Release notes¶

Support for returning dictionaries in spiders¶

Declaring and returning Scrapy Items is no longer necessary to collect the scraped data from your spider, you can now return explicit dictionaries instead.

Classic version

class MyItem(scrapy.Item):
    url = scrapy.Field()

class MySpider(scrapy.Spider):
    def parse(self, response):
        return MyItem(url=response.url)

New version

class MySpider(scrapy.Spider):
    def parse(self, response):
        return {'url': response.url}

Per-spider settings (GSoC 2014)¶

Last Google Summer of Code project accomplished an important redesign of the mechanism used for populating settings, introducing explicit priorities to override any given setting. As an extension of that goal, we included a new level of priority for settings that act exclusively for a single spider, allowing them to redefine project settings.

Start using it by defining a custom_settings class variable in your spider:

class MySpider(scrapy.Spider):
    custom_settings = {
        "DOWNLOAD_DELAY": 5.0,
        "RETRY_ENABLED": False,
    }

Read more about settings population: Settings

Python Logging¶

Scrapy 1.0 has moved away from Twisted logging to support Python built in’s as default logging system. We’re maintaining backward compatibility for most of the old custom interface to call logging functions, but you’ll get warnings to switch to the Python logging API entirely.

Old version

from scrapy import log
log.msg('MESSAGE', log.INFO)

New version

import logging
logging.info('MESSAGE')

Logging with spiders remains the same, but on top of the log() method you’ll have access to a custom logger created for the spider to issue log events:

class MySpider(scrapy.Spider):
    def parse(self, response):
        self.logger.info('Response received')

Read more in the logging documentation: Logging

Crawler API refactoring (GSoC 2014)¶

Another milestone for last Google Summer of Code was a refactoring of the internal API, seeking a simpler and easier usage. Check new core interface in: Core API

A common situation where you will face these changes is while running Scrapy from scripts. Here’s a quick example of how to run a Spider manually with the new API:

from scrapy.crawler import CrawlerProcess

process = CrawlerProcess({
    'USER_AGENT': 'Mozilla/4.0 (compatible; MSIE 7.0; Windows NT 5.1)'
})
process.crawl(MySpider)
process.start()

Bear in mind this feature is still under development and its API may change until it reaches a stable status.

See more examples for scripts running Scrapy: Common Practices

Module Relocations¶

There’s been a large rearrangement of modules trying to improve the general structure of Scrapy. Main changes were separating various subpackages into new projects and dissolving both scrapy.contrib and scrapy.contrib_exp into top level packages. Backward compatibility was kept among internal relocations, while importing deprecated modules expect warnings indicating their new place.

Changelog¶

New Features and Enhancements

• Python logging (issue 1060, issue 1235, issue 1236, issue 1240, issue 1259, issue 1278, issue 1286)
• FEED_EXPORT_FIELDS option (issue 1159, issue 1224)
• Dns cache size and timeout options (issue 1132)
• support namespace prefix in xmliter_lxml (issue 963)
• Reactor threadpool max size setting (issue 1123)
• Allow spiders to return dicts. (issue 1081)
• Add Response.urljoin() helper (issue 1086)
• look in ~/.config/scrapy.cfg for user config (issue 1098)
• handle TLS SNI (issue 1101)
• Selectorlist extract first (issue 624, issue 1145)
• Added JmesSelect (issue 1016)
• add gzip compression to filesystem http cache backend (issue 1020)
• CSS support in link extractors (issue 983)
• httpcache dont_cache meta #19 #689 (issue 821)
• add signal to be sent when request is dropped by the scheduler (issue 961)
• avoid download large response (issue 946)
• Allow to specify the quotechar in CSVFeedSpider (issue 882)
• Add referer to “Spider error processing” log message (issue 795)
• process robots.txt once (issue 896)
• GSoC Per-spider settings (issue 854)
• Add project name validation (issue 817)
• GSoC API cleanup (issue 816, issue 1128, issue 1147, issue 1148, issue 1156, issue 1185, issue 1187, issue 1258, issue 1268, issue 1276, issue 1285, issue 1284)
• Be more responsive with IO operations (issue 1074 and issue 1075)
• Do leveldb compaction for httpcache on closing (issue 1297)

