public WebBrowserDocument(string url, IHTMLDocument2 msDoc)
{
_domModifications = new IDomModification[]
{
new RemoveOrphanElements(),
new RemoveUnknownElements(),
new RemoveCommentElements(),
new RemoveEmptyFreeTextNodes(),
new ReplaceFreeTextNodes(),
new ReplacePreservedWhitespace(),
new ReplaceFreeTextNodes(),
};
_elementClassifier = new PreZoningClassification(HtmlElementType.SignificantBlock, HtmlElementType.SignificantInline, HtmlElementType.SignificantLinebreak, HtmlElementType.SignificantInvisible, HtmlElementType.BreakDown, HtmlElementType.Aname, HtmlElementType.Hidden);
_zoneTreeBuilder = ZoneTreeBuilder.Create(HtmlElementType.SignificantBlock, HtmlElementType.SignificantInline, HtmlElementType.SignificantLinebreak, HtmlElementType.SignificantInvisible, HtmlElementType.BreakDown, HtmlElementType.Aname, HtmlElementType.Hidden);
var defaultStyleLookup = DefaultStyleLookup.CreateForInternetExplorer();
_documentFactory = new MsHtmlDocumentFactory(defaultStyleLookup);
_url = url;
_msDoc = msDoc;
_highlights = new List <HtmlHighlight>();
_domModificationsApplied = false;
}
/// <summary>
/// Initializes a new instance of the <see cref="AppServices"/> class
/// </summary>
public AppServices()
{
var defaultStyleLookup = DefaultStyleLookup.CreateForInternetExplorer();
bool validateSerializationResult = false;
_htmlDocumentSerializer = new HtmlDocumentSerializer(defaultStyleLookup, validateSerializationResult);
_zoneTreeSerializer = new ZoneTreeSerializer(defaultStyleLookup);
_elementClassifier = new PreZoningClassification(HtmlElementType.SignificantBlock, HtmlElementType.SignificantInline, HtmlElementType.SignificantLinebreak, HtmlElementType.SignificantInvisible, HtmlElementType.BreakDown, HtmlElementType.Aname, HtmlElementType.Hidden);
_zoneTreeBuilder = ZoneTreeBuilder.Create(HtmlElementType.SignificantBlock, HtmlElementType.SignificantInline, HtmlElementType.SignificantLinebreak, HtmlElementType.SignificantInvisible, HtmlElementType.BreakDown, HtmlElementType.Aname, HtmlElementType.Hidden);
_columnTreeBuilder = ColumnTreeBuilder.Create();
var naturalLanguageProcessor = new OpenNaturalLanguageProcessor();
_layoutAnalysisArticleContentLabeler = new ArticleContentLabeler(naturalLanguageProcessor, ZoneLabel.Paragraph, ZoneLabel.ArticleContent, ZoneFeature.Common_Tokens);
_articleTagArticleContentLabeler = new SemanticTagArticleContentLabeler(Html.Tags.ARTICLE, ZoneLabel.ArticleContent);
_mainTagArticleContentLabeler = new SemanticTagArticleContentLabeler(Html.Tags.MAIN, ZoneLabel.ArticleContent);
}
/// <summary>
/// Initializes the <see cref="HtmlDocumentSerializer" /> class
/// </summary>
/// <param name="defaultStyleLookup">The default style lookup</param>
/// <param name="validateResult">Whether to validate the serialization result</param>
public HtmlDocumentSerializer(DefaultStyleLookup defaultStyleLookup, bool validateResult = false)
{
_serializer = new XmlSerializer(typeof(SerializableElementList));
_writerSettings = new XmlWriterSettings
{
Indent = true,
CheckCharacters = false,
NewLineHandling = NewLineHandling.Entitize,
};
_readerSettings = new XmlReaderSettings
{
CheckCharacters = false,
IgnoreWhitespace = true,
};
_namespaces = new XmlSerializerNamespaces(new[] { XmlQualifiedName.Empty });
_defaultStyleLookup = defaultStyleLookup;
_validateResult = validateResult;
}
/// <summary>
/// Initializes the <see cref="HtmlDocumentSerializer" /> class
/// </summary>
public HtmlDocumentSerializer()
: this(DefaultStyleLookup.CreateForInternetExplorer())
{
}
/// <summary>
/// Initializes a new instance of the <see cref="SerializableDocument" /> class
/// </summary>
