NEWS ( CHANGELOG and HISTORY ) HTMLPurifier ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| = KEY ==================== # Breaks back-compat ! Feature - Bugfix + Sub-comment . Internal change ========================== 4.8.0, released 2016-07-16 # By default, when a link has a target attribute associated with it, we now also add rel="noreferrer" in order to prevent the new window from being able to overwrite the original frame. To disable this protection, set %HTML.TargetNoreferrer to FALSE. ! Full PHP 7 compatibility, the test suite is ALL GO. ! %CSS.AllowDuplicates permits duplicate CSS properties. ! Support for 'tel' URIs. ! Partial support for 'border-radius' properties when %CSS.AllowProprietary is true. The slash syntax, i.e., 'border-radius: 2em 1em 4em / 0.5em 3em' is not yet supported. ! %Attr.ID.HTML5 turns on HTML5-style ID handling. - alt truncation could result in malformed UTF-8 sequence. Don't truncate. Thanks Brandon Farber for reporting. - Linkify regex is smarter, based off of Gruber's regex. - IDNA supported natively on PHP 5.3 and later. - Non all-numeric top-level names (e.g., foo.1f, 1f) are now allowed. - Minor bounds error fix to squash a PHP 7 notice. - Support non-/tmp temporary directories for data:// validation - Give a better error message when a user attempts to allow ul/ol without allowing li. - On some versions of PHP, the Serializer DefinitionCache could infinite loop when the directory exists but is not listable. (#49) - Don't match for inside comments with %Core.ConvertDocumentToFragment. (#67) - SafeObject is now less case sensitive. (#57) - AutoFormat.RemoveEmpty.Predicate now correctly renders in web form. (#85) 4.7.0, released 2015-08-04 # opacity is now considered a "tricky" CSS property rather than a proprietary one. ! %AutoFormat.RemoveEmpty.Predicate for specifying exactly when an element should be considered "empty" (maybe preserve if it has attributes), and modify iframe support so that the iframe is removed if it is missing a src attribute. Thanks meeva for reporting. - Don't truncate upon encountering when using DOMLex. Thanks Myrto Christina for finally convincing me to fix this. - Update YouTube filter for new code. - Fix parsing of rgb() values with spaces in them for 'border' attribute. - Don't remove foo="" attributes if foo is a boolean attribute. Thanks valME for reporting. 4.6.0, released 2013-11-30 # Secure URI munge hashing algorithm has changed to hash_hmac("sha256", $url, $secret). Please update any verification scripts you may have. # URI parsing algorithm was made more strict, so only prefixes which looks like schemes will actually be schemes. Thanks Michael Gusev for fixing. # %Core.EscapeInvalidChildren is no longer supported, and no longer does anything. ! New directive %Core.AllowHostnameUnderscore which allows underscores in hostnames. - Eliminate quadratic behavior in DOMLex by using a proper queue. Thanks Ole Laursen for noticing this. - Rewritten MakeWellFormed/FixNesting implementation eliminates quadratic behavior in the rest of the purificaiton pipeline. Thanks Chedburn Networks for sponsoring this work. - Made Linkify URL parser a bit less permissive, so that non-breaking spaces and commas are not included as part of URL. Thanks nAS for fixing. - Fix some bad interactions with %HTML.Allowed and injectors. Thanks David Hirtz for reporting. - Fix infinite loop in DirectLex. Thanks Ashar Javed (@soaj1664ashar) for reporting. 4.5.0, released 2013-02-17 # Fix bug where stacked attribute transforms clobber each other; this also means it's no longer possible to override attribute transforms in later modules. No internal code was using this but this may break some clients. # We now use SHA-1 to identify cached definitions, instead of MD5. ! Support display:inline-block ! Support for more white-space CSS values. ! Permit underscores in font families ! Support for page-break-* CSS3 properties when proprietary properties are enabled. ! New directive %Core.DisableExcludes; can be set to 'true' to turn off SGML excludes checking. If HTML Purifier is removing too much text and you don't care about full standards compliance, try setting this to 'true'. - Use prepend for SPL autoloading on PHP 5.3 and later. - Fix bug with nofollow transform when pre-existing rel exists. - Fix bug where background:url() always gets lower-cased (but not background-image:url()) - Fix bug with non lower-case color names in HTML - Fix bug where data URI validation doesn't remove temporary files. Thanks Javier Marín Ros for reporting. - Don't remove certain empty tags on RemoveEmpty. 