pjs/content/base/public/nsIDocument.h

1032 строки
35 KiB
C++

/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* ***** BEGIN LICENSE BLOCK *****
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
*
* The contents of this file are subject to the Mozilla Public License Version
* 1.1 (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
* http://www.mozilla.org/MPL/
*
* Software distributed under the License is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
* for the specific language governing rights and limitations under the
* License.
*
* The Original Code is mozilla.org code.
*
* The Initial Developer of the Original Code is
* Netscape Communications Corporation.
* Portions created by the Initial Developer are Copyright (C) 1998
* the Initial Developer. All Rights Reserved.
*
* Contributor(s):
*
* Alternatively, the contents of this file may be used under the terms of
* either of the GNU General Public License Version 2 or later (the "GPL"),
* or the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
* in which case the provisions of the GPL or the LGPL are applicable instead
* of those above. If you wish to allow use of your version of this file only
* under the terms of either the GPL or the LGPL, and not to allow others to
* use your version of this file under the terms of the MPL, indicate your
* decision by deleting the provisions above and replace them with the notice
* and other provisions required by the GPL or the LGPL. If you do not delete
* the provisions above, a recipient may use your version of this file under
* the terms of any one of the MPL, the GPL or the LGPL.
*
* ***** END LICENSE BLOCK ***** */
#ifndef nsIDocument_h___
#define nsIDocument_h___
#include "nsINode.h"
#include "nsStringGlue.h"
#include "nsIDocumentObserver.h" // for nsUpdateType
#include "nsCOMPtr.h"
#include "nsIURI.h"
#include "nsIBindingManager.h"
#include "nsWeakPtr.h"
#include "nsIWeakReferenceUtils.h"
#include "nsILoadGroup.h"
#include "nsCRT.h"
#include "mozFlushType.h"
#include "nsIAtom.h"
class nsIContent;
class nsPresContext;
class nsIPresShell;
class nsIStreamListener;
class nsIStreamObserver;
class nsStyleSet;
class nsIStyleSheet;
class nsIStyleRule;
class nsIViewManager;
class nsIScriptGlobalObject;
class nsPIDOMWindow;
class nsIDOMEvent;
class nsIDeviceContext;
class nsIParser;
class nsIDOMNode;
class nsIDOMDocumentFragment;
class nsILineBreaker;
class nsIWordBreaker;
class nsISelection;
class nsIChannel;
class nsIPrincipal;
class nsIDOMDocument;
class nsIDOMDocumentType;
class nsIObserver;
class nsISupportsArray;
class nsIScriptLoader;
class nsIContentSink;
class nsIScriptEventManager;
class nsNodeInfoManager;
class nsICSSLoader;
class nsHTMLStyleSheet;
class nsIHTMLCSSStyleSheet;
class nsILayoutHistoryState;
class nsIVariant;
class nsIDOMUserDataHandler;
template<class E> class nsCOMArray;
class nsIDocumentObserver;
// IID for the nsIDocument interface
#define NS_IDOCUMENT_IID \
{ 0x3b3cb52f, 0xf604, 0x48d5, \
{ 0xb5, 0x70, 0xf3, 0x25, 0xbb, 0x0a, 0x7a, 0x22 } }
// Flag for AddStyleSheet().
#define NS_STYLESHEET_FROM_CATALOG (1 << 0)
//----------------------------------------------------------------------
// Document interface. This is implemented by all document objects in
// Gecko.
class nsIDocument : public nsINode
{
public:
NS_DECLARE_STATIC_IID_ACCESSOR(NS_IDOCUMENT_IID)
NS_DECL_AND_IMPL_ZEROING_OPERATOR_NEW
#ifdef MOZILLA_INTERNAL_API
nsIDocument()
: nsINode(nsnull),
mCharacterSet(NS_LITERAL_CSTRING("ISO-8859-1")),
mNodeInfoManager(nsnull),
mPartID(0)
{
}
#endif
/**
* Let the document know that we're starting to load data into it.
* @param aCommand The parser command
* XXXbz It's odd to have that here.
* @param aChannel The channel the data will come from
* @param aLoadGroup The loadgroup this document should use from now on.
* Note that the document might not be the only thing using
* this loadgroup.
