= 4.0.0b7 () =

* Upon decoding to string, any characters that can't be represented in
  your chosen encoding will be converted into numeric XML entity
  references.

* Issue a warning if characters were replaced with REPLACEMENT
  CHARACTER during Unicode conversion.

* Restored compatibility with Python 2.6.

= 4.0.0b6 (20110216) =

* Multi-valued attributes like "class" always have a list of values,
  even if there's only one value in the list.

* Added a number of multi-valued attributes defined in HTML5.

* Stopped generating a space before the slash that closes an
  empty-element tag. This may come back if I add a special XHTML mode
  (http://www.w3.org/TR/xhtml1/#C_2), but right now it's pretty
  useless.

* Passing text along with tag-specific arguments to a find* method:

   find("a", text="Click here")

  will find tags that contain the given text as their
  .string. Previously, the tag-specific arguments were ignored and
  only strings were searched.

* Fixed a bug that caused the html5lib tree builder to build a
  partially disconnected tree. Generally cleaned up the html5lib tree
  builder.

* If you restrict a multi-valued attribute like "class" to a string
  that contains spaces, Beautiful Soup will only consider it a match
  if the values correspond to that specific string.

= 4.0.0b5 (20120209) =

* Rationalized Beautiful Soup's treatment of CSS class. A tag
  belonging to multiple CSS classes is treated as having a list of
  values for the 'class' attribute. Searching for a CSS class will
  match *any* of the CSS classes.

  This actually affects all attributes that the HTML standard defines
  as taking multiple values (class, rel, rev, archive, accept-charset,
  and headers), but 'class' is by far the most common. [bug=41034]

* If you pass anything other than a dictionary as the second argument
  to one of the find* methods, it'll assume you want to use that
  object to search against a tag's CSS classes. Previously this only
  worked if you passed in a string.

* Fixed a bug that caused a crash when you passed a dictionary as an
  attribute value (possibly because you mistyped "attrs"). [bug=842419]

* Unicode, Dammit now detects the encoding in HTML 5-style <meta> tags
  like <meta charset="utf-8" />. [bug=837268]

* If Unicode, Dammit can't figure out a consistent encoding for a
  page, it will try each of its guesses again, with errors="replace"
  instead of errors="strict". This may mean that some data gets
  replaced with REPLACEMENT CHARACTER, but at least most of it will
  get turned into Unicode. [bug=754903]

* Patched over a bug in html5lib (?) that was crashing Beautiful Soup
  on certain kinds of markup. [bug=838800]

* Fixed a bug that wrecked the tree if you replaced an element with an
  empty string. [bug=728697]

* Improved Unicode, Dammit's behavior when you give it Unicode to
  begin with.

= 4.0.0b4 (20120208) =

* Added BeautifulSoup.new_string() to go along with BeautifulSoup.new_tag()

* BeautifulSoup.new_tag() will follow the rules of whatever
  tree-builder was used to create the original BeautifulSoup object. A
  new <p> tag will look like "<p />" if the soup object was created to
  parse XML, but it will look like "<p></p>" if the soup object was
  created to parse HTML.

* We pass in strict=False to html.parser on Python 3, greatly
  improving html.parser's ability to handle bad HTML.

* We also monkeypatch a serious bug in html.parser that made
  strict=False disastrous on Python 3.2.2.

* Replaced the "substitute_html_entities" argument with the
  more general "formatter" argument.

* Bare ampersands and angle brackets are always converted to XML
  entities unless the user prevents it.

* Added PageElement.insert_before() and PageElement.insert_after(),
  which let you put an element into the parse tree with respect to
  some other element.

* Raise an exception when the user tries to do something nonsensical
  like insert a tag into itself.


= 4.0.0b3 (20120203) =

Beautiful Soup 4 is a nearly-complete rewrite that removes Beautiful
Soup's custom HTML parser in favor of a system that lets you write a
little glue code and plug in any HTML or XML parser you want.

Beautiful Soup 4.0 comes with glue code for four parsers:

 * Python's standard HTMLParser (html.parser in Python 3)
 * lxml's HTML and XML parsers
 * html5lib's HTML parser

HTMLParser is the default, but I recommend you install lxml if you
can.