4.4.0, released 2012-01-18 # Removed PEARSax3 handler. # URI.Munge now munges URIs inside the same host that go from https to http. Reported by Neike Taika-Tessaro. # Core.EscapeNonASCIICharacters now always transforms entities to entities, even if target encoding is UTF-8. # Tighten up selector validation in ExtractStyleBlocks. Non-syntactically valid selectors are now rejected, along with some of the more obscure ones such as attribute selectors, the :lang pseudoselector, and anything not in CSS2.1. Furthermore, ID and class selectors now work properly with the relevant configuration attributes. Also, mute errors when parsing CSS with CSS Tidy. Reported by Mario Heiderich and Norman Hippert. ! Added support for 'scope' attribute on tables. ! Added %HTML.TargetBlank, which adds target="blank" to all outgoing links. ! Properly handle sub-lists directly nested inside of lists in a standards compliant way, by moving them into the preceding
  • ! Added %HTML.AllowedComments and %HTML.AllowedCommentsRegexp for limited allowed comments in untrusted situations. ! Implement iframes, and allow them to be used in untrusted mode with %HTML.SafeIframe and %URI.SafeIframeRegexp. Thanks Bradley M. Froehle for submitting an initial version of the patch. ! The Forms module now works properly for transitional doctypes. ! Added support for internationalized domain names. You need the PEAR Net_IDNA2 module to be in your path; if it is installed, ensure the class can be loaded and then set %Core.EnableIDNA to true. - Color keywords are now case insensitive. Thanks Yzmir Ramirez for reporting. - Explicitly initialize anonModule variable to null. - Do not duplicate nofollow if already present. Thanks 178 for reporting. - Do not add nofollow if hostname matches our current host. Thanks 178 for reporting, and Neike Taika-Tessaro for helping diagnose. - Do not unset parser variable; this fixes intermittent serialization problems. Thanks Neike Taika-Tessaro for reporting, bill <10010tiger@gmail.com> for diagnosing. - Fix iconv truncation bug, where non-UTF-8 target encodings see output truncated after around 8000 characters. Thanks Jörg Ludwig for reporting. - Fix broken table content model for XHTML1.1 (and also earlier versions, although the W3C validator doesn't catch those violations). Thanks GlitchMr for reporting. 4.3.0, released 2011-03-27 # Fixed broken caching of customized raw definitions, but requires an API change. The old API still works but will emit a warning, see http://htmlpurifier.org/docs/enduser-customize.html#optimized for how to upgrade your code. # Protect against Internet Explorer innerHTML behavior by specially treating attributes with backticks but no angled brackets, quotes or spaces. This constitutes a slight semantic change, which can be reverted using %Output.FixInnerHTML. Reported by Neike Taika-Tessaro and Mario Heiderich. # Protect against cssText/innerHTML by restricting allowed characters used in fonts further than mandated by the specification and encoding some extra special characters in URLs. Reported by Neike Taika-Tessaro and Mario Heiderich. ! Added %HTML.Nofollow to add rel="nofollow" to external links. ! More types of SPL autoloaders allowed on later versions of PHP. ! Implementations for position, top, left, right, bottom, z-index when %CSS.Trusted is on. ! Add %Cache.SerializerPermissions option for custom serializer directory/file permissions ! Fix longstanding bug in Flash support for non-IE browsers, and allow more wmode attributes. ! Add %CSS.AllowedFonts to restrict permissible font names. - Switch to an iterative traversal of the DOM, which prevents us from running out of stack space for deeply nested documents. Thanks Maxim Krizhanovsky for contributing a patch. - Make removal of conditional IE comments ungreedy; thanks Bernd for reporting. - Escape CDATA before removing Internet Explorer comments. - Fix removal of id attributes under certain conditions by ensuring armor attributes are preserved when recreating tags. - Check if schema.ser was corrupted. - Check if zend.ze1_compatibility_mode is on, and error out if it is. This safety check is only done for HTMLPurifier.auto.php; if you are using standalone or the specialized includes files, you're expected to know what you're doing. - Stop repeatedly writing the cache file after I'm done customizing a raw definition. Reported by ajh. - Switch to using require_once in the Bootstrap to work around bad interaction with Zend Debugger and APC. Reported by Antonio Parraga. - Fix URI handling when hostname is missing but scheme is present. Reported by Neike Taika-Tessaro. - Fix missing numeric entities on DirectLex; thanks Neike Taika-Tessaro for reporting. - Fix harmless notice from indexing into empty string. Thanks Matthijs Kooijman for reporting. - Don't autoclose no parent elements are able to support the element that triggered the autoclose. In particular fixes strange behavior of stray
