Frontera scheduler for Scrapy

More flexible and featured Frontera scheduler for scrapy, which don't force to reimplement capabilities already present in scrapy, so it provides:

• Scrapy handled request dupefilter
• Scrapy handled disk and memory request queues
• Only send to frontera requests marked to be processed by it (using request meta attribute cf_store to True), thus avoiding lot of conflicts.
• Allows to set frontera settings from spider constructor, by loading frontera manager after spider instantiation.
• Allows frontera components to access scrapy stat manager instance by adding STATS_MANAGER frontera setting
• Better request/response converters, fully compatible with ScrapyCloud and Scrapy
• Emulates dont_filter=True scrapy Request flag
• Frontier fingerprint is same as scrapy request fingerprint (can be overridden by passing 'frontier_fingerprint' to request meta)
• allow custom preprocessing or ignoring of request from frontier before actually being enqueued
• Thoroughly tested, used and featured

The result is that crawler using this scheduler will not work differently than a crawler that doesn't use frontier, and reingeneering of a spider in order to be adapted to work with frontier is minimal.

Versions:

Up to version 0.1.8, frontera==0.3.3 and python2 are required. Version 0.2.x requires frontera==0.7.1 and is compatible with python3.

Installation:

pip install scrapy-frontera

Usage and features:

Note: In the context of this doc, a producer spider is the spider that writes requests to the frontier, and the consumer is the one that reads them from the frontier. They can be either the same spider or separated ones.

In your project settings.py:

SCHEDULER = 'scrapy_frontera.scheduler.FronteraScheduler'

DOWNLOADER_MIDDLEWARES = {
    'scrapy_frontera.middlewares.SchedulerDownloaderMiddleware': 0,
}

SPIDER_MIDDLEWARES = {
    'scrapy_frontera.middlewares.SchedulerSpiderMiddleware': 0,
}

# Set to True if you want start requests to be redirected to frontier
# By default they go directly to scrapy downloader
# FRONTERA_SCHEDULER_START_REQUESTS_TO_FRONTIER = False

# Allows to redirect to frontier, the requests with the given callback names
# Important: this setting doesn't affect start requests.
# FRONTERA_SCHEDULER_REQUEST_CALLBACKS_TO_FRONTIER = []

# Spider attributes that need to be passed to the requests redirected to frontier
# Some previous callbacks may have generated some state needed for following ones.
# This setting allows to transmit that state between different jobs
# FRONTERA_SCHEDULER_STATE_ATTRIBUTES = []

# map specific requests to specific slot prefix by its callback name.
# FRONTERA_SCHEDULER_CALLBACK_SLOT_PREFIX_MAP = {}

Plus the usual Frontera setup. For instance, for hcf-backend:

BACKEND = 'hcf_backend.HCFBackend'
HCF_PROJECT_ID = 11111

(etc...)

You can also set up spider specific frontera settings via the spider class attribute dict frontera_settings. Example with hcf backend:

class MySpider(Spider):

    name = 'my-producer'

    frontera_settings = {
        'HCF_AUTH': 'xxxxxxxxxx',
        'HCF_PROJECT_ID': 11111,
        'HCF_PRODUCER_FRONTIER': 'myfrontier',
        'HCF_PRODUCER_NUMBER_OF_SLOTS': 8,
    }

Scrapy-frontera also accepts the spider attribute frontera_settings_json. This is specially useful for consumers, which need per job setup of reading slot.For example, you can configure a consumer spider in this way, for usage with hcf backend:

class MySpider(Spider):

    name = 'my-consumer'

    frontera_settings = {
        'HCF_AUTH': 'xxxxxxxxxx',
        'HCF_PROJECT_ID': 11111,
        'HCF_CONSUMER_FRONTIER': 'myfrontier',
    }

and invoke it via:

scrapy crawl my-consumer -a frontera_settings_json='{"HCF_CONSUMER_SLOT": "0"}'

Settings provided through frontera_settings_json overrides those provided using frontera_settings, which in turn overrides those provided in the project settings.py file.

Requests will go through the Frontera pipeline only if the flag cf_store with value True is included in the request meta. If cf_store is not present or is False, requests will be processed as normal scrapy request. An alternative to cf_store flag are the scrapy settings FRONTERA_SCHEDULER_START_REQUESTS_TO_FRONTIER and FRONTERA_SCHEDULER_REQUEST_CALLBACKS_TO_FRONTIER (see above about usage of these settings)

Requests read from the frontier are directly enqueued by the scheduler. This means that they are not processed by spider middleware. Their processing entrypoint is downloader middleware process_request() pipeline. But if you need to preprocess requests incoming from the frontier in the spider, you can define the spider method preprocess_request_from_frontier(request: scrapy.Request). If defined, the scheduler will invoke it before actually enqueuing it. This method must returns either None or a request (same from the call, or another). This return value is what will be actually enqueued, so if it is None, request is skipped (not enqueued).

If requests read from frontier doesn't already have an errback defined, the scheduler will automatically assign the consumer spider errback method, if it exists, to them. This is specially useful when consumer spider is not the same as the producer one.

Another useful setting is FRONTERA_SCHEDULER_CALLBACK_SLOT_PREFIX_MAP. This is a dict which allows to map requests with a specific callback, to a specific slot prefix, and optionally a number of slots, different than the default one assigned by frontera backend (this feature has to be supported by the specific frontera backend you will use, last versions of hcf-backend does supports it). For example:

class MySpider(Spider):

    name = 'my-producer'

    frontera_settings = {
        'HCF_AUTH': 'xxxxxxxxxx',
        'HCF_PROJECT_ID': 11111,
        'HCF_PRODUCER_FRONTIER': 'myfrontier',
        'HCF_PRODUCER_SLOT_PREFIX': 'my-consumer'
        'HCF_PRODUCER_NUMBER_OF_SLOTS': 8,
    }

    custom_settings = {
        'FRONTERA_SCHEDULER_CALLBACK_SLOT_PREFIX_MAP': {'parse': 'my-producer/4'},
        'FRONTERA_SCHEDULER_REQUEST_CALLBACKS_TO_FRONTIER': ['parse', 'parse_consumer']
    }

    def parse_consumer(self, response):
        assert False

    def parse(self, response):
        (...)

Under this configuration, requests with callback parse() will be saved on 4 slots with prefix my-producer, while requests with callback parse_consumer() will use the configuration from hcf settings, that is, 8 slot with prefix my-consumer.

An integrated tutorial is available at shub-workflow Tutorial