Revision history for HTML::DOM 0.039 18 April, 2010 Bug fix: Submitting a GET form with a data: URL (or any URI object without a query_form method) now just uses the URL itself, discarding any data. 0.038 28 March, 2010 Bug fix: A run-time require that was causing problems in taint mode has been removed. 0.037 3 January, 2010 New feature: getElementsByClassName 0.036 27 December, 2009 Incompatible change: • A newly-created HTML::DOM object no longer has html, head and body elements. These are not created until open, write or parse_file is called. New features: • New innerText method on HTML elements • css_url_fetcher used to have no return value. Now it returns a previously assigned callback function. • css_url_fetcher no longer clobbers an existing link element handler, because it now uses a new, separate mechanism. Bug fixes: • $elem->innerHTML now encodes HTML character entities in text nodes that are the element’s direct children. • Modifying the href attribute of a link element whose rel attribute is set to stylesheet now resets the style sheet. • The target and currentTarget methods of events now return empty lists (instead of lists containing undef) before event dispatch. • If an elem_handler removes the root element from the docu- ment, it no longer causes an error when HTML::DOM next tries to call an elem_handler. • parse_file no longer dies if called after ‘close’ or when the root element has been removed. • The close, innerHTML and elem_handler methods of the docu- ment no longer die if the root element has been removed. • HTML::DOM’s innerHTML (as opposed to Element’s) now returns an empty string when the document has no children. It used to return undef. Other changes: • HTML::DOM::Interface no longer lists HTML::DOM::TreeBuilder, as that internal-only class has gone. 0.035 1 November, 2009 Bug fixes: • Silly me. I fixed a bug in getElementById in the previous release, and in doing so caused a more serious bug. I blinded getElementById to sub-elements of forms. This is now fixed. • The note below under 0.033 about base being ‘better at deal- ing with invalid documents’ is wrong. That change actually made it worse. Sometimes it would work, and sometimes not, depending on previous regular expression matches. (This is now fixed, of course.) 0.034 25 October, 2009 Incompatible change: • Node insertion methods no longer produce WRONG_DOCUMENT_ERR errors. It turns out that every web browser (except IE 6) completely disregards the DOM spec in this regard. • When attr_event_listener was renamed event_handler, the old name was kept for temporary backward compatibility. The internals still used the old name, so subclasses overriding it would have to use it. Now the internals use the new name, so subclasses have to change. Bug fixes: • HTML::DOM’s title method no longer dies when the title ele- ment does not exist. • Event dispatch no longer leaves junk in $@. • Whitespace is now preserved when a <pre> element’s inner- HTML is assigned to. We fixed this by preserving all white- space everywhere, which is what web browsers do anyway. • The event handler wrapper that get_event_listeners returns no longer closes over the event target, but gets it from the event object passed as its first argument. This allows the listener to be used on another object (e.g., if a sub- class overrides get_event_listeners to retrieve listeners from another object). • getElementById now works when the element’s ID attribute is set to a reference. 0.033 18 October, 2009 New features: • HTML::DOM’s base method now falls back to the HTTP headers (of the response object) if the document lacks a <base> elem. It is also better at dealing with invalid documents that include a <base target> before a <base href>. • Coderefs registered via elem_handler are now triggered when a node is inserted into the document, as well as dur- ing parsing. Bug fix: • HTML::DOM::Interface now includes the contentWindow method in the HTMLFrameElement and HTMLIFrameElement interfaces. 0.032 9 October, 2009 Incompatible change: • An event_attr_handler that provides a subroutine ending with a line like defined $ret and !