  • tags. Thanks pkuliga@gmail.com for reporting and Neike Taika-Tessaro for debugging assistance. 4.2.0, released 2010-09-15 ! Added %Core.RemoveProcessingInstructions, which lets you remove statements. ! Added %URI.DisableResources functionality; the directive originally did nothing. Thanks David Rothstein for reporting. ! Add documentation about configuration directive types. ! Add %CSS.ForbiddenProperties configuration directive. ! Add %HTML.FlashAllowFullScreen to permit embedded Flash objects to utilize full-screen mode. ! Add optional support for the file URI scheme, enable by explicitly setting %URI.AllowedSchemes. ! Add %Core.NormalizeNewlines options to allow turning off newline normalization. - Fix improper handling of Internet Explorer conditional comments by parser. Thanks zmonteca for reporting. - Fix missing attributes bug when running on Mac Snow Leopard and APC. Thanks sidepodcast for the fix. - Warn if an element is allowed, but an attribute it requires is not allowed. 4.1.1, released 2010-05-31 - Fix undefined index warnings in maintenance scripts. - Fix bug in DirectLex for parsing elements with a single attribute with entities. - Rewrite CSS output logic for font-family and url(). Thanks Mario Heiderich for reporting and Takeshi Terada for suggesting the fix. - Emit an error for CollectErrors if a body is extracted - Fix bug where in background-position for center keyword handling. - Fix infinite loop when a wrapper element is inserted in a context where it's not allowed. Thanks Lars for reporting. - Remove +x bit and shebang from index.php; only supported mode is to explicitly call it with php. - Make test script less chatty when log_errors is on. 4.1.0, released 2010-04-26 ! Support proprietary height attribute on table element ! Support YouTube slideshows that contain /cp/ in their URL. ! Support for data: URI scheme; not enabled by default, add it using %URI.AllowedSchemes ! Support flashvars when using %HTML.SafeObject and %HTML.SafeEmbed. ! Support for Internet Explorer compatibility with %HTML.SafeObject using %Output.FlashCompat. ! Handle
        properly, by inserting the necessary
      1. tag. - Always quote the insides of url(...) in CSS. 4.0.0, released 2009-07-07 # APIs for ConfigSchema subsystem have substantially changed. See docs/dev-config-bcbreaks.txt for details; in essence, anything that had both namespace and directive now have a single unified key. # Some configuration directives were renamed, specifically: %AutoFormatParam.PurifierLinkifyDocURL -> %AutoFormat.PurifierLinkify.DocURL %FilterParam.ExtractStyleBlocksEscaping -> %Filter.ExtractStyleBlocks.Escaping %FilterParam.ExtractStyleBlocksScope -> %Filter.ExtractStyleBlocks.Scope %FilterParam.ExtractStyleBlocksTidyImpl -> %Filter.ExtractStyleBlocks.TidyImpl As usual, the old directive names will still work, but will throw E_NOTICE errors. # The allowed values for class have been relaxed to allow all of CDATA for doctypes that are not XHTML 1.1 or XHTML 2.0. For old behavior, set %Attr.ClassUseCDATA to false. # Instead of appending the content model to an old content model, a blank element will replace the old content model. You can use #SUPER to get the old content model. ! More robust support for name="" and id="" ! HTMLPurifier_Config::inherit($config) allows you to inherit one configuration, and have changes to that configuration be propagated to all of its children. ! Implement %HTML.Attr.Name.UseCDATA, which relaxes validation rules on the name attribute when set. Use with care. Thanks Ian Cook for sponsoring. ! Implement %AutoFormat.RemoveEmpty.RemoveNbsp, which removes empty tags that contain non-breaking spaces as well other whitespace. You can also modify which tags should have   maintained with %AutoFormat.RemoveEmpty.RemoveNbsp.Exceptions. ! Implement %Attr.AllowedClasses, which allows administrators to restrict classes users can use to a specified finite set of classes, and %Attr.ForbiddenClasses, which is the logical inverse. ! You can now maintain your own configuration schema directories by creating a config-schema.php file or passing an extra argument. Check docs/dev-config-schema.html for more details. ! Added HTMLPurifier_Config->serialize() method, which lets you save away your configuration in a compact serial file, which you can unserialize and use directly without having to go through the overhead of setup. - Fix bug where URIDefinition would not get cleared if it's directives got changed. - Fix fatal error in HTMLPurifier_Encoder on certain platforms (probably NetBSD 5.0) - Fix bug in Linkify autoformatter involving http://foo - Make %URI.Munge not apply