Deprecations and Removals

• Deprecate htmlparser link extractor (issue 1205)
• remove deprecated code from FeedExporter (issue 1155)
• a leftover for.15 compatibility (issue 925)
• drop support for CONCURRENT_REQUESTS_PER_SPIDER (issue 895)
• Drop old engine code (issue 911)
• Deprecate SgmlLinkExtractor (issue 777)

Relocations

• Move exporters/__init__.py to exporters.py (issue 1242)
• Move base classes to their packages (issue 1218, issue 1233)
• Module relocation (issue 1181, issue 1210)
• rename SpiderManager to SpiderLoader (issue 1166)
• Remove djangoitem (issue 1177)
• remove scrapy deploy command (issue 1102)
• dissolve contrib_exp (issue 1134)
• Deleted bin folder from root, fixes #913 (issue 914)
• Remove jsonrpc based webservice (issue 859)
• Move Test cases under project root dir (issue 827, issue 841)
• Fix backward incompatibility for relocated paths in settings (issue 1267)

Documentation

• CrawlerProcess documentation (issue 1190)
• Favoring web scraping over screen scraping in the descriptions (issue 1188)
• Some improvements for Scrapy tutorial (issue 1180)
• Documenting Files Pipeline together with Images Pipeline (issue 1150)
• deployment docs tweaks (issue 1164)
• Added deployment section covering scrapyd-deploy and shub (issue 1124)
• Adding more settings to project template (issue 1073)
• some improvements to overview page (issue 1106)
• Updated link in docs/topics/architecture.rst (issue 647)
• DOC reorder topics (issue 1022)
• updating list of Request.meta special keys (issue 1071)
• DOC document download_timeout (issue 898)
• DOC simplify extension docs (issue 893)
• Leaks docs (issue 894)
• DOC document from_crawler method for item pipelines (issue 904)
• Spider_error doesn’t support deferreds (issue 1292)
• Corrections & Sphinx related fixes (issue 1220, issue 1219, issue 1196, issue 1172, issue 1171, issue 1169, issue 1160, issue 1154, issue 1127, issue 1112, issue 1105, issue 1041, issue 1082, issue 1033, issue 944, issue 866, issue 864, issue 796, issue 1260, issue 1271, issue 1293, issue 1298)

Bugfixes

• Item multi inheritance fix (issue 353, issue 1228)
• ItemLoader.load_item: iterate over copy of fields (issue 722)
• Fix Unhandled error in Deferred (RobotsTxtMiddleware) (issue 1131, issue 1197)
• Force to read DOWNLOAD_TIMEOUT as int (issue 954)
• scrapy.utils.misc.load_object should print full traceback (issue 902)
• Fix bug for ”.local” host name (issue 878)
• Fix for Enabled extensions, middlewares, pipelines info not printed anymore (issue 879)
• fix dont_merge_cookies bad behaviour when set to false on meta (issue 846)

Python 3 In Progress Support

• disable scrapy.telnet if twisted.conch is not available (issue 1161)
• fix Python 3 syntax errors in ajaxcrawl.py (issue 1162)
• more python3 compatibility changes for urllib (issue 1121)
• assertItemsEqual was renamed to assertCountEqual in Python 3. (issue 1070)
• Import unittest.mock if available. (issue 1066)
• updated deprecated cgi.parse_qsl to use six’s parse_qsl (issue 909)
• Prevent Python 3 port regressions (issue 830)
• PY3: use MutableMapping for python 3 (issue 810)
• PY3: use six.BytesIO and six.moves.cStringIO (issue 803)
• PY3: fix xmlrpclib and email imports (issue 801)
• PY3: use six for robotparser and urlparse (issue 800)
• PY3: use six.iterkeys, six.iteritems, and tempfile (issue 799)
• PY3: fix has_key and use six.moves.configparser (issue 798)
• PY3: use six.moves.cPickle (issue 797)
• PY3 make it possible to run some tests in Python3 (issue 776)

Tests

• remove unnecessary lines from py3-ignores (issue 1243)
• Fix remaining warnings from pytest while collecting tests (issue 1206)
• Add docs build to travis (issue 1234)
• TST don’t collect tests from deprecated modules. (issue 1165)
• install service_identity package in tests to prevent warnings (issue 1168)
• Fix deprecated settings API in tests (issue 1152)
• Add test for webclient with POST method and no body given (issue 1089)
• py3-ignores.txt supports comments (issue 1044)
• modernize some of the asserts (issue 835)
• selector.__repr__ test (issue 779)

