Scrapy Very Basic Example

B.Mr.W. picture B.Mr.W. · Sep 17, 2013 · Viewed 22.5k times · Source

Hi I have Python Scrapy installed on my mac and I was trying to follow the very first example on their web.

They were trying to run the command:

scrapy crawl mininova.org -o scraped_data.json -t json

I don't quite understand what does this mean? looks like scrapy turns out to be a separate program. And I don't think they have a command called crawl. In the example, they have a paragraph of code, which is the definition of the class MininovaSpider and the TorrentItem. I don't know where these two classes should go to, go to the same file and what is the name of this python file?

Answer

alecxe picture alecxe · Jan 2, 2015

TL;DR: see Self-contained minimum example script to run scrapy.

First of all, having a normal Scrapy project with a separate .cfg, settings.py, pipelines.py, items.py, spiders package etc is a recommended way to keep and handle your web-scraping logic. It provides a modularity, separation of concerns that keeps things organized, clear and testable.

If you are following the official Scrapy tutorial to create a project, you are running web-scraping via a special scrapy command-line tool:

scrapy crawl myspider

But, Scrapy also provides an API to run crawling from a script.

There are several key concepts that should be mentioned:

  • Settings class - basically a key-value "container" which is initialized with default built-in values
  • Crawler class - the main class that acts like a glue for all the different components involved in web-scraping with Scrapy
  • Twisted reactor - since Scrapy is built-in on top of twisted asynchronous networking library - to start a crawler, we need to put it inside the Twisted Reactor, which is in simple words, an event loop:

The reactor is the core of the event loop within Twisted – the loop which drives applications using Twisted. The event loop is a programming construct that waits for and dispatches events or messages in a program. It works by calling some internal or external “event provider”, which generally blocks until an event has arrived, and then calls the relevant event handler (“dispatches the event”). The reactor provides basic interfaces to a number of services, including network communications, threading, and event dispatching.

Here is a basic and simplified process of running Scrapy from script:

  • create a Settings instance (or use get_project_settings() to use existing settings):

    settings = Settings()  # or settings = get_project_settings()
    
  • instantiate Crawler with settings instance passed in:

    crawler = Crawler(settings)
    
  • instantiate a spider (this is what it is all about eventually, right?):

    spider = MySpider()
    
  • configure signals. This is an important step if you want to have a post-processing logic, collect stats or, at least, to ever finish crawling since the twisted reactor needs to be stopped manually. Scrapy docs suggest to stop the reactor in the spider_closed signal handler:

Note that you will also have to shutdown the Twisted reactor yourself after the spider is finished. This can be achieved by connecting a handler to the signals.spider_closed signal.

def callback(spider, reason):
    stats = spider.crawler.stats.get_stats()
    # stats here is a dictionary of crawling stats that you usually see on the console        

    # here we need to stop the reactor
    reactor.stop()

crawler.signals.connect(callback, signal=signals.spider_closed)
  • configure and start crawler instance with a spider passed in:

    crawler.configure()
    crawler.crawl(spider)
    crawler.start()
    
  • optionally start logging:

    log.start()
    
  • start the reactor - this would block the script execution:

    reactor.run()
    

Here is an example self-contained script that is using DmozSpider spider and involves item loaders with input and output processors and item pipelines:

import json

from scrapy.crawler import Crawler
from scrapy.contrib.loader import ItemLoader
from scrapy.contrib.loader.processor import Join, MapCompose, TakeFirst
from scrapy import log, signals, Spider, Item, Field
from scrapy.settings import Settings
from twisted.internet import reactor


# define an item class
class DmozItem(Item):
    title = Field()
    link = Field()
    desc = Field()


# define an item loader with input and output processors
class DmozItemLoader(ItemLoader):
    default_input_processor = MapCompose(unicode.strip)
    default_output_processor = TakeFirst()

    desc_out = Join()


# define a pipeline
class JsonWriterPipeline(object):
    def __init__(self):
        self.file = open('items.jl', 'wb')

    def process_item(self, item, spider):
        line = json.dumps(dict(item)) + "\n"
        self.file.write(line)
        return item


# define a spider
class DmozSpider(Spider):
    name = "dmoz"
    allowed_domains = ["dmoz.org"]
    start_urls = [
        "http://www.dmoz.org/Computers/Programming/Languages/Python/Books/",
        "http://www.dmoz.org/Computers/Programming/Languages/Python/Resources/"
    ]

    def parse(self, response):
        for sel in response.xpath('//ul/li'):
            loader = DmozItemLoader(DmozItem(), selector=sel, response=response)
            loader.add_xpath('title', 'a/text()')
            loader.add_xpath('link', 'a/@href')
            loader.add_xpath('desc', 'text()')
            yield loader.load_item()


# callback fired when the spider is closed
def callback(spider, reason):
    stats = spider.crawler.stats.get_stats()  # collect/log stats?

    # stop the reactor
    reactor.stop()


# instantiate settings and provide a custom configuration
settings = Settings()
settings.set('ITEM_PIPELINES', {
    '__main__.JsonWriterPipeline': 100
})

# instantiate a crawler passing in settings
crawler = Crawler(settings)

# instantiate a spider
spider = DmozSpider()

# configure signals
crawler.signals.connect(callback, signal=signals.spider_closed)

# configure and start the crawler
crawler.configure()
crawler.crawl(spider)
crawler.start()

# start logging
log.start()

# start the reactor (blocks execution)
reactor.run()

Run it in a usual way:

python runner.py

and observe items exported to items.jl with the help of the pipeline:

{"desc": "", "link": "/", "title": "Top"}
{"link": "/Computers/", "title": "Computers"}
{"link": "/Computers/Programming/", "title": "Programming"}
{"link": "/Computers/Programming/Languages/", "title": "Languages"}
{"link": "/Computers/Programming/Languages/Python/", "title": "Python"}
...

Gist is available here (feel free to improve):


Notes:

If you define settings by instantiating a Settings() object - you'll get all the defaults Scrapy settings. But, if you want to, for example, configure an existing pipeline, or configure a DEPTH_LIMIT or tweak any other setting, you need to either set it in the script via settings.set() (as demonstrated in the example):

pipelines = {
    'mypackage.pipelines.FilterPipeline': 100,
    'mypackage.pipelines.MySQLPipeline': 200
}
settings.set('ITEM_PIPELINES', pipelines, priority='cmdline')

or, use an existing settings.py with all the custom settings preconfigured:

from scrapy.utils.project import get_project_settings

settings = get_project_settings()

Other useful links on the subject: