Пример #1
0
        private void TagEnd(string sTagName)
        {
            if (_normalizeXml)
            {
                HtmlTagType tagType = HtmlTags.GetHtmlTagType(sTagName);
                switch (tagType)
                {
                case HtmlTagType.Html:
                case HtmlTagType.Head:
                case HtmlTagType.Body:
                    return;

                case HtmlTagType.Title:
                    _currentNode = _currentTreeNode;
                    return;

                case HtmlTagType.Table:
                    if (_table == null)
                    {
                        return;
                    }
                    _currentNode = _currentTreeNode = GetParentXXNode(_table.Table);
                    _table       = null;
                    if (_tableStack.Count != 0)
                    {
                        _table = _tableStack.Pop();
                    }
                    return;

                case HtmlTagType.DL:
                    if (_definitionList == null)
                    {
                        return;
                    }
                    _currentNode    = _currentTreeNode = GetParentXXNode(_definitionList);
                    _definitionList = null;
                    if (_definitionListStack.Count != 0)
                    {
                        _definitionList = _definitionListStack.Pop();
                    }
                    return;
                }
                if (_table != null)
                {
                    switch (tagType)
                    {
                    case HtmlTagType.THead:
                    case HtmlTagType.TBody:
                    case HtmlTagType.TFoot:
                        _currentNode = _currentTreeNode = _table.Table;
                        _table.Body  = null;
                        return;

                    case HtmlTagType.ColGroup:
                        _currentNode    = _currentTreeNode = _table.Table;
                        _table.ColGroup = null;
                        return;

                    case HtmlTagType.Col:
                        if (_table.Col != null)
                        {
                            _currentNode = _currentTreeNode = GetParentXXNode(_table.Col);
                            _table.Col   = null;
                        }
                        return;

                    case HtmlTagType.TR:
                        if (_table.Row != null)
                        {
                            _currentNode = _currentTreeNode = GetParentXXNode(_table.Row);
                            _table.Row   = null;
                        }
                        return;

                    case HtmlTagType.TH:
                    case HtmlTagType.TD:
                        if (_table.Data != null)
                        {
                            _currentNode = _currentTreeNode = GetParentXXNode(_table.Data);
                            _table.Data  = null;
                        }
                        return;
                    }
                }
            }
            XXXNode_v2 node = GetParentXXNodeByName(_currentTreeNode, sTagName);

            if (node != null)
            {
                _currentTreeNode = GetParentXXNode(node);
            }
            _currentNode = _currentTreeNode;
        }
Пример #2
0
        private void TagBegin(string tagName, bool tagEnd)
        {
            _noTag = false;
            HtmlTagType tagType = HtmlTags.GetHtmlTagType(tagName);
            HtmlTag     tag     = HtmlTags.GetHtmlTag(tagType);

            if (_normalizeXml)
            {
                if (tagType == HtmlTagType.Html || tagType == HtmlTagType.Head)
                {
                    _noTag = true;
                    return;
                }
                if (tagType == HtmlTagType.Body)
                {
                    _noTag = true;
                    if (!_body)
                    {
                        _body        = true;
                        _currentNode = _currentTreeNode = _bodyNode;
                    }
                    return;
                }
                if (tagType == HtmlTagType.Title)
                {
                    if (!_title)
                    {
                        if (!tagEnd)
                        {
                            _title       = true;
                            _currentNode = _titleNode;
                        }
                    }
                    else
                    {
                        _noTag = true;
                    }
                    return;
                }
                // $$pb modif le 11/01/2015
                //if (!_body && tag.TagCategory != HtmlTagCategory.Head)
                //{
                //    _body = true;
                //    _currentNode = _currentTreeNode = _bodyNode;
                //}
            }
            _currentNode = CreateElement(tagName);
            if (_normalizeXml)
            {
                if (tagType == HtmlTagType.Table && !tagEnd)
                {
                    if (_table != null)
                    {
                        _tableStack.Push(_table);
                    }
                    _table       = new HtmlTable_v2();
                    _table.Table = _currentNode;
                    AddElement(_currentTreeNode, _currentNode);
                    _currentTreeNode = _currentNode;
                    return;
                }
                if (TagBeginTableCategory(tag, tagEnd))
                {
                    return;
                }
                if (tagType == HtmlTagType.DL && !tagEnd)
                {
                    if (_definitionList != null)
                    {
                        _definitionListStack.Push(_definitionList);
                    }
                    _definitionList = _currentNode;
                    AddElement(_currentTreeNode, _currentNode);
                    _currentTreeNode = _currentNode;
                    return;
                }
                if (TagBeginDefinitionListCategory(tag, tagEnd))
                {
                    return;
                }