$ret and $event->preventDefault will now end up triggering preventDefault in the case of an undefined $ret. This is the result of the first new feature: New features: • The event dispatch mechanism now distinguishes between event listeners (added via addEventListener) and event han- dlers (those assigned via attr_event_listener or one of the on* methods), calling preventDefault as appropriate based on the latter’s return value. To achieve this, we have made get_event_listener return not the event handler itself, but a closure that calls the event handler and checks its return value. • The closure returned by get_event_listener also checks whether the handler is an object with a call_with method. If it is, it calls it with the current event target as its first argument and the event object as the second. • insertAdjacentHTML and insertAdjacentElement Bug fixes: • HTML::DOM::Interface now exports the UTF16 constant. • innerHTML no longer omits </p> before <table>. (In fact, it simply turns of TreeBuilder’s tag-omitting feature, which has bugs in it.) • as_text and as_HTML used not to pass arguments to the super- class’s methods. Other changes: • attr_event_listener has been renamed to event_handler (the original name was a big mistake). The old name is still available but will be removed in a future release. • Makefile.PL now lists Date::Format, which I forgot to include in 0.030 (lastModified uses it). Thanks to Andreas König for reporting this (RT #50273). 0.031 4 October, 2009 Bug fix: To conform to what ‘real’ web browsers do, a form’s elements collection no longer includes image buttons. Thanks to Steven Nikkel for reporting this problem. 0.030 17 September, 2009 New features: • EventTarget’s attr_event_listener method • The body element now delegates attribute event handlers to the ownerDocument’s ‘event_parent’ if there is one. (This makes the ‘onload’ in <body onload='...'> apply to the window.) • The init method of event objects now accepts target as an argument. • The same init method now returns the event object itself. • The write and writeln methods now accept multiple arguments. • lastModified • Form’s encoding method • A and area elements now have hash, host, etc. properties that refer to parts of the linked URL. Bug fixes: • HTML::DOM’s title method no longer dies when the title ele- ment is empty. • EventTarget methods beginning with ‘on’ (like ‘onclick’) used to return a hash ref of all non-capture handlers assigned to the named event. (Assignment always worked.) • Invalid method calls on any class inheriting from EventTar- get used to have a scrambled error message. • HTML::DOM::Interface now includes getComputedStyle. 0.029 28 August, 2009 Bug fixes: • HTML::DOM::Element::Form’s elements method, when called in list context, was ignoring the magic associations added in the previous version. Thanks to Terry Simons once more, for the bug report. 0.028 25 August, 2009 New features: • Form elements are now associated during parsing with the last-opened form element, even if it has been implicitly closed. This is to deal with badly-formed documents that contain aberrations like <td><form><td><input>. Thanks to Terry Simons for bringing this to my attention. 0.027 16 August, 2009 • Updated a test to match a change in CSS::DOM with regard to !important user agent rules. CSS::DOM 0.07 is now required. New features: • createEvent now accepts HTMLEvents as an event type. trig- ger_event now has the defaults for all the HTML event types listed in the DOM Level 2 Events specification. (These may need to be changed later to conform to HTML 5, which more closely resembles reality.) Bug fixes: • HTML collections and node lists created by getElements- By(Tag)Name used not to be updated when the write method was called or when an element handler was triggered. 0.026 27 July, 2009 Incompatible change: • The sheet method of link elements used to create an empty CSS::DOM object and would not allow assignment, which behav- iour became utterly useless with the API changes of CSS::DOM 0.03. This behaviour has been removed. sheet is now a simple accessor (except that it only works when the ‘rel’ attribute contains ‘stylesheet’). New features: • css_url_fetcher: Provide a URL-fetching subroutine and external style sheets will be parsed and turned into CSS::DOM objects. • Comments are now included in the HTML tree, as requested by an anonymous individual who filed RT #48168. 0.025 24 June, 2009 Bug fixes: • HTML::DOM::Interface now maps HTML::DOM::TreeBuilder to HTMLHtmlElement, instead of HTMLElement. (This class is used for the <html> element when the tree is being built.) • innerHTML no longer dies when serialising a text node that was created from an object with string overloading. 0.024 8 April, 2009 Forgot to increase the CSS::DOM required version to 0.06 in the previous release. 0.023 8 April, 2009 New features: • getComputedStyle Bug fixes: • A form element’s value method is now capable of unchecking a checkbox. • If the document was rewritten (via close and write) after styleSheets had already been accessed, a subsequent access would return the first list of style sheets again. • form-charset.t no longer leaves temporary files lying around (RT #41470). 