For complete documentation, see the Sphinx documentation in
bs4/doc/source/. What follows is a summary of the changes from
Beautiful Soup 3.

=== The module name has changed ===

Previously you imported the BeautifulSoup class from a module also
called BeautifulSoup. To save keystrokes and make it clear which
version of the API is in use, the module is now called 'bs4':

    >>> from bs4 import BeautifulSoup

=== It works with Python 3 ===

Beautiful Soup 3.1.0 worked with Python 3, but the parser it used was
so bad that it barely worked at all. Beautiful Soup 4 works with
Python 3, and since its parser is pluggable, you don't sacrifice
quality.

Special thanks to Thomas Kluyver and Ezio Melotti for getting Python 3
support to the finish line. Ezio Melotti is also to thank for greatly
improving the HTML parser that comes with Python 3.2.

=== CDATA sections are normal text, if they're understood at all. ===

Currently, the lxml and html5lib HTML parsers ignore CDATA sections in
markup:

 <p><![CDATA[foo]]></p> => <p></p>

A future version of html5lib will turn CDATA sections into text nodes,
but only within tags like <svg> and <math>:

 <svg><![CDATA[foo]]></svg> => <p>foo</p>

The default XML parser (which uses lxml behind the scenes) turns CDATA
sections into ordinary text elements:

 <p><![CDATA[foo]]></p> => <p>foo</p>

In theory it's possible to preserve the CDATA sections when using the
XML parser, but I don't see how to get it to work in practice.

=== Miscellaneous other stuff ===

If the BeautifulSoup instance has .is_xml set to True, an appropriate
XML declaration will be emitted when the tree is transformed into a
string:

    <?xml version="1.0" encoding="utf-8">
    <markup>
     ...
    </markup>

The ['lxml', 'xml'] tree builder sets .is_xml to True; the other tree
builders set it to False. If you want to parse XHTML with an HTML
parser, you can set it manually.


= 3.2.0 =

The 3.1 series wasn't very useful, so I renamed the 3.0 series to 3.2
to make it obvious which one you should use.

= 3.1.0 =

A hybrid version that supports 2.4 and can be automatically converted
to run under Python 3.0. There are three backwards-incompatible
changes you should be aware of, but no new features or deliberate
behavior changes.

1. str() may no longer do what you want. This is because the meaning
of str() inverts between Python 2 and 3; in Python 2 it gives you a
byte string, in Python 3 it gives you a Unicode string.

The effect of this is that you can't pass an encoding to .__str__
anymore. Use encode() to get a string and decode() to get Unicode, and
you'll be ready (well, readier) for Python 3.

2. Beautiful Soup is now based on HTMLParser rather than SGMLParser,
which is gone in Python 3. There's some bad HTML that SGMLParser
handled but HTMLParser doesn't, usually to do with attribute values
that aren't closed or have brackets inside them:

  <a href="foo</a>, </a><a href="bar">baz</a>
  <a b="<a>">', '<a b="&lt;a&gt;"></a><a>"></a>

A later version of Beautiful Soup will allow you to plug in different
parsers to make tradeoffs between speed and the ability to handle bad
HTML.

3. In Python 3 (but not Python 2), HTMLParser converts entities within
attributes to the corresponding Unicode characters. In Python 2 it's
possible to parse this string and leave the &eacute; intact.

 <a href="http://crummy.com?sacr&eacute;&bleu">

In Python 3, the &eacute; is always converted to \xe9 during
parsing.


= 3.0.7a =

Added an import that makes BS work in Python 2.3.


= 3.0.7 =

Fixed a UnicodeDecodeError when unpickling documents that contain
non-ASCII characters.

Fixed a TypeError that occured in some circumstances when a tag
contained no text.

Jump through hoops to avoid the use of chardet, which can be extremely
slow in some circumstances. UTF-8 documents should never trigger the
use of chardet.

Whitespace is preserved inside <pre> and <textarea> tags that contain
nothing but whitespace.

Beautiful Soup can now parse a doctype that's scoped to an XML namespace.


= 3.0.6 =

Got rid of a very old debug line that prevented chardet from working.

Added a Tag.decompose() method that completely disconnects a tree or a
subset of a tree, breaking it up into bite-sized pieces that are
easy for the garbage collecter to collect.

Tag.extract() now returns the tag that was extracted.

Tag.findNext() now does something with the keyword arguments you pass
it instead of dropping them on the floor.

Fixed a Unicode conversion bug.

Fixed a bug that garbled some <meta> tags when rewriting them.


= 3.0.5 =

Soup objects can now be pickled, and copied with copy.deepcopy.

Tag.append now works properly on existing BS objects. (It wasn't
originally intended for outside use, but it can be now.) (Giles
Radford)

Passing in a nonexistent encoding will no longer crash the parser on
Python 2.4 (John Nagle).

Fixed an underlying bug in SGMLParser that thinks ASCII has 255
characters instead of 127 (John Nagle).

Entities are converted more consistently to Unicode characters.

Entity references in attribute values are now converted to Unicode
characters when appropriate. Numeric entities are always converted,
because SGMLParser always converts them outside of attribute values.

ALL_ENTITIES happens to just be the XHTML entities, so I renamed it to
XHTML_ENTITIES.

The regular expression for bare ampersands was too loose. In some
cases ampersands were not being escaped. (Sam Ruby?)

Non-breaking spaces and other special Unicode space characters are no
longer folded to ASCII spaces. (Robert Leftwich)

Information inside a TEXTAREA tag is now parsed literally, not as HTML
tags. TEXTAREA now works exactly the same way as SCRIPT. (Zephyr Fang)

= 3.0.4 =

Fixed a bug that crashed Unicode conversion in some cases.

Fixed a bug that prevented UnicodeDammit from being used as a
general-purpose data scrubber.

Fixed some unit test failures when running against Python 2.5.

When considering whether to convert smart quotes, UnicodeDammit now
looks at the original encoding in a case-insensitive way.

= 3.0.3 (20060606) =

Beautiful Soup is now usable as a way to clean up invalid XML/HTML (be
sure to pass in an appropriate value for convertEntities, or XML/HTML
entities might stick around that aren't valid in HTML/XML). The result
may not validate, but it should be good enough to not choke a
real-world XML parser. Specifically, the output of a properly
constructed soup object should always be valid as part of an XML
document, but parts may be missing if they were missing in the
original. As always, if the input is valid XML, the output will also
be valid.

= 3.0.2 (20060602) =

Previously, Beautiful Soup correctly handled attribute values that
contained embedded quotes (sometimes by escaping), but not other kinds
of XML character. Now, it correctly handles or escapes all special XML
characters in attribute values.

I aliased methods to the 2.x names (fetch, find, findText, etc.) for
backwards compatibility purposes. Those names are deprecated and if I
ever do a 4.0 I will remove them. I will, I tell you!

Fixed a bug where the findAll method wasn't passing along any keyword
arguments.

When run from the command line, Beautiful Soup now acts as an HTML
pretty-printer, not an XML pretty-printer.

= 3.0.1 (20060530) =

Reintroduced the "fetch by CSS class" shortcut. I thought keyword
arguments would replace it, but they don't. You can't call soup('a',
class='foo') because class is a Python keyword.

If Beautiful Soup encounters a meta tag that declares the encoding,
but a SoupStrainer tells it not to parse that tag, Beautiful Soup will
no longer try to rewrite the meta tag to mention the new
encoding. Basically, this makes SoupStrainers work in real-world
applications instead of crashing the parser.

= 3.0.0 "Who would not give all else for two p" (20060528) =

This release is not backward-compatible with previous releases. If
you've got code written with a previous version of the library, go
ahead and keep using it, unless one of the features mentioned here
really makes your life easier. Since the library is self-contained,
you can include an old copy of the library in your old applications,
and use the new version for everything else.

The documentation has been rewritten and greatly expanded with many
more examples.