* @param aContainer The container this document is in. This may be null.
* XXXbz maybe we should make it more explicit (eg make the
* container an nsIWebNavigation or nsIDocShell or
* something)?
* @param [out] aDocListener the listener to pump data from the channel into.
* Generally this will be the parser this document
* sets up, or some sort of data-handler for media
* documents.
* @param aReset whether the document should call Reset() on itself. If this
* is false, the document will NOT set its principal to the
* channel's owner, will not clear any event listeners that are
* already set on it, etc.
* @param aSink The content sink to use for the data. If this is null and
* the document needs a content sink, it will create one based
* on whatever it knows about the data it's going to load.
*/
virtual nsresult StartDocumentLoad(const char* aCommand,
nsIChannel* aChannel,
nsILoadGroup* aLoadGroup,
nsISupports* aContainer,
nsIStreamListener **aDocListener,
PRBool aReset,
nsIContentSink* aSink = nsnull) = 0;
virtual void StopDocumentLoad() = 0;
/**
* Return the title of the document. This will return a void string
* if there is no title for this document).
*/
const nsString& GetDocumentTitle() const
{
return mDocumentTitle;
}
/**
* Return the URI for the document. May return null.
*/
nsIURI* GetDocumentURI() const
{
return mDocumentURI;
}
void SetDocumentURI(nsIURI* aURI)
{
mDocumentURI = aURI;
}
/**
* Set the principal responsible for this document.
*/
virtual void SetPrincipal(nsIPrincipal *aPrincipal) = 0;
/**
* Return the LoadGroup for the document. May return null.
*/
already_AddRefed<nsILoadGroup> GetDocumentLoadGroup() const
{
nsILoadGroup *group = nsnull;
if (mDocumentLoadGroup)
CallQueryReferent(mDocumentLoadGroup.get(), &group);
return group;
}
/**
* Return the base URI for relative URIs in the document (the document uri
* unless it's overridden by SetBaseURI, HTML <base> tags, etc.). The
* returned URI could be null if there is no document URI.
*/
nsIURI* GetBaseURI() const
{
return mDocumentBaseURI ? mDocumentBaseURI : mDocumentURI;
}
virtual nsresult SetBaseURI(nsIURI* aURI) = 0;
/**
* Get/Set the base target of a link in a document.
*/
virtual void GetBaseTarget(nsAString &aBaseTarget) const = 0;
virtual void SetBaseTarget(const nsAString &aBaseTarget) = 0;
/**
* Return a standard name for the document's character set. This
* will trigger a startDocumentLoad if necessary to answer the
* question.
*/
const nsCString& GetDocumentCharacterSet() const
{
return mCharacterSet;
}
/**
* Set the document's character encoding. |aCharSetID| should be canonical.
* That is, callers are responsible for the charset alias resolution.
*/
virtual void SetDocumentCharacterSet(const nsACString& aCharSetID) = 0;
PRInt32 GetDocumentCharacterSetSource() const
{
return mCharacterSetSource;
}
void SetDocumentCharacterSetSource(PRInt32 aCharsetSource)
{
mCharacterSetSource = aCharsetSource;
}
/**
* Add an observer that gets notified whenever the charset changes.
*/
virtual nsresult AddCharSetObserver(nsIObserver* aObserver) = 0;
/**
* Remove a charset observer.
*/
virtual void RemoveCharSetObserver(nsIObserver* aObserver) = 0;
/**
* Get the Content-Type of this document.
* (This will always return NS_OK, but has this signature to be compatible
* with nsIDOMNSDocument::GetContentType())
*/
NS_IMETHOD GetContentType(nsAString& aContentType) = 0;
/**
* Set the Content-Type of this document.
*/
virtual void SetContentType(const nsAString& aContentType) = 0;
/**
* Return the language of this document.
*/
void GetContentLanguage(nsAString& aContentLanguage) const
{
CopyASCIItoUTF16(mContentLanguage, aContentLanguage);
}
// The state BidiEnabled should persist across multiple views
// (screen, print) of the same document.
/**
* Check if the document contains bidi data.
* If so, we have to apply the Unicode Bidi Algorithm.
*/
PRBool GetBidiEnabled() const
{
return mBidiEnabled;
}
/**
* Indicate the document contains bidi data.