Code refactoring

• CSVFeedSpider cleanup: use iterate_spider_output (issue 1079)
• remove unnecessary check from scrapy.utils.spider.iter_spider_output (issue 1078)
• Pydispatch pep8 (issue 992)
• Removed unused ‘load=False’ parameter from walk_modules() (issue 871)
• For consistency, use job_dir helper in SpiderState extension. (issue 805)
• rename “sflo” local variables to less cryptic “log_observer” (issue 775)

Enhancements¶

• Improve Scrapy top-level namespace (issue 494, issue 684)
• Add selector shortcuts to responses (issue 554, issue 690)
• Add new lxml based LinkExtractor to replace unmantained SgmlLinkExtractor (issue 559, issue 761, issue 763)
• Cleanup settings API - part of per-spider settings GSoC project (issue 737)
• Add UTF8 encoding header to templates (issue 688, issue 762)
• Telnet console now binds to 127.0.0.1 by default (issue 699)
• Update debian/ubuntu install instructions (issue 509, issue 549)
• Disable smart strings in lxml XPath evaluations (issue 535)
• Restore filesystem based cache as default for http cache middleware (issue 541, issue 500, issue 571)
• Expose current crawler in Scrapy shell (issue 557)
• Improve testsuite comparing CSV and XML exporters (issue 570)
• New offsite/filtered and offsite/domains stats (issue 566)
• Support process_links as generator in CrawlSpider (issue 555)
• Verbose logging and new stats counters for DupeFilter (issue 553)
• Add a mimetype parameter to MailSender.send() (issue 602)
• Generalize file pipeline log messages (issue 622)
• Replace unencodeable codepoints with html entities in SGMLLinkExtractor (issue 565)
• Converted SEP documents to rst format (issue 629, issue 630, issue 638, issue 632, issue 636, issue 640, issue 635, issue 634, issue 639, issue 637, issue 631, issue 633, issue 641, issue 642)
• Tests and docs for clickdata’s nr index in FormRequest (issue 646, issue 645)
• Allow to disable a downloader handler just like any other component (issue 650)
• Log when a request is discarded after too many redirections (issue 654)
• Log error responses if they are not handled by spider callbacks (issue 612, issue 656)
• Add content-type check to http compression mw (issue 193, issue 660)
• Run pypy tests using latest pypi from ppa (issue 674)
• Run test suite using pytest instead of trial (issue 679)
• Build docs and check for dead links in tox environment (issue 687)
• Make scrapy.version_info a tuple of integers (issue 681, issue 692)
• Infer exporter’s output format from filename extensions (issue 546, issue 659, issue 760)
• Support case-insensitive domains in url_is_from_any_domain() (issue 693)
• Remove pep8 warnings in project and spider templates (issue 698)
• Tests and docs for request_fingerprint function (issue 597)
• Update SEP-19 for GSoC project per-spider settings (issue 705)
• Set exit code to non-zero when contracts fails (issue 727)
• Add a setting to control what class is instanciated as Downloader component (issue 738)
• Pass response in item_dropped signal (issue 724)
• Improve scrapy check contracts command (issue 733, issue 752)
• Document spider.closed() shortcut (issue 719)
• Document request_scheduled signal (issue 746)
• Add a note about reporting security issues (issue 697)
• Add LevelDB http cache storage backend (issue 626, issue 500)
• Sort spider list output of scrapy list command (issue 742)
• Multiple documentation enhancemens and fixes (issue 575, issue 587, issue 590, issue 596, issue 610, issue 617, issue 618, issue 627, issue 613, issue 643, issue 654, issue 675, issue 663, issue 711, issue 714)

Bugfixes¶

• Encode unicode URL value when creating Links in RegexLinkExtractor (issue 561)
• Ignore None values in ItemLoader processors (issue 556)
• Fix link text when there is an inner tag in SGMLLinkExtractor and HtmlParserLinkExtractor (issue 485, issue 574)
• Fix wrong checks on subclassing of deprecated classes (issue 581, issue 584)
• Handle errors caused by inspect.stack() failures (issue 582)
• Fix a reference to unexistent engine attribute (issue 593, issue 594)
• Fix dynamic itemclass example usage of type() (issue 603)
• Use lucasdemarchi/codespell to fix typos (issue 628)
• Fix default value of attrs argument in SgmlLinkExtractor to be tuple (issue 661)
• Fix XXE flaw in sitemap reader (issue 676)
• Fix engine to support filtered start requests (issue 707)
• Fix offsite middleware case on urls with no hostnames (issue 745)
• Testsuite doesn’t require PIL anymore (issue 585)