Beautiful Soup autodetects the encoding of a document (or uses the one
you specify), and converts it from its native encoding to
Unicode. Internally, it only deals with Unicode strings. When you
print out the document, it converts to UTF-8 (or another encoding you
specify). [Doc reference]

It's now easy to make large-scale changes to the parse tree without
screwing up the navigation members. The methods are extract,
replaceWith, and insert. [Doc reference. See also Improving Memory
Usage with extract]

Passing True in as an attribute value gives you tags that have any
value for that attribute. You don't have to create a regular
expression. Passing None for an attribute value gives you tags that
don't have that attribute at all.

Tag objects now know whether or not they're self-closing. This avoids
the problem where Beautiful Soup thought that tags like <BR /> were
self-closing even in XML documents. You can customize the self-closing
tags for a parser object by passing them in as a list of
selfClosingTags: you don't have to subclass anymore.

There's a new built-in parser, MinimalSoup, which has most of
BeautifulSoup's HTML-specific rules, but no tag nesting rules. [Doc
reference]

You can use a SoupStrainer to tell Beautiful Soup to parse only part
of a document. This saves time and memory, often making Beautiful Soup
about as fast as a custom-built SGMLParser subclass. [Doc reference,
SoupStrainer reference]

You can (usually) use keyword arguments instead of passing a
dictionary of attributes to a search method. That is, you can replace
soup(args={"id" : "5"}) with soup(id="5"). You can still use args if
(for instance) you need to find an attribute whose name clashes with
the name of an argument to findAll. [Doc reference: **kwargs attrs]

The method names have changed to the better method names used in
Rubyful Soup. Instead of find methods and fetch methods, there are
only find methods. Instead of a scheme where you can't remember which
method finds one element and which one finds them all, we have find
and findAll. In general, if the method name mentions All or a plural
noun (eg. findNextSiblings), then it finds many elements
method. Otherwise, it only finds one element. [Doc reference]

Some of the argument names have been renamed for clarity. For instance
avoidParserProblems is now parserMassage.

Beautiful Soup no longer implements a feed method. You need to pass a
string or a filehandle into the soup constructor, not with feed after
the soup has been created. There is still a feed method, but it's the
feed method implemented by SGMLParser and calling it will bypass
Beautiful Soup and cause problems.

The NavigableText class has been renamed to NavigableString. There is
no NavigableUnicodeString anymore, because every string inside a
Beautiful Soup parse tree is a Unicode string.

findText and fetchText are gone. Just pass a text argument into find
or findAll.

Null was more trouble than it was worth, so I got rid of it. Anything
that used to return Null now returns None.

Special XML constructs like comments and CDATA now have their own
NavigableString subclasses, instead of being treated as oddly-formed
data. If you parse a document that contains CDATA and write it back
out, the CDATA will still be there.

When you're parsing a document, you can get Beautiful Soup to convert
XML or HTML entities into the corresponding Unicode characters. [Doc
reference]

= 2.1.1 (20050918) =

Fixed a serious performance bug in BeautifulStoneSoup which was
causing parsing to be incredibly slow.

Corrected several entities that were previously being incorrectly
translated from Microsoft smart-quote-like characters.

Fixed a bug that was breaking text fetch.

Fixed a bug that crashed the parser when text chunks that look like
HTML tag names showed up within a SCRIPT tag.

THEAD, TBODY, and TFOOT tags are now nestable within TABLE
tags. Nested tables should parse more sensibly now.

BASE is now considered a self-closing tag.

= 2.1.0 "Game, or any other dish?" (20050504) =

Added a wide variety of new search methods which, given a starting
point inside the tree, follow a particular navigation member (like
nextSibling) over and over again, looking for Tag and NavigableText
objects that match certain criteria. The new methods are findNext,
fetchNext, findPrevious, fetchPrevious, findNextSibling,
fetchNextSiblings, findPreviousSibling, fetchPreviousSiblings,
findParent, and fetchParents. All of these use the same basic code
used by first and fetch, so you can pass your weird ways of matching
things into these methods.

The fetch method and its derivatives now accept a limit argument.

You can now pass keyword arguments when calling a Tag object as though
it were a method.