                // $$pb à revérifier
                // il faut au moins annuler gLastPNode quand un des parents de gLastPNode se ferme
                //if (tagType == HtmlTagTypeEnum.P)
                //{
                //    // pour gérer une balise <p> qui n'a pas de fin de balise </p>
                //    if (gLastPNode != null)
                //    {
                //        gCurrentTreeNode = GetParentXXNode(gLastPNode);
                //        gLastPNode = null;
                //    }
                //    if (!bTagEnd) gLastPNode = gCurrentNode;
                //}
            }
            AddElement(_currentTreeNode, _currentNode);
            //if (!tagEnd && tag.EndBoundType != HtmlBoundType.Forbidden)
            //    _currentTreeNode = _currentNode;

            if (!tagEnd)
            {
                if (tag.EndBoundType != HtmlBoundType.Forbidden)
                {
                    _currentTreeNode = _currentNode;
                }
            }
            else if (_correctionMarkBeginEnd)
            {
                _currentNode = _currentTreeNode;
            }
        }
Пример #3
0
        private void GenerateXml()
        {
            // gbNormalizeXml = true :
            //   - les tag html, head, title et body sont créés automatiquement
            //   - les tag html, head, title et body rencontrés ne sont pas pris en compte
            //   - seul les tag title et meta sont mis dans la partie head les autre tag sont mis dans la partie body
            //   - si un tag meta est placé après le début de la partie body, ce tag reste dans la partie body
            //   - seul le premier tag title est pris en compte et placé dans la partie head, les autre tag title ne sont pas pris en compte

            try
            {
                //cTrace.Trace("GenerateXml NewGenerateXml  : {0}", XmlConfig.CurrentConfig.Get("NewGenerateXml"));

                InitXml();

                _tableStack = new Stack <HtmlTable_v2>();
                _table      = null;

                _definitionListStack = new Stack <XXXNode_v2>();
                _definitionList      = null;

                _noTag = false;
                _body  = false;
                _title = false;
                while (_htmlReader.Read())
                {
                    if (_htmlReader.IsText || _htmlReader.IsComment)
                    {
                        // $$pb modif le 11/01/2015
                        //if (_htmlReader.IsText && !_htmlReader.IsTextSeparator && !_body)
                        //if (_htmlReader.IsText && !_htmlReader.IsTextSeparator && !_htmlReader.IsScript && !_body)
                        //{
                        //    _body = true;
                        //    _currentNode = _currentTreeNode = _bodyNode;
                        //}
                        if (!_generateXmlNodeOnly)
                        {
                            if (_readCommentInText)
                            {
                                AddText(_currentNode, _htmlReader.Value);
                            }
                            else
                            {
                                if (_htmlReader.IsText)
                                {
                                    AddText(_currentNode, _htmlReader.Value);
                                }
                                else
                                {
                                    string s = _htmlReader.Value;
                                    s = _commentCorrection.Replace(s, "-");
                                    if (s.EndsWith("-"))
                                    {
                                        s += " ";
                                    }
                                    AddComment(_currentNode, s);
                                }
                            }
                        }
                    }
                    else if (_htmlReader.IsDocType)
                    {
                        AddAttribute(_htmlNode, "doctype", _htmlReader.DocType);
                    }
                    else if (_htmlReader.IsProperty)
                    {
                        if (_generateXmlNodeOnly || _noTag)
                        {
                            continue;
                        }
                        try
                        {
                            string sPropertyName = _htmlReader.PropertyName;
                            //sPropertyName = sPropertyName.Replace("\"", "");
                            //sPropertyName = sPropertyName.Replace("/", "");
                            //sPropertyName = sPropertyName.Replace("\\", "");
                            //sPropertyName = sPropertyName.Replace("-", "");
                            //sPropertyName = sPropertyName.Replace(",", "");
                            sPropertyName = _replace.Replace(sPropertyName, "");
                            sPropertyName = sPropertyName.ToLower();
                            if (sPropertyName == "")
                            {
                                sPropertyName = "__value";
                            }