to links that have the same host as your host. - Prevent stray tag from truncating output, if a second is present. . Created script maintenance/rename-config.php for renaming a configuration directive while maintaining its alias. This script does not change source code. . Implement namespace locking for definition construction, to prevent bugs where a directive is used for definition construction but is not used to construct the cache hash. 3.3.0, released 2009-02-16 ! Implement CSS property 'overflow' when %CSS.AllowTricky is true. ! Implement generic property list classess - Fix bug with testEncodingSupportsASCII() algorithm when iconv() implementation does not do the "right thing" with characters not supported in the output set. - Spellcheck UTF-8: The Secret To Character Encoding - Fix improper removal of the contents of elements with only whitespace. Thanks Eric Wald for reporting. - Fix broken test suite in versions of PHP without spl_autoload_register() - Fix degenerate case with YouTube filter involving double hyphens. Thanks Pierre Attar for reporting. - Fix YouTube rendering problem on certain versions of Firefox. - Fix CSSDefinition Printer problems with decorators - Add text parameter to unit tests, forces text output . Add verbose mode to command line test runner, use (--verbose) . Turn on unit tests for UnitConverter . Fix missing version number in configuration %Attr.DefaultImageAlt (added 3.2.0) . Fix newline errors that caused spurious failures when CRLF HTML Purifier was tested on Linux. . Removed trailing whitespace from all text files, see remote-trailing-whitespace.php maintenance script. . Convert configuration to use property list backend. 3.2.0, released 2008-10-31 # Using %Core.CollectErrors forces line number/column tracking on, whereas previously you could theoretically turn it off. # HTMLPurifier_Injector->notifyEnd() is formally deprecated. Please use handleEnd() instead. ! %Output.AttrSort for when you need your attributes in alphabetical order to deal with a bug in FCKEditor. Requested by frank farmer. ! Enable HTML comments when %HTML.Trusted is on. Requested by Waldo Jaquith. ! Proper support for name attribute. It is now allowed and equivalent to the id attribute in a and img tags, and is only converted to id when %HTML.TidyLevel is heavy (for all doctypes). ! %AutoFormat.RemoveEmpty to remove some empty tags from documents. Please don't use on hand-written HTML. ! Add error-cases for unsupported elements in MakeWellFormed. This enables the strategy to be used, standalone, on untrusted input. ! %Core.AggressivelyFixLt is on by default. This causes more sensible processing of left angled brackets in smileys and other whatnot. ! Test scripts now have a 'type' parameter, which lets you say 'htmlpurifier', 'phpt', 'vtest', etc. in order to only execute those tests. This supercedes the --only-phpt parameter, although for backwards-compatibility the flag will still work. ! AutoParagraph auto-formatter will now preserve double-newlines upon output. Users who are not performing inbound filtering, this may seem a little useless, but as a bonus, the test suite and handling of edge cases is also improved. ! Experimental implementation of forms for %HTML.Trusted ! Track column numbers when maintain line numbers is on ! Proprietary 'background' attribute on table-related elements converted into corresponding CSS. Thanks Fusemail for sponsoring this feature! ! Add forward(), forwardUntilEndToken(), backward() and current() to Injector supertype. ! HTMLPurifier_Injector->handleEnd() permits modification to end tokens. The time of operation varies slightly from notifyEnd() as *all* end tokens are processed by the injector before they are subject to the well-formedness rules. ! %Attr.DefaultImageAlt allows overriding default behavior of setting alt to basename of image when not present. ! %AutoFormat.DisplayLinkURI neuters tags into plain text URLs. - Fix two bugs in %URI.MakeAbsolute; one involving empty paths in base URLs, the other involving an undefined $is_folder error. - Throw error when %Core.Encoding is set to a spurious value. Previously, this errored silently and returned false. - Redirected stderr to stdout for flush error output. - %URI.DisableExternal will now use the host in %URI.Base if %URI.Host is not available. - Do not re-munge URL if the output URL has the same host as the input URL. Requested by Chris. - Fix error in documentation regarding %Filter.ExtractStyleBlocks - Prevent ]]> from triggering %Core.ConvertDocumentToFragment - Fix bug with inline elements in blockquotes conflicting with strict doctype - Detect if HTML support is disabled for DOM by checking for loadHTML() method. - Fix bug where dots and double-dots