/// <param name="root">The root element</param>
/// <param name="info">The HTML Document information</param>
/// <param name="defaultStyleLookup">The default style lookup</param>
/// <param name="html">The html</param>
/// <param name="text">The text</param>
public SerializableDocument(SerializableElement root, HtmlDocumentInfo info, DefaultStyleLookup defaultStyleLookup, string html, string text)
: base(root, info, defaultStyleLookup)
{
this.Html = html;
this.Text = text;
}
/// <summary>
/// Initializes a new instance of the <see cref="WebBrowser" /> class
/// </summary>
/// <param name="size">The size</param>
/// <param name="defaultStyles">The default styles</param>
public WebBrowser(Size size, DefaultStyleLookup defaultStyles)
{
_size = size;
this.DefaultStyles = defaultStyles;
this.RenderingComplete = false;
}
/// <summary>
/// Initializes a new instance of the <see cref="MsHtmlElementFactory"/> class
/// </summary>
/// <param name="sourceIndexOffset">The source index offset</param>
/// <param name="defaultStyleLookup">The default style lookup</param>
public MsHtmlElementFactory(int sourceIndexOffset, DefaultStyleLookup defaultStyleLookup)
{
_sourceIndexOffset = sourceIndexOffset;
_defaultStyleLookup = defaultStyleLookup;
}
/// <summary>
/// Initializes a new instance of the <see cref="SerializableElement" /> class
/// </summary>
/// <param name="id">The id</param>
/// <param name="displayOrder">The display order</param>
/// <param name="parentId">The parent id</param>
/// <param name="childrenIds">The children ids</param>
/// <param name="boundingBox">The bounding box</param>
/// <param name="classifications">The classifications</param>
/// <param name="tagName">The tag name</param>
/// <param name="attributes">The attributes</param>
/// <param name="styles">The styles</param>
/// <param name="defaultStyleLookup">The default style lookup</param>
/// <param name="outerHtmlStartPos">The outer html start position</param>
/// <param name="outerHtmlEndPos">The outer html end position</param>
/// <param name="innerHtmlStartPos">The inner html start position</param>
/// <param name="innerHtmlEndPos">The inner html end position</param>
/// <param name="textStartPos">The text start position</param>
/// <param name="textEndPos">The text end position</param>
public SerializableElement(int id, int displayOrder, int?parentId, IReadOnlyList <int> childrenIds, Rectangle boundingBox, IEnumerable <string> classifications, string tagName,
IDictionary <string, string> attributes, IDictionary <string, string> styles, DefaultStyleLookup defaultStyleLookup,
int outerHtmlStartPos, int outerHtmlEndPos,
int innerHtmlStartPos, int innerHtmlEndPos,
int textStartPos, int textEndPos)
{
this.Initialize(id, displayOrder, parentId, childrenIds, boundingBox, classifications,
tagName, attributes, styles, defaultStyleLookup,
outerHtmlStartPos, outerHtmlEndPos,
innerHtmlStartPos, innerHtmlEndPos,
textStartPos, textEndPos);
}
/// <summary>
/// Initializes a new instance of the <see cref="MsHtmlDocument" /> class
/// </summary>
/// <param name="root">The root element</param>
/// <param name="info">The Html Document information</param>
/// <param name="defaultStyleLookup">The default style lookup</param>
public MsHtmlDocument(MsHtmlElement root, HtmlDocumentInfo info, DefaultStyleLookup defaultStyleLookup)
: base(root, info, defaultStyleLookup)
{
}
/// <summary>
/// Initializes a new instance of the <see cref="MsHtmlDocumentFactory"/> class
/// </summary>
/// <param name="defaultStyleLookup">The default style lookup</param>
public MsHtmlDocumentFactory(DefaultStyleLookup defaultStyleLookup)
{
_defaultStyleLookup = defaultStyleLookup;