0.022 17 October, 2008 A few memory leak fixes: • Whitespace after the </body> tag would cause the close meth- od to create a circularity. • Accessing an attribute node’s text node (i.e., autovivify- ing it) would create a circularity. • Deep clones were indestructible. • replaceChild would create a circular reference. 0.021 25 September, 2008 New feature: • (I)frame elements now have a contentWindow attribute, which you can set, although it is listed as read-only in H:D:Interface. contentDocument now uses this. Bug fixes: • Fixed some tests that failed on Windoze because they relied on an LWP bug that was fixed in 5.810 (filenames in file upload fields not being escaped) 0.020 20 September, 2008 Incompatible changes: • HTML::DOM’s defaultView attribute no longer has anything assigned to it by default. • The defaultView attribute now holds a weak reference to the view, and the view holds a strong reference to the document. New features: • HTML::DOM now has an event_parent method, with which you can add an object to the top of the event dispatch chain. • HTML::DOM::NodeList::Magic’s constructor is now officially public. (For those of you who read the source, it now accepts a second arg; which it didn’t before). Use this if you need to create a custom ‘live’ node list. • HTML::DOM now has an event_listeners_enabled method that allows event handlers to be disabled. (Default event han- dlers still run.) EventTarget uses this method, so any sub- class can implement it to achieve the same effect. Bug fixes: • HTML::DOM’s defaultView method and HTML::DOM::View’s docu- ment method now return empty list (as opposed to lists con- taining undef) if called in list context when there is no object assigned. • When there is no view object, HTML::DOM::Node’s trigger_event method no longer mangles the arguments passed to the superclass (preventing default actions and causing the event’s properties other than the event type not to be set). • innerHTML no longer dies when there are event attributes in the HTML code fed to it. 0.019 7 September, 2008 New feature: • EventTarget is now listed as a separate interface in %H:D:Interface. Whether you call this a bug fix or a new feature is up to you: • The DOM Level 0 methods for accessing HTML attributes now trigger DOMAttrModified events. (So mutation events are fin- ished; the implementation object also claims to support it now.) By fixing this I accidentally uncovered another bug: Two bug fixes: • Event attribute methods ($elem->onclick, etc.) are now list- ed in %HTML::DOM::Interface. • Unspecified attribute nodes (attribute nodes corresponding to HTML attributes that are not specified in the document but have default values in the DTD) are not supposed to be-- and are no longer-- created if the attribute *is* speci- fied but set to the empty string. This problem affected getAttribute, but not getAttributeNode. 0.018 4 September, 2008 New features/miscellaneous changes: • HTML::DOM’s defaultView attribute is now writable. • HTML::DOM::View is now an inside-out class, so anything can subclass it; and its ‘document’ attribute is writable. The constructor has been removed. • The location method has been added to HTML::DOM. Bug fixes: • innerHTML is now listed in %HTML::DOM::Interface. • An element’s innerHTML now serialises text nodes that are direct children of the element properly. • When the removeAttribute method is called and the attribute has previously been accessed as a node, the event object corresponding to the DOMAttrModified event that occurs now has its attrName set to the attribute's name, rather than undef, and its attrChange set to REMOVAL, rather than ADDITION. 0.017 2 September, 2008 New features/miscellaneous changes: • innerHTML is now writable. • Mutation events are now triggered automatic- ally by HTML::DOM::Element’s various get/set/ removeAttribute(node) methods. It does not yet happen with the shorthand methods (those named after the attributes; aka the DOM Level 0 interface). • Modification of the contents of a text or comment node now triggers a DOMCharacterDataModified event. • The EventTarget interface is now a separate module from Node, so that anything can inherit from it. • Event attribute methods ($elem->onclick, etc.) have been added. • An Attr node can now be used as an EventTarget. • Assigning a value to an Attr node now modifies its child text node in place instead of replacing it. Bug fixes: • Modifying a style object (as returned by an element’s style method) now immediately causes the