* Currently, we cannot disable bidi, because once bidi is enabled,
* it affects a frame model irreversibly, and plays even though
* the document no longer contains bidi data.
*/
void SetBidiEnabled(PRBool aBidiEnabled)
{
mBidiEnabled = aBidiEnabled;
}
/**
* Get the bidi options for this document.
* @see nsBidiUtils.h
*/
PRUint32 GetBidiOptions() const
{
return mBidiOptions;
}
/**
* Set the bidi options for this document. This just sets the bits;
* callers are expected to take action as needed if they want this
* change to actually change anything immediately.
* @see nsBidiUtils.h
*/
void SetBidiOptions(PRUint32 aBidiOptions)
{
mBidiOptions = aBidiOptions;
}
/**
* Access HTTP header data (this may also get set from other
* sources, like HTML META tags).
*/
virtual void GetHeaderData(nsIAtom* aHeaderField, nsAString& aData) const = 0;
virtual void SetHeaderData(nsIAtom* aheaderField, const nsAString& aData) = 0;
/**
* Create a new presentation shell that will use aContext for its
* presentation context (presentation contexts <b>must not</b> be
* shared among multiple presentation shells).
*/
virtual nsresult CreateShell(nsPresContext* aContext,
nsIViewManager* aViewManager,
nsStyleSet* aStyleSet,
nsIPresShell** aInstancePtrResult) = 0;
virtual PRBool DeleteShell(nsIPresShell* aShell) = 0;
virtual PRUint32 GetNumberOfShells() const = 0;
virtual nsIPresShell *GetShellAt(PRUint32 aIndex) const = 0;
virtual void SetShellsHidden(PRBool aHide) = 0;
/**
* Return the parent document of this document. Will return null
* unless this document is within a compound document and has a
* parent. Note that this parent chain may cross chrome boundaries.
*/
nsIDocument *GetParentDocument() const
{
return mParentDocument;
}
/**
* Set the parent document of this document.
*/
void SetParentDocument(nsIDocument* aParent)
{
mParentDocument = aParent;
}
/**
* Set the sub document for aContent to aSubDoc.
*/
virtual nsresult SetSubDocumentFor(nsIContent *aContent,
nsIDocument* aSubDoc) = 0;
/**
* Get the sub document for aContent
*/
virtual nsIDocument *GetSubDocumentFor(nsIContent *aContent) const = 0;
/**
* Find the content node for which aDocument is a sub document.
*/
virtual nsIContent *FindContentForSubDocument(nsIDocument *aDocument) const = 0;
/**
* Return the root content object for this document.
*/
nsIContent *GetRootContent() const
{
return mRootContent;
}
/**
* Accessors to the collection of stylesheets owned by this document.
* Style sheets are ordered, most significant last.
*/
/**
* Get the number of stylesheets
*
* @return the number of stylesheets
* @throws no exceptions
*/
virtual PRInt32 GetNumberOfStyleSheets() const = 0;
/**
* Get a particular stylesheet
* @param aIndex the index the stylesheet lives at. This is zero-based
* @return the stylesheet at aIndex. Null if aIndex is out of range.
* @throws no exceptions
*/
virtual nsIStyleSheet* GetStyleSheetAt(PRInt32 aIndex) const = 0;
/**
* Insert a sheet at a particular spot in the stylesheet list (zero-based)
* @param aSheet the sheet to insert
* @param aIndex the index to insert at. This index will be
* adjusted for the "special" sheets.
* @throws no exceptions
*/
virtual void InsertStyleSheetAt(nsIStyleSheet* aSheet, PRInt32 aIndex) = 0;
/**
* Get the index of a particular stylesheet. This will _always_
* consider the "special" sheets as part of the sheet list.
* @param aSheet the sheet to get the index of
* @return aIndex the index of the sheet in the full list
*/
virtual PRInt32 GetIndexOfStyleSheet(nsIStyleSheet* aSheet) const = 0;
/**
* Replace the stylesheets in aOldSheets with the stylesheets in
* aNewSheets. The two lists must have equal length, and the sheet
* at positon J in the first list will be replaced by the sheet at
* position J in the second list. Some sheets in the second list
* may be null; if so the corresponding sheets in the first list
* will simply be removed.