Enhancements¶

• [Backwards incompatible] Switched HTTPCacheMiddleware backend to filesystem (issue 541) To restore old backend set HTTPCACHE_STORAGE to scrapy.contrib.httpcache.DbmCacheStorage
• Proxy https:// urls using CONNECT method (issue 392, issue 397)
• Add a middleware to crawl ajax crawleable pages as defined by google (issue 343)
• Rename scrapy.spider.BaseSpider to scrapy.spider.Spider (issue 510, issue 519)
• Selectors register EXSLT namespaces by default (issue 472)
• Unify item loaders similar to selectors renaming (issue 461)
• Make RFPDupeFilter class easily subclassable (issue 533)
• Improve test coverage and forthcoming Python 3 support (issue 525)
• Promote startup info on settings and middleware to INFO level (issue 520)
• Support partials in get_func_args util (issue 506, issue:504)
• Allow running indiviual tests via tox (issue 503)
• Update extensions ignored by link extractors (issue 498)
• Add middleware methods to get files/images/thumbs paths (issue 490)
• Improve offsite middleware tests (issue 478)
• Add a way to skip default Referer header set by RefererMiddleware (issue 475)
• Do not send x-gzip in default Accept-Encoding header (issue 469)
• Support defining http error handling using settings (issue 466)
• Use modern python idioms wherever you find legacies (issue 497)
• Improve and correct documentation (issue 527, issue 524, issue 521, issue 517, issue 512, issue 505, issue 502, issue 489, issue 465, issue 460, issue 425, issue 536)

Fixes¶

• Update Selector class imports in CrawlSpider template (issue 484)
• Fix unexistent reference to engine.slots (issue 464)
• Do not try to call body_as_unicode() on a non-TextResponse instance (issue 462)
• Warn when subclassing XPathItemLoader, previously it only warned on instantiation. (issue 523)
• Warn when subclassing XPathSelector, previously it only warned on instantiation. (issue 537)
• Multiple fixes to memory stats (issue 531, issue 530, issue 529)
• Fix overriding url in FormRequest.from_response() (issue 507)
• Fix tests runner under pip 1.5 (issue 513)
• Fix logging error when spider name is unicode (issue 479)

Enhancements¶

• New Selector’s API including CSS selectors (issue 395 and issue 426),
• Request/Response url/body attributes are now immutable (modifying them had been deprecated for a long time)
• ITEM_PIPELINES is now defined as a dict (instead of a list)
• Sitemap spider can fetch alternate URLs (issue 360)
• Selector.remove_namespaces() now remove namespaces from element’s attributes. (issue 416)
• Paved the road for Python 3.3+ (issue 435, issue 436, issue 431, issue 452)
• New item exporter using native python types with nesting support (issue 366)
• Tune HTTP1.1 pool size so it matches concurrency defined by settings (commit b43b5f575)
• scrapy.mail.MailSender now can connect over TLS or upgrade using STARTTLS (issue 327)
• New FilesPipeline with functionality factored out from ImagesPipeline (issue 370, issue 409)
• Recommend Pillow instead of PIL for image handling (issue 317)
• Added debian packages for Ubuntu quantal and raring (commit 86230c0)
• Mock server (used for tests) can listen for HTTPS requests (issue 410)
• Remove multi spider support from multiple core components (issue 422, issue 421, issue 420, issue 419, issue 423, issue 418)
• Travis-CI now tests Scrapy changes against development versions of w3lib and queuelib python packages.
• Add pypy 2.1 to continuous integration tests (commit ecfa7431)
• Pylinted, pep8 and removed old-style exceptions from source (issue 430, issue 432)
• Use importlib for parametric imports (issue 445)
• Handle a regression introduced in Python 2.7.5 that affects XmlItemExporter (issue 372)
• Bugfix crawling shutdown on SIGINT (issue 450)
• Do not submit reset type inputs in FormRequest.from_response (commit b326b87)
• Do not silence download errors when request errback raises an exception (commit 684cfc0)