in absolute URLs without hostname were not collapsed by URIFilter_MakeAbsolute. - Fix bug with anonymous modules operating on SafeEmbed or SafeObject elements by reordering their addition. - Will now throw exception on many error conditions during lexer creation; also throw an exception when MaintainLineNumbers is true, but a non-tracksLineNumbers is being used. - Detect if domxml extension is loaded, and use DirectLEx accordingly. - Improve handling of big numbers with floating point arithmetic in UnitConverter. Reported by David Morton. . Strategy_MakeWellFormed now operates in-place, saving memory and allowing for more interesting filter-backtracking . New HTMLPurifier_Injector->rewind() functionality, allows injectors to rewind index to reprocess tokens. . StringHashParser now allows for multiline sections with "empty" content; previously the section would remain undefined. . Added --quick option to multitest.php, which tests only the most recent release for each series. . Added --distro option to multitest.php, which accepts either 'normal' or 'standalone'. This supercedes --exclude-normal and --exclude-standalone 3.1.1, released 2008-06-19 # %URI.Munge now, by default, does not munge resources (for example, ) In order to enable this again, please set %URI.MungeResources to true. ! More robust imagecrash protection with height/width CSS with %CSS.MaxImgLength, and height/width HTML with %HTML.MaxImgLength. ! %URI.MungeSecretKey for secure URI munging. Thanks Chris for sponsoring this feature. Check out the corresponding documentation for details. (Att Nightly testers: The API for this feature changed before the general release. Namely, rename your directives %URI.SecureMungeSecretKey => %URI.MungeSecretKey and and %URI.SecureMunge => %URI.Munge) ! Implemented post URI filtering. Set member variable $post to true to set a URIFilter as such. ! Allow modules to define injectors via $info_injector. Injectors are automatically disabled if injector's needed elements are not found. ! Support for "safe" objects added, use %HTML.SafeObject and %HTML.SafeEmbed. Thanks Chris for sponsoring. If you've been using ad hoc code from the forums, PLEASE use this instead. ! Added substitutions for %e, %n, %a and %p in %URI.Munge (in order, embedded, tag name, attribute name, CSS property name). See %URI.Munge for more details. Requested by Jochem Blok. - Disable percent height/width attributes for img. - AttrValidator operations are now atomic; updates to attributes are not manifest in token until end of operations. This prevents naughty internal code from directly modifying CurrentToken when they're not supposed to. This semantics change was requested by frank farmer. - Percent encoding checks enabled for URI query and fragment - Fix stray backslashes in font-family; CSS Unicode character escapes are now properly resolved (although *only* in font-family). Thanks Takeshi Terada for reporting. - Improve parseCDATA algorithm to take into account newline normalization - Account for browser confusion between Yen character and backslash in Shift_JIS encoding. This fix generalizes to any other encoding which is not a strict superset of printable ASCII. Thanks Takeshi Terada for reporting. - Fix missing configuration parameter in Generator calls. Thanks vs for the partial patch. - Improved adherence to Unicode by checking for non-character codepoints. Thanks Geoffrey Sneddon for reporting. This may result in degraded performance for extremely large inputs. - Allow CSS property-value pair ''text-decoration: none''. Thanks Jochem Blok for reporting. . Added HTMLPurifier_UnitConverter and HTMLPurifier_Length for convenient handling of CSS-style lengths. HTMLPurifier_AttrDef_CSS_Length now uses this class. . API of HTMLPurifier_AttrDef_CSS_Length changed from __construct($disable_negative) to __construct($min, $max). __construct(true) is equivalent to __construct('0'). . Added HTMLPurifier_AttrDef_Switch class . Rename HTMLPurifier_HTMLModule_Tidy->construct() to setup() and bubble method up inheritance hierarchy to HTMLPurifier_HTMLModule. All HTMLModules get this called with the configuration object. All modules now use this rather than __construct(), although legacy code using constructors will still work--the new format, however, lets modules access the configuration object for HTML namespace dependant tweaks. . AttrDef_HTML_Pixels now takes a single construction parameter, pixels. . ConfigSchema data-structure heavily optimized; on average it uses a third the memory it did previously. The interface has changed accordingly, consult changes to HTMLPurifier_Config for details. . Variable parsing types now are magic integers instead of strings . Added benchmark for ConfigSchema . HTMLPurifier_Generator requires $config and $context parameters. If you don't know what they should be, use HTMLPurifier_Config::createDefault() and new HTMLPurifier_Context(). . Printers now properly distinguish between output configuration, and target configuration. This is not applicable to scripts using the Printers for HTML Purifier related tasks. . HTML/CSS Printers must be primed with prepareGenerator($gen_config), otherwise fatal errors will ensue. . URIFilter->prepare can return false in order to abort loading of the filter . Factory for AttrDef_URI implemented, URI#embedded to indicate URI that embeds an external resource. . %URI.Munge functionality factored out into a post-filter class. . Added CurrentCSSProperty context variable during CSS validation 3.1.0, released 2008-05-18 # Unnecessary references to objects (vestiges of PHP4) removed from method signatures. The following methods do not need references when assigning from them and will result in E_STRICT errors if you try: + HTMLPurifier_Config->get*Definition() [* = HTML, CSS] + HTMLPurifier_ConfigSchema::instance() + HTMLPurifier_DefinitionCacheFactory::instance() + HTMLPurifier_DefinitionCacheFactory->create() + HTMLPurifier_DoctypeRegistry->register() + HTMLPurifier_DoctypeRegistry->get() + HTMLPurifier_HTMLModule->addElement() + HTMLPurifier_HTMLModule->addBlankElement() + HTMLPurifier_LanguageFactory::instance() # Printer_ConfigForm's get*() functions were static-ified # %HTML.ForbiddenAttributes requires attribute declarations to be in the form of tag@attr, NOT tag.attr (which will throw an error and won't do anything). This is for forwards compatibility with XML; you'd do best to migrate an %HTML.AllowedAttributes directives to this syntax too. ! Allow index to be false for config from form creation ! Added HTMLPurifier::VERSION constant ! Commas, not dashes, used for serializer IDs. This change is forwards-compatible and allows for version numbers like "3.1.0-dev". ! %HTML.Allowed deals gracefully with whitespace anywhere, anytime! ! HTML Purifier's URI handling is a lot more robust, with much stricter validation checks and better percent encoding handling. Thanks Gareth Heyes for indicating security vulnerabilities from lax percent encoding. ! Bootstrap autoloader deals more robustly with classes that don't exist, preventing class_exists($class, true) from barfing. - InterchangeBuilder now alphabetizes its lists - Validation error in configdoc output fixed - Iconv and other encoding errors muted even with custom error handlers that do not honor error_reporting - Add protection against imagecrash attack with CSS height/width - HTMLPurifier::instance() created for consistency, is equivalent to getInstance() - Fixed and revamped broken ConfigForm smoketest - Bug with bool/null fields in Printer_ConfigForm fixed - Bug with global forbidden attributes fixed - Improved error messages for allowed and forbidden HTML elements and attributes - Missing (or null) in configdoc documentation restored - If DOM throws and exception during parsing with PH5P (occurs in newer versions of DOM), HTML Purifier punts to DirectLex - Fatal error with unserialization of ScriptRequired - Created directories are now chmod'ed properly - Fixed bug with fallback languages in LanguageFactory - Standalone testing setup properly with autoload . Out-of-date documentation revised . UTF-8 encoding check optimization as suggested by Diego . HTMLPurifier_Error removed in favor of exceptions . More copy() function removed; should use clone instead . More extensive unit tests for HTMLDefinition . assertPurification moved to central harness . HTMLPurifier_Generator accepts $config and $context parameters during instantiation, not runtime . Double-quotes outside of attribute values are now unescaped 3.1.0rc1, released 2008-04-22 # Autoload support added. Internal require_once's removed in favor of an explicit require list or autoloading. To use HTML Purifier, you must now either use HTMLPurifier.auto.php or HTMLPurifier.includes.php; setting the include path and including HTMLPurifier.php is insufficient--in such cases include HTMLPurifier.autoload.php as well to register our autoload handler (or modify your autoload function to check HTMLPurifier_Bootstrap::getPath($class)). You can also use HTMLPurifier.safe-includes.php for a less performance friendly but more user-friendly library load. # HTMLPurifier_ConfigSchema static functions are officially deprecated. Schema information is stored in the ConfigSchema directory, and the maintenance/generate-schema-cache.php generates the schema.ser file, which is now instantiated. Support for userland schema changes coming soon! # HTMLPurifier_Config will now throw E_USER_NOTICE when you use a directive alias; to get rid of these errors just modify your configuration to use the new directive name. # HTMLPurifier->addFilter is deprecated; built-in filters can now be enabled using %Filter.