*/
virtual void UpdateStyleSheets(nsCOMArray<nsIStyleSheet>& aOldSheets,
nsCOMArray<nsIStyleSheet>& aNewSheets) = 0;
/**
* Add a stylesheet to the document
*/
virtual void AddStyleSheet(nsIStyleSheet* aSheet) = 0;
/**
* Remove a stylesheet from the document
*/
virtual void RemoveStyleSheet(nsIStyleSheet* aSheet) = 0;
/**
* Notify the document that the applicable state of the sheet changed
* and that observers should be notified and style sets updated
*/
virtual void SetStyleSheetApplicableState(nsIStyleSheet* aSheet,
PRBool aApplicable) = 0;
/**
* Just like the style sheet API, but for "catalog" sheets,
* extra sheets inserted at the UA level.
*/
virtual PRInt32 GetNumberOfCatalogStyleSheets() const = 0;
virtual nsIStyleSheet* GetCatalogStyleSheetAt(PRInt32 aIndex) const = 0;
virtual void AddCatalogStyleSheet(nsIStyleSheet* aSheet) = 0;
virtual void EnsureCatalogStyleSheet(const char *aStyleSheetURI) = 0;
/**
* Get this document's CSSLoader. This is guaranteed to not return null.
*/
nsICSSLoader* CSSLoader() const {
return mCSSLoader;
}
/**
* Get the channel that was passed to StartDocumentLoad or Reset for this
* document. Note that this may be null in some cases (eg if
* StartDocumentLoad or Reset were never called)
*/
virtual nsIChannel* GetChannel() const = 0;
/**
* Get this document's attribute stylesheet. May return null if
* there isn't one.
*/
virtual nsHTMLStyleSheet* GetAttributeStyleSheet() const = 0;
/**
* Get this document's inline style sheet. May return null if there
* isn't one
*/
virtual nsIHTMLCSSStyleSheet* GetInlineStyleSheet() const = 0;
/**
* Get/set the object from which a document can get a script context
* and scope. This is the context within which all scripts (during
* document creation and during event handling) will run. Note that
* this is the *inner* window object.
*/
virtual nsIScriptGlobalObject* GetScriptGlobalObject() const = 0;
virtual void SetScriptGlobalObject(nsIScriptGlobalObject* aGlobalObject) = 0;
/**
* Get the object that is used as the scope for all of the content
* wrappers whose owner document is this document. Unlike the script
* global object, this never changes once it's set. Use this object
* when you're trying to find a content wrapper in XPConnect.
*/
virtual nsIScriptGlobalObject* GetScopeObject() = 0;
/**
* Return the window containing the document (the outer window).
*/
virtual nsPIDOMWindow *GetWindow() = 0;
/**
* Return the inner window used as the script compilation scope for
* this document. If you're not absolutely sure you need this, use
* GetWindow().
*/
virtual nsPIDOMWindow *GetInnerWindow() = 0;
/**
* Get the script loader for this document
*/
virtual nsIScriptLoader* GetScriptLoader() = 0;
//----------------------------------------------------------------------
// Document notification API's
/**
* Add a new observer of document change notifications. Whenever
* content is changed, appended, inserted or removed the observers are
* informed.
*/
virtual void AddObserver(nsIDocumentObserver* aObserver) = 0;
/**
* Remove an observer of document change notifications. This will
* return false if the observer cannot be found.
*/
virtual PRBool RemoveObserver(nsIDocumentObserver* aObserver) = 0;
// Observation hooks used to propagate notifications to document observers.
// BeginUpdate must be called before any batch of modifications of the
// content model or of style data, EndUpdate must be called afterward.