Fixed a bug that caused all hand-created tags to share a single set of
attributes.

= 2.0.3 (20050501) =

Fixed Python 2.2 support for iterators.

Fixed a bug that gave the wrong representation to tags within quote
tags like <script>.

Took some code from Mark Pilgrim that treats CDATA declarations as
data instead of ignoring them.

Beautiful Soup's setup.py will now do an install even if the unit
tests fail. It won't build a source distribution if the unit tests
fail, so I can't release a new version unless they pass.

= 2.0.2 (20050416) =

Added the unit tests in a separate module, and packaged it with
distutils.

Fixed a bug that sometimes caused renderContents() to return a Unicode
string even if there was no Unicode in the original string.

Added the done() method, which closes all of the parser's open
tags. It gets called automatically when you pass in some text to the
constructor of a parser class; otherwise you must call it yourself.

Reinstated some backwards compatibility with 1.x versions: referencing
the string member of a NavigableText object returns the NavigableText
object instead of throwing an error.

= 2.0.1 (20050412) =

Fixed a bug that caused bad results when you tried to reference a tag
name shorter than 3 characters as a member of a Tag, eg. tag.table.td.

Made sure all Tags have the 'hidden' attribute so that an attempt to
access tag.hidden doesn't spawn an attempt to find a tag named
'hidden'.

Fixed a bug in the comparison operator.

= 2.0.0 "Who cares for fish?" (20050410)

Beautiful Soup version 1 was very useful but also pretty stupid. I
originally wrote it without noticing any of the problems inherent in
trying to build a parse tree out of ambiguous HTML tags. This version
solves all of those problems to my satisfaction. It also adds many new
clever things to make up for the removal of the stupid things.

== Parsing ==

The parser logic has been greatly improved, and the BeautifulSoup
class should much more reliably yield a parse tree that looks like
what the page author intended. For a particular class of odd edge
cases that now causes problems, there is a new class,
ICantBelieveItsBeautifulSoup.

By default, Beautiful Soup now performs some cleanup operations on
text before parsing it. This is to avoid common problems with bad
definitions and self-closing tags that crash SGMLParser. You can
provide your own set of cleanup operations, or turn it off
altogether. The cleanup operations include fixing self-closing tags
that don't close, and replacing Microsoft smart quotes and similar
characters with their HTML entity equivalents.

You can now get a pretty-print version of parsed HTML to get a visual
picture of how Beautiful Soup parses it, with the Tag.prettify()
method.

== Strings and Unicode ==

There are separate NavigableText subclasses for ASCII and Unicode
strings. These classes directly subclass the corresponding base data
types. This means you can treat NavigableText objects as strings
instead of having to call methods on them to get the strings.

str() on a Tag always returns a string, and unicode() always returns
Unicode. Previously it was inconsistent.

== Tree traversal ==

In a first() or fetch() call, the tag name or the desired value of an
attribute can now be any of the following:

 * A string (matches that specific tag or that specific attribute value)
 * A list of strings (matches any tag or attribute value in the list)
 * A compiled regular expression object (matches any tag or attribute
   value that matches the regular expression)
 * A callable object that takes the Tag object or attribute value as a
   string. It returns None/false/empty string if the given string
   doesn't match, and any other value if it does.

This is much easier to use than SQL-style wildcards (see, regular
expressions are good for something). Because of this, I took out
SQL-style wildcards. I'll put them back if someone complains, but
their removal simplifies the code a lot.

You can use fetch() and first() to search for text in the parse tree,
not just tags. There are new alias methods fetchText() and firstText()
designed for this purpose. As with searching for tags, you can pass in
a string, a regular expression object, or a method to match your text.

If you pass in something besides a map to the attrs argument of
fetch() or first(), Beautiful Soup will assume you want to match that
thing against the "class" attribute. When you're scraping
well-structured HTML, this makes your code a lot cleaner.