Bugfixes¶

• Fix tests under Django 1.6 (commit b6bed44c)
• Lot of bugfixes to retry middleware under disconnections using HTTP 1.1 download handler
• Fix inconsistencies among Twisted releases (issue 406)
• Fix scrapy shell bugs (issue 418, issue 407)
• Fix invalid variable name in setup.py (issue 429)
• Fix tutorial references (issue 387)
• Improve request-response docs (issue 391)
• Improve best practices docs (issue 399, issue 400, issue 401, issue 402)
• Improve django integration docs (issue 404)
• Document bindaddress request meta (commit 37c24e01d7)
• Improve Request class documentation (issue 226)

Other¶

• Dropped Python 2.6 support (issue 448)
• Add cssselect python package as install dependency
• Drop libxml2 and multi selector’s backend support, lxml is required from now on.
• Minimum Twisted version increased to 10.0.0, dropped Twisted 8.0 support.
• Running test suite now requires mock python library (issue 390)

Thanks¶

Thanks to everyone who contribute to this release!

List of contributors sorted by number of commits:

69 Daniel Graña <dangra@...>
37 Pablo Hoffman <pablo@...>
13 Mikhail Korobov <kmike84@...>
 9 Alex Cepoi <alex.cepoi@...>
 9 alexanderlukanin13 <alexander.lukanin.13@...>
 8 Rolando Espinoza La fuente <darkrho@...>
 8 Lukasz Biedrycki <lukasz.biedrycki@...>
 6 Nicolas Ramirez <nramirez.uy@...>
 3 Paul Tremberth <paul.tremberth@...>
 2 Martin Olveyra <molveyra@...>
 2 Stefan <misc@...>
 2 Rolando Espinoza <darkrho@...>
 2 Loren Davie <loren@...>
 2 irgmedeiros <irgmedeiros@...>
 1 Stefan Koch <taikano@...>
 1 Stefan <cct@...>
 1 scraperdragon <dragon@...>
 1 Kumara Tharmalingam <ktharmal@...>
 1 Francesco Piccinno <stack.box@...>
 1 Marcos Campal <duendex@...>
 1 Dragon Dave <dragon@...>
 1 Capi Etheriel <barraponto@...>
 1 cacovsky <amarquesferraz@...>
 1 Berend Iwema <berend@...>

New features and settings¶

• Support for AJAX crawleable urls

• New persistent scheduler that stores requests on disk, allowing to suspend and resume crawls (r2737)

• added -o option to scrapy crawl, a shortcut for dumping scraped items into a file (or standard output using -)

• Added support for passing custom settings to Scrapyd schedule.json api (r2779, r2783)

• New ChunkedTransferMiddleware (enabled by default) to support chunked transfer encoding (r2769)

• Add boto 2.0 support for S3 downloader handler (r2763)

• Added marshal to formats supported by feed exports (r2744)

• In request errbacks, offending requests are now received in failure.request attribute (r2738)

• Big downloader refactoring to support per domain/ip concurrency limits (r2732)

  • CONCURRENT_REQUESTS_PER_SPIDER setting has been deprecated and replaced by:

    • CONCURRENT_REQUESTS, CONCURRENT_REQUESTS_PER_DOMAIN, CONCURRENT_REQUESTS_PER_IP

  • check the documentation for more details

• Added builtin caching DNS resolver (r2728)

• Moved Amazon AWS-related components/extensions (SQS spider queue, SimpleDB stats collector) to a separate project: [scaws](https://github.com/scrapinghub/scaws) (r2706, r2714)

• Moved spider queues to scrapyd: scrapy.spiderqueue -> scrapyd.spiderqueue (r2708)

• Moved sqlite utils to scrapyd: scrapy.utils.sqlite -> scrapyd.sqlite (r2781)

• Real support for returning iterators on start_requests() method. The iterator is now consumed during the crawl when the spider is getting idle (r2704)

• Added REDIRECT_ENABLED setting to quickly enable/disable the redirect middleware (r2697)

• Added RETRY_ENABLED setting to quickly enable/disable the retry middleware (r2694)

• Added CloseSpider exception to manually close spiders (r2691)

• Improved encoding detection by adding support for HTML5 meta charset declaration (r2690)

• Refactored close spider behavior to wait for all downloads to finish and be processed by spiders, before closing the spider (r2688)

• Added SitemapSpider (see documentation in Spiders page) (r2658)

• Added LogStats extension for periodically logging basic stats (like crawled pages and scraped items) (r2657)

• Make handling of gzipped responses more robust (#319, r2643). Now Scrapy will try and decompress as much as possible from a gzipped response, instead of failing with an IOError.