}
/// <summary>
/// Initializes a new instance of the <see cref="InternetExplorer" /> class
/// </summary>
/// <param name="winFormsBrowser">The windows forms browser</param>
/// <param name="size">The size</param>
/// <param name="defaultStyles">The default styles</param>
public InternetExplorer(System.Windows.Forms.WebBrowser winFormsBrowser, Size size, DefaultStyleLookup defaultStyles)
: base(size, defaultStyles)
{
_winFormsBrowser = winFormsBrowser;
_msDocFactory = new MsHtmlDocumentFactory(defaultStyles);
this.ConfigureBrowser();
}
/// <summary>
/// Initializes a new instance of the <see cref="InternetExplorer" /> class
/// </summary>
/// <param name="winFormsBrowser">The windows forms browser</param>
/// <param name="size">The size</param>
public InternetExplorer(System.Windows.Forms.WebBrowser winFormsBrowser, Size size)
: this(winFormsBrowser, size, DefaultStyleLookup.CreateForInternetExplorer())
{
}
/// <summary>
/// Initializes a new instance of the <see cref="SerializableElement" /> class
/// </summary>
/// <param name="element">The html element</param>
/// <param name="outerHtmlStartPos">The outer html start position</param>
/// <param name="outerHtmlEndPos">The outer html end position</param>
/// <param name="innerHtmlStartPos">The inner html start position</param>
/// <param name="innerHtmlEndPos">The inner html end position</param>
/// <param name="textStartPos">The text start position</param>
/// <param name="textEndPos">The text end position</param>
/// <param name="defaultStyleLookup">The default style lookup</param>
public SerializableElement(HtmlElement element, int outerHtmlStartPos, int outerHtmlEndPos, int innerHtmlStartPos, int innerHtmlEndPos, int textStartPos, int textEndPos, DefaultStyleLookup defaultStyleLookup)
{
this.Initialize(element, outerHtmlStartPos, outerHtmlEndPos, innerHtmlStartPos, innerHtmlEndPos, textStartPos, textEndPos, defaultStyleLookup);
}
/// <summary>
/// Initializes the serializable element
/// </summary>
/// <param name="element">The html element</param>
/// <param name="outerHtmlStartPos">The outer html start position</param>
/// <param name="outerHtmlEndPos">The outer html end position</param>
/// <param name="innerHtmlStartPos">The inner html start position</param>
/// <param name="innerHtmlEndPos">The inner html end position</param>
/// <param name="textStartPos">The text start position</param>
/// <param name="textEndPos">The text end position</param>
/// <param name="defaultStyleLookup">The default style lookup</param>
public void Initialize(HtmlElement element, int outerHtmlStartPos, int outerHtmlEndPos, int innerHtmlStartPos, int innerHtmlEndPos, int textStartPos, int textEndPos, DefaultStyleLookup defaultStyleLookup)
{
var attributeDictionary = ConstructAttributeDictionary(element);
var styleDictionary = ConstructStyleDictionary(element);
this.Initialize(element.Id, element.DisplayOrder, element.ParentId, element.ChildrenIds, element.BoundingBox, element.Classifications,
element.TagName, attributeDictionary, styleDictionary, defaultStyleLookup,
outerHtmlStartPos, outerHtmlEndPos,
innerHtmlStartPos, innerHtmlEndPos,
textStartPos, textEndPos);
}
/// <summary>
/// Sets the default style lookup
/// </summary>
/// <param name="defaultStyleLookup">The default style lookup</param>
public new void SetDefaultStyleLookup(DefaultStyleLookup defaultStyleLookup)
{
base.SetDefaultStyleLookup(defaultStyleLookup);
}
/// <summary>
/// Initializes a new instance of the <see cref="SerializableElement" /> class
/// For testing purposes
/// </summary>
/// <param name="id">The id</param>
/// <param name="displayOrder">The display order</param>