corresponding attribute to be updated. (It used only to be updated on calls to getAttribute[Node]). • Modifying an attribute node for the style attribute or its child text node now updates the corresponding style object. • The child text node of an attr node now has its parentNode set correctly. • HTML::DOM::Element’s removeAttribute and setAttributeNode methods no longer die when the attribute is set to some ran- dom object (e.g., $elem->setAttribute(foo => bless )). • An element’s innerHTML method no longer includes the tags of the element itself (oops!); nor does it have a line break tacked on the end. • A node’s push_content method no longer dies when called with no arguments while the node is empty. • Direct changes to an event attribute node’s text node now reset the event handler, instead of leaving it as it is. (For those who don’t understand my obtuse jargon, this applies to $elem->getAttributeNode('onclick')->lastChild-> data("...").) 0.016 28 August, 2008 New features/miscellaneous changes: • The UIEvent, MouseEvent and MutationEvent interfaces have been added. (Mutation events for attributes and character data modified are not yet automatically triggered.) • Events now have an init method that works like initEvent, but takes named args. • trigger_event has been extended to allow named args specify- ing the event type and which default action to take. It also chooses the right event class and reasonable defaults for the event name passed to it. • default_event_handler_for’s submit_button and reset_button event types have been removed, and default default event handlers for specific event types (i.e., default event han- dlers already in place by default) have been replaced by another mechanism (not part of the public API). • The innerHTML has been added to HTML::DOM and HTML::DOM::Element. It is currently read-only. • Triggering a form’s reset event (or calling its reset method, which in turn triggers the event) now actually resets the form. Bug fixes: • A table cell following an unclosed table cell (<td><TD>) no longer mangles the tree. This was bro- ken in 0.011 which turned that into the nonsensical <tr><td></td><tr><td></td></tr></tr>. (This may have affected other elements, as well.) • HTMLOptionsCollection is now listed as acting as a hash and as an array in HTML::DOM::Interface. • The trigger_event method now works on the document with a string arg, instead of dying. 0.015 20 August, 2008 • Removed every last reference to CSS::DOM::StyleDecl, which was renamed in 0.03. 0.014 20 August, 2008 • Fixed the Makefile.PL so it installs properly and doesn’t downgrade CSS::DOM after installing 0.03. • HTML::DOM->createEvent now dies if the arg passed to it is an unsupported event category. 0.013 19 August, 2008 New features: • HTML::DOM::Interface now indicates which members have UTF-16 equivalents. • Changes to conform to the Level-2 HTML DOM module: • HTML::DOM::Element::Input’s ‘type’ method is no longer read-only. • HTML::DOM::Element::Object now has a contentDocument, which returns nothing for now. • HTML::DOM::Element::Table’s caption, tHead and tFoot meth- ods now die when passed something other an element of the right type. • The insertRow methods of HTML::DOM::Element::Table and ::TableSection and the insertCell method of ::TableRow now die if the index is out of range. • Frame and iframe elements now each have a contentDocument method, which, for now, returns an empty HTML::DOM object. Later I need to add support for callback routines that create the object. • HTML::DOM::Exception now has a ‘code’ method. • The brand new shiny default_event_handler_for method has been added. It makes it less cumbersome to provide default behaviours for different event types. • All HTML elements now have a click method--something from HTML 5 (I seem to be getting ahead of myself). Bug fixes: • removeChild now works on the document tree. • Modification of an Attr object for an event attribute now modifies the event handler itself. • HTML::DOM::Interface no longer lists the index property of the HTMLOptionsElement interface as read-only. • HTML::DOM::Table’s and ::TableSection’s insertRow methods and ::TableRow’s insertCell method now insert the new row or cell into the right place when called with an index other than 0 or -1. • HTML::DOM::Implementation->hasFeature now returns true for Core 1.0, and not just Core 2.0. (Oops!) 0.012 29 July, 2008 Incompatible changes: • The parse_file method now uses HTML::Encoding to detect the encoding of the file, if given a file name, and if the encoding has not been set with the new ‘charset’ method or the constructor’s new ‘charset’ option. It used to assume ISO Latin 1. It still treats file handles the same way, though; that is, it assumes the file handle is produc- ing Unicode text (you have to ‘binmode’ to set the encoding if that’s not the case). New features: • The constructor now accepts a ‘charset’ option which is used by parse_file and by the make_request method of form form elements. • A ‘charset’ method has been added to HTML::DOM, which returns the encoding passed to the constructor or that parse_file used. You can also use it to set it. • parse_file’s return value is now useful: true upon success; false upon failure (to open the file) • Upward references in the object tree are now weak references. • I’ve started work on Level-2 HTML features. So far, • HTML::DOM::Interface lists HTMLOptionsCollection as a sepa- rate interface from HTMLCollection and • HTML::DOM::Collection::Options now throws an error if length is set. • A callback routine passed to event_attr_handler now receives the offset within the source of the attribute’s value as the fourth argument. • Attribute methods of elements now support implicit attri- butes (those with default values in the DTD). Bug fixes: • Suppressed an erroneous warning triggered in bleadperl (5.11 to be) by HTML::DOM::Element::Form’s inputs method. • HTML::DOM’s close method had an erroneous eval, which would catch errors that should have been propagated, in addition to the specific error it was intended to catch. • HTML::DOM::Collection::Option’s name and form_name_value methods no longer die when there are no options. • HTML::DOM::Element::Option’s text method now supports multi- ple child nodes, and no longer assumes that there is just one text node within the option. • In cases where the HTML DTD specifies a list of alternates for an attribute’s value, the corresponding DOM 0 methods return a normalised (lc) value. • HTML::DOM::Element’s removeAttributeNode method used to emit warnings when the node did not belong to the element and the element either had no attribute with the node’s name or an attribute that had never been accessed as a node. 0.011 19 May, 2008 New features: • Link and style elements now have their ‘sheet’ attribute (in other words, the LinkStyle interface has been added). • HTML::DOM now has a C<styleSheets> method (the DocumentStyle interface has been implemented). One bug fix: • The so-called ‘content_offset’ bug (RT #34404) has been fixed. It actually had nothing to do with content_offset (which merely exposed the bug), but text after the </html> tag that was inserted in the document, rather than the HTML element. 0.010 23 December, 2007 New features: • HTML::DOM::Node’s trigger_event method now returns nothing. (It used to return the return value of the default event handler.) As a side effect, this fixes a few bugs. • The classes for HTML tables and frames have been added. • HTML::DOM::Element now has a content_offset method. • HTML::DOM::Element::Form now has its reset method. • The DOM Level-2 core interfaces are now supported. • I’ve started work on CSS features. So far, elements have a ‘style’ method. • You can now access forms by using an HTML::DOM object as a hash ref. • A checkbox’s click method now toggles the checkbox’s state. Fixed bugs: • HTML::DOM::Element::Form’s submit method now returns nothing. • HTML::DOM::Element::Select’s blur and focus methods now return nothing. • HTML::DOM::Element::Button’s type method now works. • HTML::DOM::Element::Label’s htmlFor method now works. • Heading elements (h1..h6) are now blessed into the HTML::DOM::Element::Heading class, so the align method works. • HTML::DOM::Element::Object’s useMap method now works. • HTML::DOM::Element::Applet’s codeBase method now works. • tbody elements are now inserted into the tree if a <tr> occurs directly within a <table>. • HTML::DOM::Interface now includes HTML::DOM::TreeBuilder, which is used for the document’s <html> element while the HTML code is being parsed. • The as_text and as_HTML methods now work on the document’s <html> element during parsing. • The <html> element’s ‘version’ method now works dur- ing parsing. • getElementById and getElementsByTagName now stringify the name or ID (in order to make HTML::Element’s look_down method happy). These both used to return false negatives when passed objects with string overloading. • HTML::DOM::Element’s normalize method (which has been moved to HTML::DOM::Node) now actually does something. • HTML::DOM::Node’s cloneNode method used not to erase the clone’s parentNode attribute unless the cloning was deep. Nor did it erase the list of child nodes in a shallow clone, so those nodes would have multiple parents. • cloneNode now works with HTML elements (the attributes are now cloned). • HTML::DOM::CharacterData’s deleteData16 method now works properly, instead of getting offsets wrong and then convert- ing all the text into UTF-16. • HTML::DOM::Collection’s namedItem method used to return undef for form, img and map elements. • HTML::DOM::Collection::Elements’s namedItem method used to die if there were multiple form fields with the same name. • HTML::DOM’s open method no longer clobbers element handlers. 