$filter_name or by setting your own filters using %Filter.Custom # Directive-level safety properties superceded in favor of module-level safety. Internal method HTMLModule->addElement() has changed, although the externally visible HTMLDefinition->addElement has *not* changed. ! Extra utility classes for testing and non-library operations can be found in extras/. Specifically, these are FSTools and ConfigDoc. You may find a use for these in your own project, but right now they are highly experimental and volatile. ! Integration with PHPT allows for automated smoketests ! Limited support for proprietary HTML elements, namely , sponsored by Chris. You can enable them with %HTML.Proprietary if your client demands them. ! Support for !important CSS cascade modifier. By default, this will be stripped from CSS, but you can enable it using %CSS.AllowImportant ! Support for display and visibility CSS properties added, set %CSS.AllowTricky to true to use them. ! HTML Purifier now has its own Exception hierarchy under HTMLPurifier_Exception. Developer error (not enduser error) can cause these to be triggered. ! Experimental kses() wrapper introduced with HTMLPurifier.kses.php ! Finally %CSS.AllowedProperties for tweaking allowed CSS properties without mucking around with HTMLPurifier_CSSDefinition ! ConfigDoc output has been enhanced with version and deprecation info. ! %HTML.ForbiddenAttributes and %HTML.ForbiddenElements implemented. - Autoclose now operates iteratively, i.e.
        now has both span tags closed. - Various HTMLPurifier_Config convenience functions now accept another parameter $schema which defines what HTMLPurifier_ConfigSchema to use besides the global default. - Fix bug with trusted script handling in libxml versions later than 2.6.28. - Fix bug in ExtractStyleBlocks with comments in style tags - Fix bug in comment parsing for DirectLex - Flush output now displayed when in command line mode for unit tester - Fix bug with rgb(0, 1, 2) color syntax with spaces inside shorthand syntax - HTMLPurifier_HTMLDefinition->addAttribute can now be called multiple times on the same element without emitting errors. - Fixed fatal error in PH5P lexer with invalid tag names . Plugins now get their own changelogs according to project conventions. . Convert tokens to use instanceof, reducing memory footprint and improving comparison speed. . Dry runs now supported in SimpleTest; testing facilities improved . Bootstrap class added for handling autoloading functionality . Implemented recursive glob at FSTools->globr . ConfigSchema now has instance methods for all corresponding define* static methods. . A couple of new historical maintenance scripts were added. . HTMLPurifier/HTMLModule/Tidy/XHTMLAndHTML4.php split into two files . tests/index.php can now be run from any directory. . HTMLPurifier_Token subclasses split into seperate files . HTMLPURIFIER_PREFIX now is defined in Bootstrap.php, NOT HTMLPurifier.php . HTMLPURIFIER_PREFIX can now be defined outside of HTML Purifier . New --php=php flag added, allows PHP executable to be specified (command line only!) . htmlpurifier_add_test() preferred method to translate test files in to classes, because it handles PHPT files too. . Debugger class is deprecated and will be removed soon. . Command line argument parsing for testing scripts revamped, now --opt value format is supported. . Smoketests now cleanup after magic quotes . Generator now can output comments (however, comments are still stripped from HTML Purifier output) . HTMLPurifier_ConfigSchema->validate() deprecated in favor of HTMLPurifier_VarParser->parse() . Integers auto-cast into float type by VarParser. . HTMLPURIFIER_STRICT removed; no validation is performed on runtime, only during cache generation . Reordered script calls in maintenance/flush.php . Command line scripts now honor exit codes . When --flush fails in unit testers, abort tests and print message . Improved documentation in docs/dev-flush.html about the maintenance scripts . copy() methods removed in favor of clone keyword 3.0.0, released 2008-01-06 # HTML Purifier is PHP 5 only! The 2.1.x branch will be maintained until PHP 4 is completely deprecated, but no new features will be added to it. + Visibility declarations added + Constructor methods renamed to __construct() + PHP4 reference cruft removed (in progress) ! CSS properties are now case-insensitive ! DefinitionCacheFactory now can register new implementations ! New HTMLPurifier_Filter_ExtractStyleBlocks for extracting