/// <param name="parentId">The parent id</param>
/// <param name="childrenIds">The children ids</param>
/// <param name="boundingBox">The bounding box</param>
/// <param name="classifications">The classifications</param>
/// <param name="tagName">Tag name</param>
/// <param name="outerHtml">The outer html</param>
/// <param name="innerHtml">The inner html</param>
/// <param name="outerText">The outer text</param>
/// <param name="attributes">The attributes</param>
/// <param name="styles">The styles</param>
/// <param name="defaultStyleLookup">The default style lookup</param>
public SerializableElement(int id, int displayOrder, int?parentId, IReadOnlyList <int> childrenIds, Rectangle boundingBox, IEnumerable <string> classifications,
string tagName, string outerHtml, string innerHtml, string outerText,
IDictionary <string, string> attributes, IDictionary <string, string> styles, DefaultStyleLookup defaultStyleLookup)
{
base.Initialize(id, displayOrder, parentId, childrenIds, boundingBox, classifications, tagName, outerHtml, innerHtml, outerText, attributes, styles, defaultStyleLookup);
}
/// <summary>
/// Initializes a new instance of the <see cref="HtmlElement" /> class
/// </summary>
/// <param name="id">The id</param>
/// <param name="displayOrder">The display order</param>
/// <param name="parentId">The parent id</param>
/// <param name="childrenIds">The children ids</param>
/// <param name="classifications">The classifications</param>
/// <param name="tagName">The tag name</param>
/// <param name="outerHtml">The outer html</param>
/// <param name="innerHtml">The inner html</param>
/// <param name="text">The text</param>
/// <param name="attributes">The attributes</param>
/// <param name="styles">The styles</param>
/// <param name="defaultStyleLookup">The default style lookup</param>
/// <param name="offsetParentId">The offset parent id</param>
/// <param name="offsetRectangle">The offset rectangle</param>
public MsHtmlElement(int id, int displayOrder, int?parentId, IReadOnlyList <int> childrenIds, IEnumerable <string> classifications,
string tagName, string outerHtml, string innerHtml, string text,
IDictionary <string, string> attributes, IDictionary <string, string> styles, DefaultStyleLookup defaultStyleLookup,
int?offsetParentId, Rectangle offsetRectangle)
: base(id, id, parentId, childrenIds, Rectangle.Empty, classifications, tagName, outerHtml, innerHtml, text, attributes, styles, defaultStyleLookup)
{
_offsetParentId = offsetParentId;
_offsetRectangle = offsetRectangle;
}
/// <summary>
/// Creates the serializable elements
/// </summary>
/// <param name="element">The original element</param>
/// <param name="sElements">All serializable element</param>
/// <param name="htmlBuilder">The HTML builder</param>
/// <param name="textBuilder">The text builder</param>
/// <param name="defaultStyleLookup">The default style lookup</param>
public void CreateElements(HtmlElement element, List <SerializableElement> sElements, StringBuilder htmlBuilder, StringBuilder textBuilder, DefaultStyleLookup defaultStyleLookup)
{
// Create element
var sElement = new SerializableElement();
sElements.Add(sElement);
// The variables need to be set
int outerHtmlStartPos;
int outerHtmlEndPos;
int innerHtmlStartPos;
int innerHtmlEndPos;
int textStartPos;
int textEndPos;
// Record outer start positions
outerHtmlStartPos = htmlBuilder.Length;
// Add start tag & then record inner html start position
htmlBuilder.Append(element.StartTag);
innerHtmlStartPos = htmlBuilder.Length;
// Add start text
string startText;
string endText;
GetElementText(element, out startText, out endText);
