
+def setup_crawler_engine(crawler: Crawler):
+    """Run the crawl steps until engine setup, so that crawler.engine is not
+    None.
+    https://github.com/scrapy/scrapy/blob/8fbebfa943c3352f5ba49f46531a6ccdd0b52b60/scrapy/crawler.py#L116-L122
+    """
+
+    crawler.crawling = True
+    crawler.spider = crawler._create_spider()
+    crawler.engine = crawler._create_engine()
+
+    handler = get_download_handler(crawler, "https")
+    if hasattr(handler, "engine_started"):
+        handler.engine_started()
+
+
+class DummySpider(Spider):
+    name = "dummy"
+
+
+def get_crawler(settings=None, spider_cls=DummySpider, setup_engine=True):
+    settings = settings or {}
+    crawler = _get_crawler(settings_dict=settings, spidercls=spider_cls)
+    if setup_engine:
+        setup_crawler_engine(crawler)
+    return crawler
+
+
 class CollectorPipeline:
     def open_spider(self, spider):
         spider.collected_items = []