                            // modif le 28/01/2014
                            //   hexadecimal value 0x03, is an invalid character
                            //   found in http://www.reseau-gesat.com/Gesat/Yvelines,78/Fontenay-le-Fleury,31443/esat-cotra,e1596/
                            //   <html><head><meta name="keywords" content="Conditionnement, travaux &amp;agrave; fa&amp;ccedil;onToutes activit&amp;eacute;s en entreprise Entretien et cr&amp;eacute;ation despaces verts" />
                            string propertyValue = _htmlReader.PropertyValue;
                            if (propertyValue != null)
                            {
                                propertyValue = propertyValue.Replace("\x03", "");
                            }
                            AddAttribute(_currentNode, sPropertyName, propertyValue);
                            if (_htmlReader.IsMarkBeginEnd)
                            {
                                TagEnd(_htmlReader.MarkName.ToLower());
                            }
                        }
                        catch (Exception ex)
                        {
                            Trace.WriteLine("error in HtmlToXml.GenerateXml() : line {0} column {1}", _htmlReader.Line, _htmlReader.Column);
                            Trace.WriteLine(ex.Message);
                        }
                    }
                    else if (_htmlReader.IsMarkBeginEnd)
                    {
                        string sTagName = _htmlReader.MarkName.ToLower();
                        sTagName = _replace.Replace(sTagName, "_");
                        if (sTagName == "")
                        {
                            sTagName = "_";
                        }

                        TagBegin(sTagName, true);
                    }
                    else if (_htmlReader.IsMarkBegin)
                    {
                        string sTagName = _htmlReader.MarkName.ToLower();
                        sTagName = _replace.Replace(sTagName, "_");
                        if (sTagName == "")
                        {
                            sTagName = "_";
                        }

                        TagBegin(sTagName, false);
                    }
                    else if (_htmlReader.IsMarkEnd)
                    {
                        string sTagName = _htmlReader.MarkName.ToLower();
                        sTagName = _replace.Replace(sTagName, "_");
                        if (sTagName == "")
                        {
                            sTagName = "_";
                        }

                        TagEnd(sTagName);
                    }
                }
            }
            finally
            {
                _htmlReader.Close();
            }
        }
Пример #4
0
        private void TagEnd(string sTagName)
        {
            if (gbNormalizeXml)
            {
                HtmlTagType tagType = HtmlTags.GetHtmlTagType(sTagName);
                switch (tagType)
                {
                case HtmlTagType.Html:
                case HtmlTagType.Head:
                case HtmlTagType.Body:
                    return;

                case HtmlTagType.Title:
                    gCurrentNode = gCurrentTreeNode;
                    return;

                case HtmlTagType.Table:
                    if (gTable == null)
                    {
                        return;
                    }
                    //gCurrentNode = gCurrentTreeNode = gTable.Table.ParentNode;
                    gCurrentNode = gCurrentTreeNode = GetParentXXNode(gTable.Table);
                    gTable       = null;
                    if (gTableStack.Count != 0)
                    {
                        gTable = gTableStack.Pop();
                    }
                    return;

                case HtmlTagType.DL:
                    if (gDefinitionList == null)
                    {
                        return;
                    }
                    //gCurrentNode = gCurrentTreeNode = gDefinitionList.ParentNode;
                    gCurrentNode    = gCurrentTreeNode = GetParentXXNode(gDefinitionList);
                    gDefinitionList = null;
                    if (gDefinitionListStack.Count != 0)
                    {
                        gDefinitionList = gDefinitionListStack.Pop();
                    }
                    return;
                }
                if (gTable != null)
                {
                    switch (tagType)
                    {
                    case HtmlTagType.THead:
                    case HtmlTagType.TBody:
                    case HtmlTagType.TFoot:
                        gCurrentNode = gCurrentTreeNode = gTable.Table;
                        gTable.Body  = null;
                        return;