// To make this easy and painless, use the mozAutoDocUpdate helper class.
virtual void BeginUpdate(nsUpdateType aUpdateType) = 0;
virtual void EndUpdate(nsUpdateType aUpdateType) = 0;
virtual void BeginLoad() = 0;
virtual void EndLoad() = 0;
virtual void CharacterDataChanged(nsIContent* aContent, PRBool aAppend) = 0;
// notify that one or two content nodes changed state
// either may be nsnull, but not both
virtual void ContentStatesChanged(nsIContent* aContent1,
nsIContent* aContent2,
PRInt32 aStateMask) = 0;
virtual void AttributeWillChange(nsIContent* aChild,
PRInt32 aNameSpaceID,
nsIAtom* aAttribute) = 0;
virtual void AttributeChanged(nsIContent* aChild,
PRInt32 aNameSpaceID,
nsIAtom* aAttribute,
PRInt32 aModType) = 0;
virtual void ContentAppended(nsIContent* aContainer,
PRInt32 aNewIndexInContainer) = 0;
virtual void ContentInserted(nsIContent* aContainer,
nsIContent* aChild,
PRInt32 aIndexInContainer) = 0;
virtual void ContentRemoved(nsIContent* aContainer,
nsIContent* aChild,
PRInt32 aIndexInContainer) = 0;
// Observation hooks for style data to propagate notifications
// to document observers
virtual void StyleRuleChanged(nsIStyleSheet* aStyleSheet,
nsIStyleRule* aOldStyleRule,
nsIStyleRule* aNewStyleRule) = 0;
virtual void StyleRuleAdded(nsIStyleSheet* aStyleSheet,
nsIStyleRule* aStyleRule) = 0;
virtual void StyleRuleRemoved(nsIStyleSheet* aStyleSheet,
nsIStyleRule* aStyleRule) = 0;
/**
* Flush notifications for this document and its parent documents
* (since those may affect the layout of this one).
*/
virtual void FlushPendingNotifications(mozFlushType aType) = 0;
nsIBindingManager* BindingManager() const
{
return mBindingManager;
}
/**
* Only to be used inside Gecko, you can't really do anything with the
* pointer outside Gecko anyway.
*/
nsNodeInfoManager* NodeInfoManager() const
{
return mNodeInfoManager;
}
/**
* Reset the document using the given channel and loadgroup. This works
* like ResetToURI, but also sets the document's channel to aChannel.
*/
virtual void Reset(nsIChannel* aChannel, nsILoadGroup* aLoadGroup) = 0;
/**
* Reset this document to aURI and aLoadGroup. aURI must not be null.
*/
virtual void ResetToURI(nsIURI *aURI, nsILoadGroup* aLoadGroup) = 0;
/**
* Set the container (docshell) for this document.
*/
void SetContainer(nsISupports *aContainer)
{
mDocumentContainer = do_GetWeakReference(aContainer);
}
/**
* Get the container (docshell) for this document.
*/
already_AddRefed<nsISupports> GetContainer() const
{
nsISupports* container = nsnull;
if (mDocumentContainer)
CallQueryReferent(mDocumentContainer.get(), &container);
return container;
}
virtual nsIScriptEventManager* GetScriptEventManager() = 0;
/**
* Set and get XML declaration. If aVersion is null there is no declaration.
* aStandalone takes values -1, 0 and 1 indicating respectively that there
* was no standalone parameter in the declaration, that it was given as no,
* or that it was given as yes.
*/
virtual void SetXMLDeclaration(const PRUnichar *aVersion,
const PRUnichar *aEncoding,
const PRInt32 aStandalone) = 0;
virtual void GetXMLDeclaration(nsAString& aVersion,
nsAString& aEncoding,
nsAString& Standalone) = 0;
virtual PRBool IsCaseSensitive()
{
return PR_TRUE;
}
virtual PRBool IsScriptEnabled() = 0;
virtual nsresult AddXMLEventsContent(nsIContent * aXMLEventsElement) = 0;
virtual PRBool IsLoadedAsData()
{
return PR_FALSE;
}
/**
* Create an element with the specified name, prefix and namespace ID.
* If aDocumentDefaultType is true we create an element of the default type
* for that document (currently XHTML in HTML documents and XUL in XUL
* documents), otherwise we use the type specified by the namespace ID.
*/
virtual nsresult CreateElem(nsIAtom *aName, nsIAtom *aPrefix,
PRInt32 aNamespaceID,
PRBool aDocumentDefaultType,
nsIContent** aResult) = 0;
/**
* Get the security info (i.e. SSL state etc) that the document got
* from the channel/document that created the content of the
* document.
*
* @see nsIChannel
*/
nsISupports *GetSecurityInfo()
{
return mSecurityInfo;
}
/**
* Returns the default namespace ID used for elements created in this
* document.
