Libearth Changelog

Version 0.4.0

To be released.

• Python 3.2 is no more supported since even pip 8.0.0 also dropped their support for Python 3.2.
• Parsing RSS 1.0 feed is available. [issue #57]
• Refactoring parser package. [issue #54]
  • Every single element parser can be specified using ParserBase and its decorator. When calling root element parser, the children elements are also parsed in hierarchical order.
  • Basic parsing information is stored in SessionBase and passed from parent parser to chlidren parsers.
  • Added get_element_id(). It returns the string consists of an XML namespace and an element tag that xml.etree.ElementTree can recognize when finding children elements.
  • Support atom feed that Text with xhtml type.
• Introduced new libearth.defaults module. This module provides small utilities and default data to fill initial state of Earth Reader apps.
• HTML sanitizer now does rebase all links in the given document on the base uri. The get_sanitized_html() method was added to Text type. The sanitize_html() function became to additionally require base_uri parameter.
• Added get_default_name() for default session name.

Version 0.3.3

Released on November 6, 2014.

• Fixed a bug that complete() never terminates for documents read() from a single chunk.

Version 0.3.2

Released on November 5, 2014.

• Fixed a bug that SubscriptionLists having Outlines without their created_at attribute failed to be merged on Python 3. [issue #65]
• Fixed a bug that a DocumentElement in streamed read mode is not properly marked as complete even when it’s completed by complete() function in some cases.

Version 0.3.1

Released on July 20, 2014.

• Fixed two backward compatibility breakages:
  • A bug that subcategory changes hadn’t been detected when SubscriptionLists are merged.
  • A bug that all children outlines become wiped when a category is deleted.

Version 0.3.0

Released on July 12, 2014.

• Root MergeableDocumentElements’ __merge_entities__() methods are not ignored anymore. Respnosibilty to merge two documents is now moved from Session.merge() method to MergeableDocumentElement.__merge_entities__() method.
• crawl() now return a set of CrawlResult objects instead of tuples.
• feeds parameter of crawl() function was renamed to feed_urls.
• Added feed_uri parameter and corresponding feed_uri attribute to CrawlError exception.
• Timeout option was added to crawler.
  • Added optional timeout parameter to crawl().
  • Added optional timeout parameter to get_feed().
  • Added DEFAULT_TIMEOUT constant which is 10 seconds.
• Added LinkList.favicon property. [issue #49]
• Link.relation attribute which had been optional now becomes required
• AutoDiscovery.find_feed_url() method (that returned feed links) was gone. Instead AutoDiscovery.find() method (that returns a pair of feed links and favicon links) was introduced. [issue #49]
• Subscription.icon_uri attribute was introduced. [issue #49]
• Added an optional icon_uri parameter to SubscriptionSet.subscribe() method. [issue #49]
• Added normalize_xml_encoding() function to workaround xml.etree.ElementTree module’s encoding detection bug. [issue #41]
• Added guess_tzinfo_by_locale() function. [issue #41]
• Added microseconds option to Rfc822 codec.
• Fixed incorrect merge of subscription/category deletion.
  • Subscriptions are now archived rather than deleted.
  • Outline (which is a common superclass of Subscription and Category) now has deleted_at attribute and deleted property.
• Fixed several rss2 parser bugs.
  • Now the parser accepts several malformed <pubDate> and <lastBuildDate> elements.
  • It become to guess the time zone according to its <language> and the ccTLD (if applicable) when the date time doesn’t give any explicit time zone (which is also malformed). [issue #41]
  • It had ignored <category> elements other than the last one, now it become to accept as many as there are.
  • It had ignored <comments> links at all, now these become to be parsed to Link objects with relation='discussion'.
  • Some RSS 2 feeds put a URI into <generator>, so the parser now treat it as uri rather than value for such situation.
  • <enclosure> links had been parsed as Link object without relation attribute, but it becomes to properly set the attribute to 'enclosure'.
  • Mixed <link> elements with Atom namespace also becomes to be parsed well.
• Fixed several atom parser bugs.
  • Now it accepts obsolete PURL Atom namespace.
  • Since some broken Atom feeds (e.g. Naver Blog) provide date time as RFC 822 format which is incorrect according to RFC 4287#section-3.3 (section 3.3), the parser becomes to accept RFC 822 format as well.
  • Some broken Atom feeds (e.g. Naver Blog) use <modified> which is not standard instead of <updated> which is standard, so the parser now treats <modified> equivalent to <updated>.
  • <content> and <summary> can has text/plain and text/html in addition to text and html.
  • <author>/<contributor> becomes ignored if it hasn’t any of <name>, <uri>, or <email>.
  • Fixed a parser bug that hadn’t interpret omission of link[rel] attribute as 'alternate'.
• Fixed the parser to work well even if there’s any file separator characters (FS, '\x1c').

Version 0.2.1

Released on July 12, 2014.

• Fixed rss2 parsing error when any empty element occurs.
• Fixed a bug that validate() function errored when any subelement has Text descriptor.

Version 0.2.0

Released on April 22, 2014.

• Session files in .sessions/ directory become to be touched only once at a transaction. [issue #43]
• Added SubscriptionSet.contains() method which provides recursively=True option. It’s useful for determining that a subcategory or subscription is in the whole tree.
• Attribute.default option becomes to accept only callable objects. Below 0.2.0, default is not a function but a value which is simply used as it is.
• libearth.parser.heuristic module is gone; and get_format() function in the module is moved to libearth.parser.autodiscovery module: get_format().
• Added Link.html property.
• Added LinkList.permalink property.
• Fixed a FileSystemRepository bug that conflicts reading buffer and emits broken mixed bytes when there are simultaneous readings and writings to the same key.
• Fixed broken functions related to repository urls on Windows.
• Fixed libearth.compat.parallel.cpu_count() function not to raise NotImplementedError in some cases.
• Fixed Rfc822 to properly work also on non-English locales e.g. ko_KR.

Version 0.1.2

Released on January 19, 2014.

• XML elements in data files are written in canonical order. For example, <title> element of the feed was at the back before, but now is in front.
• write() becomes to store length hints of children that is multiple, and read() becomes aware of the hints. When hints are read len() for the ElementList is O(1).
• Fixed a bug that autodiscovery raises AttributeError when the given HTML contains <link> to both application/atom+xml and application/rss+xml. [issue #40]
• Fill <title> to <description> if there’s no <title> (rss2).
• Fill <id> to the feed URL if there’s no <id> (atom).

Version 0.1.1

Released on January 2, 2014.

• Added a workaround for thread unsafety time.strftime() on CPython. See http://bugs.python.org/issue7980 as well. [issue #32]
• Fixed UnicodeDecodeError which is raised when a feed title contains any non-ASCII characters. [issue #34 by Jae-Myoung Yu]
• Now libearth.parser.rss2 fills Entry.updated_at if it’s not given. [issue #35]
• Fixed TypeError which is raised when any DocumentElement with multiple Child elements is passed to validate() function.
• Fixed the race condition of two FileSystemRepository objects creating the same directory. [issue #36 by klutzy]
• parallel_map() becomes to raise exceptions at the last, if any errored. [issue #38]

Version 0.1.0

Released on December 13, 2013. Initial alpha version.