                    case HtmlTagType.ColGroup:
                        gCurrentNode    = gCurrentTreeNode = gTable.Table;
                        gTable.ColGroup = null;
                        return;

                    case HtmlTagType.Col:
                        if (gTable.Col != null)
                        {
                            //gCurrentNode = gCurrentTreeNode = gTable.Col.ParentNode;
                            gCurrentNode = gCurrentTreeNode = GetParentXXNode(gTable.Col);
                            gTable.Col   = null;
                        }
                        return;

                    case HtmlTagType.TR:
                        if (gTable.Row != null)
                        {
                            //gCurrentNode = gCurrentTreeNode = gTable.Row.ParentNode;
                            gCurrentNode = gCurrentTreeNode = GetParentXXNode(gTable.Row);
                            gTable.Row   = null;
                        }
                        return;

                    case HtmlTagType.TH:
                    case HtmlTagType.TD:
                        if (gTable.Data != null)
                        {
                            //gCurrentNode = gCurrentTreeNode = gTable.Data.ParentNode;
                            gCurrentNode = gCurrentTreeNode = GetParentXXNode(gTable.Data);
                            gTable.Data  = null;
                        }
                        return;
                    }
                }
            }
            //XmlNode node = gCurrentTreeNode;
            //while (node != null)
            //{
            //    if (node.Name == sTagName)
            //    {
            //        gCurrentTreeNode = node.ParentNode;
            //        break;
            //    }
            //    node = node.ParentNode;
            //}
            XXXNode_v2 node = GetParentXXNodeByName(gCurrentTreeNode, sTagName);

            if (node != null)
            {
                gCurrentTreeNode = GetParentXXNode(node);
            }
            gCurrentNode = gCurrentTreeNode;
        }
Пример #5
0
        private void TagBegin(string sTagName, bool bTagEnd)
        {
            gbNoTag = false;

            //sTagName = sTagName.Replace('-', '_');
            //sTagName = sTagName.Replace('!', '_');
            //sTagName = sTagName.Replace('[', '_');
            //sTagName = sTagName.Replace(']', '_');
            //sTagName = gReplace.Replace(sTagName, "_");
            //if (sTagName == "") sTagName = "_";
            HtmlTagType tagType = HtmlTags.GetHtmlTagType(sTagName);
            HtmlTag     tag     = HtmlTags.GetHtmlTag(tagType);

            if (gbNormalizeXml)
            {
                if (tagType == HtmlTagType.Html || tagType == HtmlTagType.Head)
                {
                    gbNoTag = true;
                    return;
                }
                if (tagType == HtmlTagType.Body)
                {
                    gbNoTag = true;
                    if (!gbBody)
                    {
                        gbBody       = true;
                        gCurrentNode = gCurrentTreeNode = gBodyNode;
                    }
                    return;
                }
                if (tagType == HtmlTagType.Title)
                {
                    if (!gbTitle)
                    {
                        if (!bTagEnd)
                        {
                            gbTitle      = true;
                            gCurrentNode = gTitleNode;
                        }
                    }
                    else
                    {
                        gbNoTag = true;
                    }
                    return;
                }
                if (!gbBody && tag.TagCategory != HtmlTagCategory.Head)
                {
                    gbBody       = true;
                    gCurrentNode = gCurrentTreeNode = gBodyNode;
                }
            }
            //gCurrentNode = gXmlDocument.CreateElement(sTagName);
            gCurrentNode = CreateElement(sTagName);
            if (gbNormalizeXml)
            {
                if (tagType == HtmlTagType.Table && !bTagEnd)
                {
                    if (gTable != null)
                    {
                        gTableStack.Push(gTable);
                    }
                    gTable       = new HtmlTable_v2();
                    gTable.Table = gCurrentNode;
                    //gCurrentTreeNode.AppendChild(gCurrentNode);
                    AddElement(gCurrentTreeNode, gCurrentNode);
                    gCurrentTreeNode = gCurrentNode;
                    return;
                }
                if (TagBeginTableCategory(tag, bTagEnd))
                {
                    return;
                }
                if (tagType == HtmlTagType.DL && !bTagEnd)
                {
                    if (gDefinitionList != null)
                    {
                        gDefinitionListStack.Push(gDefinitionList);
                    }
                    gDefinitionList = gCurrentNode;
                    //gCurrentTreeNode.AppendChild(gCurrentNode);
                    AddElement(gCurrentTreeNode, gCurrentNode);
                    gCurrentTreeNode = gCurrentNode;
                    return;
                }
                if (TagBeginDefinitionListCategory(tag, bTagEnd))
                {
                    return;
                }