*/
virtual PRInt32 GetDefaultNamespaceID() const = 0;
nsPropertyTable* PropertyTable() { return &mPropertyTable; }
/**
* Sets the ID used to identify this part of the multipart document
*/
void SetPartID(PRUint32 aID) {
mPartID = aID;
}
/**
* Return the ID used to identify this part of the multipart document
*/
PRUint32 GetPartID() const {
return mPartID;
}
/**
* Sanitize the document by resetting all input elements and forms that have
* autocomplete=off to their default values.
*/
virtual nsresult Sanitize() = 0;
/**
* Enumerate all subdocuments.
* The enumerator callback should return PR_TRUE to continue enumerating, or
* PR_FALSE to stop.
*/
typedef PRBool (*nsSubDocEnumFunc)(nsIDocument *aDocument, void *aData);
virtual void EnumerateSubDocuments(nsSubDocEnumFunc aCallback,
void *aData) = 0;
/**
* Check whether it is safe to cache the presentation of this document
* and all of its subdocuments. This method checks the following conditions
* recursively:
* - Some document types, such as plugin documents, cannot be safely cached.
* - If there are any pending requests, we don't allow the presentation
* to be cached. Ideally these requests would be suspended and resumed,
* but that is difficult in some cases, such as XMLHttpRequest.
* - If there are any beforeunload or unload listeners, we must fire them
* for correctness, but this likely puts the document into a state where
* it would not function correctly if restored.
*
* |aNewRequest| should be the request for a new document which will
* replace this document in the docshell. The new document's request
* will be ignored when checking for active requests. If there is no
* request associated with the new document, this parameter may be null.
*/
virtual PRBool CanSavePresentation(nsIRequest *aNewRequest) = 0;
/**
* Notify the document that its associated ContentViewer is being destroyed.
* This releases circular references so that the document can go away.
* Destroy() is only called on documents that have a content viewer.
*/
virtual void Destroy() = 0;
/**
* Get the layout history state that should be used to save and restore state
* for nodes in this document. This may return null; if that happens state
* saving and restoration is not possible.
*/
virtual already_AddRefed<nsILayoutHistoryState> GetLayoutHistoryState() const = 0;
/**
* Methods that can be used to prevent onload firing while an event that
* should block onload is posted. onload is guaranteed to not fire until
* either all calls to BlockOnload() have been matched by calls to
* UnblockOnload() or the load has been stopped altogether (by the user
* pressing the Stop button, say).
*/
virtual void BlockOnload() = 0;
/**
* @param aFireSync whether to fire onload synchronously. If false,
* onload will fire asynchronously after all onload blocks have been
* removed. It will NOT fire from inside UnblockOnload. If true,
* onload may fire from inside UnblockOnload.
*/
virtual void UnblockOnload(PRBool aFireSync) = 0;
/**
* Notification that the page has been shown, for documents which are loaded
* into a DOM window. This corresponds to the completion of document load,
* or to the page's presentation being restored into an existing DOM window.
* This notification fires applicable DOM events to the content window. See
* nsIDOMPageTransitionEvent.idl for a description of the |aPersisted|
* parameter.
*/
virtual void OnPageShow(PRBool aPersisted) = 0;
/**
* Notification that the page has been hidden, for documents which are loaded
* into a DOM window. This corresponds to the unloading of the document, or
* to the document's presentation being saved but removed from an existing
* DOM window. This notification fires applicable DOM events to the content
* window. See nsIDOMPageTransitionEvent.idl for a description of the
* |aPersisted| parameter.
*/
virtual void OnPageHide(PRBool aPersisted) = 0;
/*
* We record the set of links in the document that are relevant to
* style.
*/
/**
* Notification that an element is a link with a given URI that is
* relevant to style.
*/
virtual void AddStyleRelevantLink(nsIContent* aContent, nsIURI* aURI) = 0;
/**
* Notification that an element is a link and its URI might have been
* changed or the element removed. If the element is still a link relevant
* to style, then someone must ensure that AddStyleRelevantLink is
* (eventually) called on it again.
*/
virtual void ForgetLink(nsIContent* aContent) = 0;
/**
* Notification that the visitedness state of a URI has been changed
* and style related to elements linking to that URI should be updated.