1.x and 2.x both let you call a Tag object as a shorthand for
fetch(). For instance, foo("bar") is a shorthand for
foo.fetch("bar"). In 2.x, you can also access a specially-named member
of a Tag object as a shorthand for first(). For instance, foo.barTag
is a shorthand for foo.first("bar"). By chaining these shortcuts you
traverse a tree in very little code: for header in
soup.bodyTag.pTag.tableTag('th'):

If an element relationship (like parent or next) doesn't apply to a
tag, it'll now show up Null instead of None. first() will also return
Null if you ask it for a nonexistent tag. Null is an object that's
just like None, except you can do whatever you want to it and it'll
give you Null instead of throwing an error.

This lets you do tree traversals like soup.htmlTag.headTag.titleTag
without having to worry if the intermediate stages are actually
there. Previously, if there was no 'head' tag in the document, headTag
in that instance would have been None, and accessing its 'titleTag'
member would have thrown an AttributeError. Now, you can get what you
want when it exists, and get Null when it doesn't, without having to
do a lot of conditionals checking to see if every stage is None.

There are two new relations between page elements: previousSibling and
nextSibling. They reference the previous and next element at the same
level of the parse tree. For instance, if you have HTML like this:

  <p><ul><li>Foo<br /><li>Bar</ul>

The first 'li' tag has a previousSibling of Null and its nextSibling
is the second 'li' tag. The second 'li' tag has a nextSibling of Null
and its previousSibling is the first 'li' tag. The previousSibling of
the 'ul' tag is the first 'p' tag. The nextSibling of 'Foo' is the
'br' tag.

I took out the ability to use fetch() to find tags that have a
specific list of contents. See, I can't even explain it well. It was
really difficult to use, I never used it, and I don't think anyone
else ever used it. To the extent anyone did, they can probably use
fetchText() instead. If it turns out someone needs it I'll think of
another solution.

== Tree manipulation ==

You can add new attributes to a tag, and delete attributes from a
tag. In 1.x you could only change a tag's existing attributes.

== Porting Considerations ==

There are three changes in 2.0 that break old code:

In the post-1.2 release you could pass in a function into fetch(). The
function took a string, the tag name. In 2.0, the function takes the
actual Tag object.

It's no longer to pass in SQL-style wildcards to fetch(). Use a
regular expression instead.

The different parsing algorithm means the parse tree may not be shaped
like you expect. This will only actually affect you if your code uses
one of the affected parts. I haven't run into this problem yet while
porting my code.

= Between 1.2 and 2.0 =

This is the release to get if you want Python 1.5 compatibility.

The desired value of an attribute can now be any of the following:

 * A string
 * A string with SQL-style wildcards
 * A compiled RE object
 * A callable that returns None/false/empty string if the given value
   doesn't match, and any other value otherwise.

This is much easier to use than SQL-style wildcards (see, regular
expressions are good for something). Because of this, I no longer
recommend you use SQL-style wildcards. They may go away in a future
release to clean up the code.

Made Beautiful Soup handle processing instructions as text instead of
ignoring them.

Applied patch from Richie Hindle (richie at entrian dot com) that
makes tag.string a shorthand for tag.contents[0].string when the tag
has only one string-owning child.

Added still more nestable tags. The nestable tags thing won't work in
a lot of cases and needs to be rethought.

Fixed an edge case where searching for "%foo" would match any string
shorter than "foo".

= 1.2 "Who for such dainties would not stoop?" (20040708) =

Applied patch from Ben Last (ben at benlast dot com) that made
Tag.renderContents() correctly handle Unicode.

Made BeautifulStoneSoup even dumber by making it not implicitly close
a tag when another tag of the same type is encountered; only when an
actual closing tag is encountered. This change courtesy of Fuzzy (mike
at pcblokes dot com). BeautifulSoup still works as before.

= 1.1 "Swimming in a hot tureen" =

Added more 'nestable' tags. Changed popping semantics so that when a
nestable tag is encountered, tags are popped up to the previously
encountered nestable tag (of whatever kind). I will revert this if
enough people complain, but it should make more people's lives easier
than harder. This enhancement was suggested by Anthony Baxter (anthony
at interlink dot com dot au).

= 1.0 "So rich and green" (20040420) =

Initial release.