textBuilder.Append(startText);
textStartPos = textBuilder.Length;
// Add inner html/text
if (element.ChildrenCount > 0)
{
foreach (HtmlElement child in element.Children)
{
CreateElements(child, sElements, htmlBuilder, textBuilder, defaultStyleLookup);
}
}
else
{
htmlBuilder.Append(element.InnerHtml);
textBuilder.Append(element.OuterText);
}
// Add end tag
innerHtmlEndPos = htmlBuilder.Length;
htmlBuilder.Append(element.EndTag);
outerHtmlEndPos = htmlBuilder.Length;
// Add end text
textEndPos = textBuilder.Length;
textBuilder.Append(endText);
// Initalize element
sElement.Initialize(element, outerHtmlStartPos, outerHtmlEndPos, innerHtmlStartPos, innerHtmlEndPos, textStartPos, textEndPos, defaultStyleLookup);
}
/// <summary>
/// Initializes a new instance of the <see cref="MsHtmlDocument" /> class
/// </summary>
/// <param name="root">The root element</param>
/// <param name="info">The Html Document information</param>
/// <param name="defaultStyleLookup">The default style lookup</param>
/// <param name="displayHtml">The display HTML</param>
public MsHtmlDocument(MsHtmlElement root, HtmlDocumentInfo info, DefaultStyleLookup defaultStyleLookup, string displayHtml)
: base(root, info, defaultStyleLookup)
{
this.DisplayHtml = displayHtml;
}
/// <summary>
/// Initializes a new instance of the <see cref="DisplayHiddenFilter"/> class
/// </summary>
/// <param name="defaultStyleLookup">The default style lookup</param>
public DisplayHiddenFilter(DefaultStyleLookup defaultStyleLookup)
{
_defaultStyleLookup = defaultStyleLookup;
}
/// <summary>
/// Initializes a new instance of the <see cref="HtmlRenderer"/> class
/// </summary>
/// <param name="defaultStyles">The default styles</param>
public HtmlRenderer(DefaultStyleLookup defaultStyles)
{
this.DefaultStyles = defaultStyles;
}
/// <summary>
/// Initializes the serializable element
/// </summary>
/// <param name="id">The id</param>
/// <param name="displayOrder">The display order</param>
/// <param name="parentId">The parent id</param>
/// <param name="childrenIds">The children ids</param>
/// <param name="boundingBox">The bounding box</param>
/// <param name="classifications">The classifications</param>
/// <param name="tagName">The tag name</param>
/// <param name="attributes">The attributes</param>
/// <param name="styles">The styles</param>
/// <param name="deafultStyleLookup">The deafult style lookup</param>
/// <param name="outerHtmlStartPos">The outer html start position</param>
/// <param name="outerHtmlEndPos">The outer html end position</param>
/// <param name="innerHtmlStartPos">The inner html start position</param>
/// <param name="innerHtmlEndPos">The inner html end position</param>
/// <param name="textStartPos">The text start position</param>
/// <param name="textEndPos">The text end position</param>
public void Initialize(int id, int displayOrder, int?parentId, IReadOnlyList <int> childrenIds, Rectangle boundingBox, IEnumerable <string> classifications, string tagName,
IDictionary <string, string> attributes, IDictionary <string, string> styles, DefaultStyleLookup deafultStyleLookup,
int outerHtmlStartPos, int outerHtmlEndPos,
int innerHtmlStartPos, int innerHtmlEndPos,
int textStartPos, int textEndPos)
{
base.Initialize(id, displayOrder, parentId, childrenIds, boundingBox, classifications, tagName, string.Empty, string.Empty, string.Empty, attributes, styles, deafultStyleLookup);
this.OuterHtmlStartPos = outerHtmlStartPos;
this.OuterHtmlEndPos = outerHtmlEndPos;
this.InnerHtmlStartPos = innerHtmlStartPos;
this.InnerHtmlEndPos = innerHtmlEndPos;
this.TextStartPos = textStartPos;
this.TextEndPos = textEndPos;
}