                // $$pb à revérifier
                // il faut au moins annuler gLastPNode quand un des parents de gLastPNode se ferme
                //if (tagType == HtmlTagTypeEnum.P)
                //{
                //    // pour gérer une balise <p> qui n'a pas de fin de balise </p>
                //    if (gLastPNode != null)
                //    {
                //        gCurrentTreeNode = GetParentXXNode(gLastPNode);
                //        gLastPNode = null;
                //    }
                //    if (!bTagEnd) gLastPNode = gCurrentNode;
                //}
            }
            //gCurrentTreeNode.AppendChild(gCurrentNode);
            AddElement(gCurrentTreeNode, gCurrentNode);
            if (!bTagEnd && tag.EndBoundType != HtmlBoundType.Forbidden)
            {
                gCurrentTreeNode = gCurrentNode;
            }
        }
Пример #6
0
        private void GenerateXml()
        {
            // gbNormalizeXml = true :
            //   - les tag html, head, title et body sont créés automatiquement
            //   - les tag html, head, title et body rencontrés ne sont pas pris en compte
            //   - seul les tag title et meta sont mis dans la partie head les autre tag sont mis dans la partie body
            //   - si un tag meta est placé après le début de la partie body, ce tag reste dans la partie body
            //   - seul le premier tag title est pris en compte et placé dans la partie head, les autre tag title ne sont pas pris en compte

            try
            {
                //cTrace.Trace("GenerateXml NewGenerateXml  : {0}", XmlConfig.CurrentConfig.Get("NewGenerateXml"));

                InitXml();

                gTableStack = new Stack <HtmlTable_v2>();
                gTable      = null;

                gDefinitionListStack = new Stack <XXXNode_v2>();
                gDefinitionList      = null;