*/
virtual void NotifyURIVisitednessChanged(nsIURI* aURI) = 0;
/**
* Associate an object aData to aKey on node aObject. Should only be used to
* implement the DOM Level 3 UserData API.
*
* @param aObject canonical nsINode pointer of the node to add aData to
* @param aKey the key to associate the object to
* @param aData the object to associate to aKey on aObject
* @param aHandler the UserDataHandler to call when the node is
* cloned/deleted/imported/renamed
* @param aResult [out] the previously registered object for aKey on aObject,
* if any
* @return whether adding the object and UserDataHandler succeeded
*/
nsresult SetUserData(const nsINode *aObject, const nsAString &aKey,
nsIVariant *aData, nsIDOMUserDataHandler *aHandler,
nsIVariant **aResult)
{
nsCOMPtr<nsIAtom> key = do_GetAtom(aKey);
if (!key) {
return NS_ERROR_OUT_OF_MEMORY;
}
return SetUserData(aObject, key, aData, aHandler, aResult);
}
/**
* Associate an object aData to aKey on node aObject. If aData is null any
* previously registered object and UserDataHandler associated to aKey on
* aObject will be removed. Should only be used to implement the DOM Level 3
* UserData API.
*
* @param aObject canonical nsINode pointer of the node to add aData to
* @param aKey the key to associate the object to
* @param aData the object to associate to aKey on aObject (may be nulll)
* @param aHandler the UserDataHandler to call when the node is
* cloned/deleted/imported/renamed (may be nulll)
* @param aResult [out] the previously registered object for aKey on aObject,
* if any
* @return whether adding the object and UserDataHandler succeeded
*/
virtual nsresult SetUserData(const nsINode *aObject, nsIAtom *aKey,
nsIVariant *aData,
nsIDOMUserDataHandler *aHandler,
nsIVariant **aResult) = 0;
/**
* Get the object associated to aKey on node aObject. This will return NS_OK
* if no object was associated to aKey on aObject. Should only be used to
* implement the DOM Level 3 UserData API.
*
* @param aObject canonical nsINode pointer of the node to get the
* object for
* @param aKey the key the object is associated to
* @param aResult [out] the registered object for aKey on aObject, if any
* @return whether an error occured while looking up the registered object
*/
nsresult GetUserData(nsINode *aObject, const nsAString &aKey,
nsIVariant **aResult)
{
nsCOMPtr<nsIAtom> key = do_GetAtom(aKey);
if (!key) {
return NS_ERROR_OUT_OF_MEMORY;
}
return GetUserData(aObject, key, aResult);
}
/**
* Get the object associated to aKey on node aObject. This will return NS_OK
* if no object was associated to aKey on aObject. Should only be used to
* implement the DOM Level 3 UserData API.
*
* @param aObject canonical nsINode pointer of the node to get the
* object for
* @param aKey the key the object is associated to
* @param aResult [out] the registered object for aKey on aObject, if any
* @return whether an error occured while looking up the registered object
*/
virtual nsresult GetUserData(const nsINode *aObject, nsIAtom *aKey,
nsIVariant **aResult) = 0;
/**
* Call the UserDataHandler associated with aKey on node aObject. Should only
* be used to implement the DOM Level 3 UserData API.
*
* @param aOperation the type of operation that is being performed on the
* node. @see nsIDOMUserDataHandler
* @param aObject canonical nsINode pointer of the node to call the
* UserDataHandler for
* @param aSource the node that aOperation is being performed on, or null if
* the operation is a deletion
* @param aDest the newly created node if any, or null
*/
virtual void CallUserDataHandler(PRUint16 aOperation,
const nsINode *aObject,
nsIDOMNode *aSource, nsIDOMNode *aDest) = 0;
/**
* Copy the objects and UserDataHandlers for node aObject to a new document.
* Should only be used to implement the DOM Level 3 UserData API.
*
* @param aObject canonical nsINode pointer of the node to copy objects
* and UserDataHandlers for
* @param aDestination the new document
*/
virtual void CopyUserData(const nsINode *aObject,
nsIDocument *aDestination) = 0;
protected:
~nsIDocument()
{
// XXX The cleanup of mNodeInfoManager (calling DropDocumentReference and
// releasing it) happens in the nsDocument destructor. We'd prefer to
// do it here but nsNodeInfoManager is a concrete class that we don't
// want to expose to users of the nsIDocument API outside of Gecko.