• Simplified !MemoryDebugger extension to use stats for dumping memory debugging info (r2639)

• Added new command to edit spiders: scrapy edit (r2636) and -e flag to genspider command that uses it (r2653)

• Changed default representation of items to pretty-printed dicts. (r2631). This improves default logging by making log more readable in the default case, for both Scraped and Dropped lines.

• Added spider_error signal (r2628)

• Added COOKIES_ENABLED setting (r2625)

• Stats are now dumped to Scrapy log (default value of STATS_DUMP setting has been changed to True). This is to make Scrapy users more aware of Scrapy stats and the data that is collected there.

• Added support for dynamically adjusting download delay and maximum concurrent requests (r2599)

• Added new DBM HTTP cache storage backend (r2576)

• Added listjobs.json API to Scrapyd (r2571)

• CsvItemExporter: added join_multivalued parameter (r2578)

• Added namespace support to xmliter_lxml (r2552)

• Improved cookies middleware by making COOKIES_DEBUG nicer and documenting it (r2579)

• Several improvements to Scrapyd and Link extractors

Code rearranged and removed¶

• Merged item passed and item scraped concepts, as they have often proved confusing in the past. This means: (r2630)

  • original item_scraped signal was removed
  • original item_passed signal was renamed to item_scraped
  • old log lines Scraped Item... were removed
  • old log lines Passed Item... were renamed to Scraped Item... lines and downgraded to DEBUG level

• Reduced Scrapy codebase by striping part of Scrapy code into two new libraries:

  • w3lib (several functions from scrapy.utils.{http,markup,multipart,response,url}, done in r2584)
  • scrapely (was scrapy.contrib.ibl, done in r2586)

• Removed unused function: scrapy.utils.request.request_info() (r2577)

• Removed googledir project from examples/googledir. There’s now a new example project called dirbot available on github: https://github.com/scrapy/dirbot

• Removed support for default field values in Scrapy items (r2616)

• Removed experimental crawlspider v2 (r2632)

• Removed scheduler middleware to simplify architecture. Duplicates filter is now done in the scheduler itself, using the same dupe fltering class as before (DUPEFILTER_CLASS setting) (r2640)

• Removed support for passing urls to scrapy crawl command (use scrapy parse instead) (r2704)

• Removed deprecated Execution Queue (r2704)

• Removed (undocumented) spider context extension (from scrapy.contrib.spidercontext) (r2780)

• removed CONCURRENT_SPIDERS setting (use scrapyd maxproc instead) (r2789)

• Renamed attributes of core components: downloader.sites -> downloader.slots, scraper.sites -> scraper.slots (r2717, r2718)

• Renamed setting CLOSESPIDER_ITEMPASSED to CLOSESPIDER_ITEMCOUNT (r2655). Backwards compatibility kept.

New features and improvements¶

• Passed item is now sent in the item argument of the item_passed (#273)
• Added verbose option to scrapy version command, useful for bug reports (#298)
• HTTP cache now stored by default in the project data dir (#279)
• Added project data storage directory (#276, #277)
• Documented file structure of Scrapy projects (see command-line tool doc)
• New lxml backend for XPath selectors (#147)
• Per-spider settings (#245)
• Support exit codes to signal errors in Scrapy commands (#248)
• Added -c argument to scrapy shell command
• Made libxml2 optional (#260)
• New deploy command (#261)
• Added CLOSESPIDER_PAGECOUNT setting (#253)
• Added CLOSESPIDER_ERRORCOUNT setting (#254)

Scrapyd changes¶

• Scrapyd now uses one process per spider
• It stores one log file per spider run, and rotate them keeping the lastest 5 logs per spider (by default)
• A minimal web ui was added, available at http://localhost:6800 by default
• There is now a scrapy server command to start a Scrapyd server of the current project

Changes to settings¶

• added HTTPCACHE_ENABLED setting (False by default) to enable HTTP cache middleware
• changed HTTPCACHE_EXPIRATION_SECS semantics: now zero means “never expire”.