                gbNoTag = false;
                gbBody  = false;
                gbTitle = false;
                while (gHTMLReader.Read())
                {
                    if (gHTMLReader.IsText || gHTMLReader.IsComment)
                    {
                        if (gHTMLReader.IsText && !gHTMLReader.IsTextSeparator && !gbBody)
                        {
                            gbBody       = true;
                            gCurrentNode = gCurrentTreeNode = gBodyNode;
                        }
                        if (!gbGenerateXmlNodeOnly)
                        {
                            //if (string.Compare(XmlConfig.CurrentConfig.Get("NewGenerateXml"), "true", true) != 0)
                            if (gbReadCommentInText)
                            {
                                //XmlText text = gXmlDocument.CreateTextNode("text");
                                //text.Value = gHTMLReader.Value;
                                //gCurrentNode.AppendChild(text);
                                AddText(gCurrentNode, gHTMLReader.Value);
                            }
                            else
                            {
                                if (gHTMLReader.IsText)
                                {
                                    //XmlText text = gXmlDocument.CreateTextNode("text");
                                    //text.Value = gHTMLReader.Value;
                                    //gCurrentNode.AppendChild(text);
                                    AddText(gCurrentNode, gHTMLReader.Value);
                                }
                                else
                                {
                                    string s = gHTMLReader.Value;
                                    s = gCommentCorrection.Replace(s, "-");
                                    if (s.EndsWith("-"))
                                    {
                                        s += " ";
                                    }
                                    //XmlComment comment = gXmlDocument.CreateComment(s);
                                    //gCurrentNode.AppendChild(comment);
                                    AddComment(gCurrentNode, s);
                                }
                            }
                        }
                    }
                    else if (gHTMLReader.IsDocType)
                    {
                        //XmlAttribute attrib = gXmlDocument.CreateAttribute("doctype");
                        //attrib.Value = gHTMLReader.DocType;
                        //gHtmlNode.Attributes.Append(attrib);
                        AddAttribute(gHtmlNode, "doctype", gHTMLReader.DocType);
                    }
                    else if (gHTMLReader.IsProperty)
                    {
                        if (gbGenerateXmlNodeOnly || gbNoTag)
                        {
                            continue;
                        }
                        try
                        {
                            string sPropertyName = gHTMLReader.PropertyName;
                            //sPropertyName = sPropertyName.Replace("\"", "");
                            //sPropertyName = sPropertyName.Replace("/", "");
                            //sPropertyName = sPropertyName.Replace("\\", "");
                            //sPropertyName = sPropertyName.Replace("-", "");
                            //sPropertyName = sPropertyName.Replace(",", "");
                            sPropertyName = gReplace.Replace(sPropertyName, "");
                            sPropertyName = sPropertyName.ToLower();
                            if (sPropertyName == "")
                            {
                                sPropertyName = "__value";
                            }
                            //XmlAttribute attrib = gXmlDocument.CreateAttribute(sPropertyName);
                            //attrib.Value = gHTMLReader.PropertyValue;
                            //gCurrentNode.Attributes.Append(attrib);


                            // modif le 28/01/2014
                            //   hexadecimal value 0x03, is an invalid character
                            //   found in http://www.reseau-gesat.com/Gesat/Yvelines,78/Fontenay-le-Fleury,31443/esat-cotra,e1596/
                            //   <html><head><meta name="keywords" content="Conditionnement, travaux &amp;agrave; fa&amp;ccedil;onToutes activit&amp;eacute;s en entreprise Entretien et cr&amp;eacute;ation despaces verts" />
                            string propertyValue = gHTMLReader.PropertyValue;
                            if (propertyValue != null)
                            {
                                propertyValue = propertyValue.Replace("\x03", "");
                            }
                            //AddAttribute(gCurrentNode, sPropertyName, gHTMLReader.PropertyValue);
                            AddAttribute(gCurrentNode, sPropertyName, propertyValue);


                            if (gHTMLReader.IsMarkBeginEnd)
                            {
                                TagEnd(gHTMLReader.MarkName.ToLower());
                            }
                        }
                        catch
                        {
                        }
                    }
                    else if (gHTMLReader.IsMarkBeginEnd)
                    {
                        //TagBegin(gHTMLReader.MarkName.ToLower(), true);

                        string sTagName = gHTMLReader.MarkName.ToLower();
                        sTagName = gReplace.Replace(sTagName, "_");
                        if (sTagName == "")
                        {
                            sTagName = "_";
                        }

                        TagBegin(sTagName, true);
                    }
                    else if (gHTMLReader.IsMarkBegin)
                    {
                        //TagBegin(gHTMLReader.MarkName.ToLower(), false);

                        string sTagName = gHTMLReader.MarkName.ToLower();
                        sTagName = gReplace.Replace(sTagName, "_");
                        if (sTagName == "")
                        {
                            sTagName = "_";
                        }

                        TagBegin(sTagName, false);
                    }
                    else if (gHTMLReader.IsMarkEnd)
                    {
                        //TagEnd(gHTMLReader.MarkName.ToLower());

                        string sTagName = gHTMLReader.MarkName.ToLower();
                        sTagName = gReplace.Replace(sTagName, "_");
                        if (sTagName == "")
                        {
                            sTagName = "_";
                        }

                        TagEnd(sTagName);
                    }
                }
            }
            finally
            {
                gHTMLReader.Close();
            }
        }