}
nsString mDocumentTitle;
nsCOMPtr<nsIURI> mDocumentURI;
nsCOMPtr<nsIURI> mDocumentBaseURI;
nsWeakPtr mDocumentLoadGroup;
nsWeakPtr mDocumentContainer;
nsCString mCharacterSet;
PRInt32 mCharacterSetSource;
// This is just a weak pointer; the parent document owns its children.
nsIDocument* mParentDocument;
// A weak reference to the only child element, or null if no
// such element exists.
nsIContent* mRootContent;
nsCOMPtr<nsIBindingManager> mBindingManager;
nsNodeInfoManager* mNodeInfoManager; // [STRONG]
nsICSSLoader* mCSSLoader; // [STRONG; not a COMPtr to avoid
// including nsICSSLoader.h; the ownership
// is managed by nsDocument]
// Table of element properties for this document.
nsPropertyTable mPropertyTable;
// True if BIDI is enabled.
PRBool mBidiEnabled;
// The bidi options for this document. What this bitfield means is
// defined in nsBidiUtils.h
PRUint32 mBidiOptions;
nsCString mContentLanguage;
nsCString mContentType;
// The document's security info
nsCOMPtr<nsISupports> mSecurityInfo;
// if this document is part of a multipart document,
// the ID can be used to distinguish it from the other parts.
PRUint32 mPartID;
};
NS_DEFINE_STATIC_IID_ACCESSOR(nsIDocument, NS_IDOCUMENT_IID)
/**
* Helper class to automatically handle batching of document updates. This
* class will call BeginUpdate on construction and EndUpdate on destruction on
* the given document with the given update type. The document could be null,
* in which case no updates will be called. The constructor also takes a
* boolean that can be set to false to prevent notifications.
*/
class mozAutoDocUpdate
{
public:
mozAutoDocUpdate(nsIDocument* aDocument, nsUpdateType aUpdateType,
PRBool aNotify) :
mDocument(aNotify ? aDocument : nsnull),
mUpdateType(aUpdateType)
{
if (mDocument) {
mDocument->BeginUpdate(mUpdateType);
}
}
~mozAutoDocUpdate()
{
if (mDocument) {
mDocument->EndUpdate(mUpdateType);
}
}
private:
nsCOMPtr<nsIDocument> mDocument;
nsUpdateType mUpdateType;
};
#define MOZ_AUTO_DOC_UPDATE_PASTE2(tok,line) tok##line
#define MOZ_AUTO_DOC_UPDATE_PASTE(tok,line) \
MOZ_AUTO_DOC_UPDATE_PASTE2(tok,line)
#define MOZ_AUTO_DOC_UPDATE(doc,type,notify) \
mozAutoDocUpdate MOZ_AUTO_DOC_UPDATE_PASTE(_autoDocUpdater_, __LINE__) \
(doc,type,notify)
// XXX These belong somewhere else
nsresult
NS_NewHTMLDocument(nsIDocument** aInstancePtrResult);
nsresult
NS_NewXMLDocument(nsIDocument** aInstancePtrResult);
#ifdef MOZ_SVG
nsresult
NS_NewSVGDocument(nsIDocument** aInstancePtrResult);
#endif
nsresult
NS_NewImageDocument(nsIDocument** aInstancePtrResult);
nsresult
NS_NewDocumentFragment(nsIDOMDocumentFragment** aInstancePtrResult,
nsNodeInfoManager *aNodeInfoManager);
// Note: it's the caller's responsibility to create or get aPrincipal as needed
// -- this method will not attempt to get a principal based on aDocumentURI.
// Also, both aDocumentURI and aBaseURI must not be null.
nsresult
NS_NewDOMDocument(nsIDOMDocument** aInstancePtrResult,
const nsAString& aNamespaceURI,
const nsAString& aQualifiedName,
nsIDOMDocumentType* aDoctype,
nsIURI* aDocumentURI,
nsIURI* aBaseURI,
nsIPrincipal* aPrincipal);
nsresult
NS_NewPluginDocument(nsIDocument** aInstancePtrResult);
#endif /* nsIDocument_h___ */