0.009 23 November, 2007 Changes/new features: * You can do something with errors produced by event handlers by passing a coderef to the new error_handler method. That coderef will be called for each error. (Before, such errors were ignored.) * HTML::DOM::Interface has been modified, such that 'Document' is no longer a separate interface--since there is no corresponding HTML::DOM module--, but all its members are now under HTMLDocument, which inherits directly from Node. Likewise, 'Element' has been merged into HTMLElement. * HTML::DOM::Element::Select now has 'add' and 'remove' methods. * 'Select' elements can now be used as array refs, so you can write $select-> instead of $select->options->. * You can now assign undef to such an array element to delete it, so '$select-> = undef' does the same thing as 'shift @$select'. 'delete $select->' also works. Bug fixes: * $form->action now works when the document has no base url (used to die) * HTML::DOM::Element::Select's type method now works. * HTML::DOM::Element::Select's selectedIndex method and HTML::DOM::Element::Option's selected method should work reliably now. * HTML::DOM::Collection::Options is now listed by HTML::DOM::Interface. * HTML::DOM::Element::Input's click method now returns an empty list. * HTML::DOM's event_attr_handler is now actually useful. (It used to be ignored most of the time.) 0.008 5 October, 2007 * New HTML::DOM::Interface module * Added type and value methods to HTML::DOM::Element::Button * Added the charset, coords and tabIndex methods to HTML::DOM::Element::A. * HTML::DOM's createElement method now throws an HTML::DOM::Exception. 0.007 29 September, 2007 Two bug fixes: * The URL method no longer warns if there is no URL. * The previous release broke parsing of forms. (The rest of the document would be within the form element.) Now it should work. 0.006 27 September, 2007 - The following interfaces have been implemented: HTMLFormElement HTMLSelectElement HTMLOptGroupElement HTMLOptionElement HTMLInputElement HTMLTextAreaElement HTMLButtonElement HTMLLabelElement HTMLFieldSetElement HTMLLegendElement HTMLUListElement HTMLOListElement HTMLDListElement HTMLDirectoryElement HTMLMenuElement HTMLLIElement HTMLDivElement HTMLParagraphElement HTMLHeadingElement HTMLQuoteElement HTMLPreElement HTMLBRElement HTMLBaseFontElement HTMLFontElement HTMLHRElement HTMLModElement HTMLAnchorElement HTMLImageElement HTMLObjectElement HTMLParamElement HTMLAppletElement HTMLMapElement HTMLAreaElement HTMLScriptElement - Some methods have also been added for compatibility with WWW::Mechanize. - The 'open' method no longer stops the ownerDocument method from working properly. - Furthermore, 'open' is no longer anywhere near as powerful as it was; it used to obliterate the coderefs that were passed to event_attr_handler and default_event_handler. - &HTML::DOM::base has been added. - The parse and eof methods have been deleted, since write and close do the same thing. 0.005 26 August, 2007 The HTMLCollection interface has now been implemented, and so have the attribute methods of the HTMLDocument interface (Level 1). The HTMLHtmlElement, HTMLHeadElement, HTMLLinkElement, HTMLTitleElement, HTMLMetaElement, HTMLBaseElement, HTMLIsIndexElement, HTMLStyleElement and HTMLBodyElement interfaces (again, Level 1) have also been implemented. 0.004 22 July, 2007 Fixed infinite recursion triggered by element handlers calling $dom_tree->write($html) when $html contains elements handled by the same handler. E.g., <script>document.write('<script>...<\/script>')</script> 0.003 15 July, 2007 - HTML::DOM::Text now has its nodeValue method, which I forgot about before. (Actually, I moved HTML::DOM::Comment's nodeValue to HTML::DOM::CharacterData.) - as_text and as_HTML have been overridden and now work. 0.002 9 July, 2007 The basic mechanism for event handling has been implemented. The various different flavours of event objects still need to be added. 0.001 27 June, 2007 The first version. So far, most of the level-1 core DOM is implemented, and a couple of HTML-specific methods, but that's all.