scrapy-test 0.5

scrapy output testing framework

scrapy-test

Scrapy test is a validation/test framework for validating scrapy results. This framework is capable of testing scrapy crawl and stats output.

See example project for hackernews crawler with full test suite.

Philosophy and Architecture

scrapy-test tries to replicate scrapy.Item definition but instead of defining fields it defines test for every field.
Tests are callables that either returns a failure message if some condition is met. Example item specification:

class MyItem(Item):
    name = Field()
    url = Field()

class TestMyItem(ItemSpec):
    item_cls = MyItem

    # define tests
    name_test = Match('some-regex-pattern')
    url_test = lamda v: 'bad url' if 'cat' in v else ''

    # define coverage
    url_cov = 100  # 100% - every item should have url field

scrapy-test also supports stats output validation. When scrapy finished crawling it outputs various stats like error count etc. StatSpec can be defined to validate these stats:

class MyStats(StatsSpec):
    spider_cls = MySpder1, MySpider2
    # or multiple spiders
    validation = {  #stat_name_pattern : tests
        'item_scraped_count': MoreThan(1),
        'downloader/response_status_count/50\d': LessThan(1),
    }
    # required stat keys
    required = ['stat_pattern.+']

Finally scrapy-test determines failure by asserting if there are any messages generated by either stat ir item specifications (exit code 1 and 0 respectively).

Usage

Setup

1. test.py module should be created in spider directory.
   For example creating test.py

   scrapy-test-example/
   ├── example
   │   ├── __init__.py
   │   └── test.py
   └── scrapy.cfg
   

2. Add test file config to scrapy.cfg:

   [settings]
   default = example.settings
   [test]
   root = example.test 
   

3. Define ItemSpec for item field validation:

   from scrapytest.tests import Match, Equal, Type, MoreThan, Map, Len, Required
   
   class TestPost(ItemSpec):
       # defining item that is being covered
       item_cls = PostItem
   
       # defining field tests
       title_test = Match('.{5,}')
       points_test = Type(int), MoreThan(0)
       author_test = Type(str), Match('.{3}')
       comments_test = Type(list), Required()
   
       # also supports methods!
       def url_test(self, value: str):
           if not value.startswith('http'):
               return f'Invalid url: {value}'
           return ''
   

   ItemSpec class should contain attributes that end in _test these attributes have be callables (functions, methods etc.) that return message(s) if failure is encountered. See the url_test example above.

4. Define StatSpec for crawl stats validation:

   class TestStats(StatsSpec):
       # stat pattern: test functions
       validate = {  # this is default
           'log_count/ERROR$': LessThan(1),
           'item_scraped_count': MoreThan(1),
           'finish_reason': Match('finished'),
       }
       # these stats shoudl be required
       required = ['some_cool_stat']  
   

   StatsSpec should contain validate attribute with pattern: tests dictionary.

5. Define Spider classes:

   from project.spiders import HackernewsSpider 
   
   class TestHackernewsSpider(HackernewsSpider):
       test_urls = [
           "https://news.ycombinator.com/item?id=19187417",
       ]
   
       def start_requests(self):
           for url in self.test_urls:
               yield Request(url, self.parse_submission)
   

   This spider should extend your production spider that simply crawls the urls without doing discovery. Alternatively you can also not extend anything for live testing.

Running

$ scrapy-test --help                                                                                                 
Usage: scrapy-test [OPTIONS] [SPIDER_NAME]

  run scrapy-test tests and output messages and appropriate exit code (1 for
  failed, 0 for passed)

Options:
  --cache  enable HTTPCACHE_ENABLED setting for this run
  --help       Show this message and exit.

To run the tests use cli command:

$ scrapy-test <spider_name>

Spider name can be skipped for running all spiders

Notifications

scrapy-test supports notification hooks on either test failure or success:

  --notify-on-error TEXT    send notification on failure, choice from:
                            ['slack']
  --notify-on-all TEXT      send notification on failure or success, choice
                            from: ['slack']
  --notify-on-success TEXT  send notification on success, choice from:
                            ['slack']

Right scrapy-test offers these notifiers:

* Slack - to configure slack notification follow slack [incoming webhooks](https://slack.com/apps/A0F7XDUAZ-incoming-webhooks) app and supply these settings in `scrapy.cfg`:

    slack_url = https://hooks.slack.com/services/AAA/BBB/CCC
    # where the message goes to
    slack_channel = #cats
    # bot's name
    slack_username = bender
    # bot's avatar
    slack_icon_emoji = :bender:
    # maintainer will be mentioned on error
    slack_maintainer = @bernard