hyperlink

(August 3, 2020)

• CPython 3.7 and 3.8 and PyPy3 added to test matrix
• Hyperlink now has type hints and they are now exported per PEP 561.
• Several bugs related to hidden state were fixed, making it so that all data on a URL object (including rooted and uses_netloc) is reflected by and consistent with its textual representation. This does mean that sometimes these constructor arguments are ignored, if it would create invalid or unparseable URL text.

(April 7, 2019)

A query parameter-centric release, with two enhancements:

• "equals sign" characters in query parameter values are no longer escaped. (see #39)
• URL.remove() now accepts value and limit parameters, allowing for removal of specific name-value pairs, as well as limiting the number of removals. (See #71)

(February 25, 2018)

Maybe the biggest release since the original. Big thanks to @glyph, @wsanchez, @Julian, @wbolster, and of course, @markrwilliams, for all the reports, reviews, and other contributions.

• When passed to str() URLs now stringify to usable URL strings. (See #49)
• Switched off of Python's built-in IDNA facilities to using the idna package. Not only is it much more modern, it's also much more strict and correct in its output. (See #19 and #56)
• Added new DecodedURL type with almost-identical API to the normal URL, except that it automatically handles reserved characters in argument values passed to its methods. (See #6, #11, #44)
• Added top-level parse() convenience function that now represents the main entrypoint to hyperlink.
• Accept dictionaries as ‘query=’ arguments, in addition to sequences of tuples (see #50)
• URL.child() will no longer fail when child gets no segments (#42 and #43)
• URL.normalize() now supports encoding stray/unmatched % characters in character-encoded fields (userinfo, path, query string, fragment) (see #61, #62)

(August 19, 2017)

• Add URL.normalize() method, which applies five normalizations from RFC 3986 (sections 2.3, 2.1, 3.2.2, 6.2.2.3, 6.2.3). See the docs for more details.
• Enable URL.click() to accept a URL object as a target.

(July 18, 2017)

Fixed a couple major decoding issues and simplified the URL API.

• limit types accepted by URL.from_text() to just text (str on py3, unicode on py2), see #20
• fix percent decoding issues surrounding multiple calls to URL.to_iri() (see #16)
• remove the socket-inspired family argument from URL's APIs. It was never consistently implemented and leaked slightly more problems than it solved.
• Improve authority parsing (see #26)
• include LICENSE, README, docs, and other resources in the package

(June 18, 2017)

A small bugfix release after yesterday's big changes. This patch version simply reverts an exception message for parameters expecting strings on Python 3, returning to compliance with Twisted's test suite.

(June 17, 2017)

Fixed a great round of issues based on the amazing community review (@wsanchez and @jvanasco) after our first listserv announcement and PyConWeb talk.

• Add checking for invalid unescaped delimiters in parameters to the URL constructor. No more slashes and question marks allowed in path segments themselves.
• More robust support for IDNA decoding on "narrow"/UCS-2 Python builds (e.g., Mac's built-in Python).
• Correctly encode colons in the first segment of relative paths for URLs with no scheme set.
• Make URLs with empty paths compare as equal (http://example.com vs. http://example.com/) per RFC 3986. If you need the stricter check, you can check the attributes directly or compare the strings.
• Automatically escape the arguments to .child() and .sibling()
• Fix some IPv6 and port parsing corner cases.

• Python 2.6 support
• Added LICENSE
• Automated CI and code coverage
• When a host and a query string are present, empty paths are now rendered as a single slash. This is slightly more in line with RFC 3986 section 6.2.3, but might need to go further and use an empty slash whenever the authority is present. This also better replicates Twisted URL's old behavior.