Deprecated/obsoleted functionality¶

• Deprecated runserver command in favor of server command which starts a Scrapyd server. See also: Scrapyd changes
• Deprecated queue command in favor of using Scrapyd schedule.json API. See also: Scrapyd changes
• Removed the !LxmlItemLoader (experimental contrib which never graduated to main contrib)

New features and improvements¶

• New Scrapy service called scrapyd for deploying Scrapy crawlers in production (#218) (documentation available)
• Simplified Images pipeline usage which doesn’t require subclassing your own images pipeline now (#217)
• Scrapy shell now shows the Scrapy log by default (#206)
• Refactored execution queue in a common base code and pluggable backends called “spider queues” (#220)
• New persistent spider queue (based on SQLite) (#198), available by default, which allows to start Scrapy in server mode and then schedule spiders to run.
• Added documentation for Scrapy command-line tool and all its available sub-commands. (documentation available)
• Feed exporters with pluggable backends (#197) (documentation available)
• Deferred signals (#193)
• Added two new methods to item pipeline open_spider(), close_spider() with deferred support (#195)
• Support for overriding default request headers per spider (#181)
• Replaced default Spider Manager with one with similar functionality but not depending on Twisted Plugins (#186)
• Splitted Debian package into two packages - the library and the service (#187)
• Scrapy log refactoring (#188)
• New extension for keeping persistent spider contexts among different runs (#203)
• Added dont_redirect request.meta key for avoiding redirects (#233)
• Added dont_retry request.meta key for avoiding retries (#234)

Command-line tool changes¶

• New scrapy command which replaces the old scrapy-ctl.py (#199) - there is only one global scrapy command now, instead of one scrapy-ctl.py per project - Added scrapy.bat script for running more conveniently from Windows
• Added bash completion to command-line tool (#210)
• Renamed command start to runserver (#209)

API changes¶

• url and body attributes of Request objects are now read-only (#230)

• Request.copy() and Request.replace() now also copies their callback and errback attributes (#231)

• Removed UrlFilterMiddleware from scrapy.contrib (already disabled by default)

• Offsite middelware doesn’t filter out any request coming from a spider that doesn’t have a allowed_domains attribute (#225)

• Removed Spider Manager load() method. Now spiders are loaded in the constructor itself.

• Changes to Scrapy Manager (now called “Crawler”):

  • scrapy.core.manager.ScrapyManager class renamed to scrapy.crawler.Crawler
  • scrapy.core.manager.scrapymanager singleton moved to scrapy.project.crawler

• Moved module: scrapy.contrib.spidermanager to scrapy.spidermanager

• Spider Manager singleton moved from scrapy.spider.spiders to the spiders` attribute of ``scrapy.project.crawler singleton.

• moved Stats Collector classes: (#204)

  • scrapy.stats.collector.StatsCollector to scrapy.statscol.StatsCollector
  • scrapy.stats.collector.SimpledbStatsCollector to scrapy.contrib.statscol.SimpledbStatsCollector

• default per-command settings are now specified in the default_settings attribute of command object class (#201)

• changed arguments of Item pipeline process_item() method from (spider, item) to (item, spider)

  • backwards compatibility kept (with deprecation warning)

• moved scrapy.core.signals module to scrapy.signals

  • backwards compatibility kept (with deprecation warning)

• moved scrapy.core.exceptions module to scrapy.exceptions

  • backwards compatibility kept (with deprecation warning)

• added handles_request() class method to BaseSpider

• dropped scrapy.log.exc() function (use scrapy.log.err() instead)

• dropped component argument of scrapy.log.msg() function

• dropped scrapy.log.log_level attribute

• Added from_settings() class methods to Spider Manager, and Item Pipeline Manager

Changes to settings¶

• Added HTTPCACHE_IGNORE_SCHEMES setting to ignore certain schemes on !HttpCacheMiddleware (#225)
• Added SPIDER_QUEUE_CLASS setting which defines the spider queue to use (#220)
• Added KEEP_ALIVE setting (#220)
• Removed SERVICE_QUEUE setting (#220)
• Removed COMMANDS_SETTINGS_MODULE setting (#201)
• Renamed REQUEST_HANDLERS to DOWNLOAD_HANDLERS and make download handlers classes (instead of functions)

New features and improvements¶

• Added SMTP-AUTH support to scrapy.mail
• New settings added: MAIL_USER, MAIL_PASS (r2065 | #149)
• Added new scrapy-ctl view command - To view URL in the browser, as seen by Scrapy (r2039)
• Added web service for controlling Scrapy process (this also deprecates the web console. (r2053 | #167)
• Support for running Scrapy as a service, for production systems (r1988, r2054, r2055, r2056, r2057 | #168)
• Added wrapper induction library (documentation only available in source code for now). (r2011)
• Simplified and improved response encoding support (r1961, r1969)
• Added LOG_ENCODING setting (r1956, documentation available)
• Added RANDOMIZE_DOWNLOAD_DELAY setting (enabled by default) (r1923, doc available)
• MailSender is no longer IO-blocking (r1955 | #146)
• Linkextractors and new Crawlspider now handle relative base tag urls (r1960 | #148)
• Several improvements to Item Loaders and processors (r2022, r2023, r2024, r2025, r2026, r2027, r2028, r2029, r2030)
• Added support for adding variables to telnet console (r2047 | #165)
• Support for requests without callbacks (r2050 | #166)

API changes¶

• Change Spider.domain_name to Spider.name (SEP-012, r1975)
• Response.encoding is now the detected encoding (r1961)
• HttpErrorMiddleware now returns None or raises an exception (r2006 | #157)
• scrapy.command modules relocation (r2035, r2036, r2037)
• Added ExecutionQueue for feeding spiders to scrape (r2034)
• Removed ExecutionEngine singleton (r2039)
• Ported S3ImagesStore (images pipeline) to use boto and threads (r2033)
• Moved module: scrapy.management.telnet to scrapy.telnet (r2047)

Changes to default settings¶

• Changed default SCHEDULER_ORDER to DFO (r1939)

New features¶

• Added DEFAULT_RESPONSE_ENCODING setting (r1809)
• Added dont_click argument to FormRequest.from_response() method (r1813, r1816)
• Added clickdata argument to FormRequest.from_response() method (r1802, r1803)
• Added support for HTTP proxies (HttpProxyMiddleware) (r1781, r1785)
• Offiste spider middleware now logs messages when filtering out requests (r1841)

Backwards-incompatible changes¶

• Changed scrapy.utils.response.get_meta_refresh() signature (r1804)

• Removed deprecated scrapy.item.ScrapedItem class - use scrapy.item.Item instead (r1838)

• Removed deprecated scrapy.xpath module - use scrapy.selector instead. (r1836)

• Removed deprecated core.signals.domain_open signal - use core.signals.domain_opened instead (r1822)

• log.msg() now receives a spider argument (r1822)

  • Old domain argument has been deprecated and will be removed in 0.9. For spiders, you should always use the spider argument and pass spider references. If you really want to pass a string, use the component argument instead.

• Changed core signals domain_opened, domain_closed, domain_idle

• Changed Item pipeline to use spiders instead of domains

  • The domain argument of process_item() item pipeline method was changed to spider, the new signature is: process_item(spider, item) (r1827 | #105)
  • To quickly port your code (to work with Scrapy 0.8) just use spider.domain_name where you previously used domain.

• Changed Stats API to use spiders instead of domains (r1849 | #113)

  • StatsCollector was changed to receive spider references (instead of domains) in its methods (set_value, inc_value, etc).
  • added StatsCollector.iter_spider_stats() method
  • removed StatsCollector.list_domains() method
  • Also, Stats signals were renamed and now pass around spider references (instead of domains). Here’s a summary of the changes:
  • To quickly port your code (to work with Scrapy 0.8) just use spider.domain_name where you previously used domain. spider_stats contains exactly the same data as domain_stats.

• CloseDomain extension moved to scrapy.contrib.closespider.CloseSpider (r1833)

  • Its settings were also renamed:

    • CLOSEDOMAIN_TIMEOUT to CLOSESPIDER_TIMEOUT
    • CLOSEDOMAIN_ITEMCOUNT to CLOSESPIDER_ITEMCOUNT

• Removed deprecated SCRAPYSETTINGS_MODULE environment variable - use SCRAPY_SETTINGS_MODULE instead (r1840)

• Renamed setting: REQUESTS_PER_DOMAIN to CONCURRENT_REQUESTS_PER_SPIDER (r1830, r1844)

• Renamed setting: CONCURRENT_DOMAINS to CONCURRENT_SPIDERS (r1830)

• Refactored HTTP Cache middleware

• HTTP Cache middleware has been heavilty refactored, retaining the same functionality except for the domain sectorization which was removed. (r1843 )

• Renamed exception: DontCloseDomain to DontCloseSpider (r1859 | #120)

• Renamed extension: DelayedCloseDomain to SpiderCloseDelay (r1861 | #121)

• Removed obsolete scrapy.utils.markup.remove_escape_chars function - use scrapy.utils.markup.replace_escape_chars instead (r1865)