private static XDocument ProcessZipFile(byte[] fileData)
{
using (var ms = new MemoryStream(fileData)) {
try {
using (var gz = new ZipArchive(ms, ZipArchiveMode.Read)) {
var entries = gz.Entries;
var docFile = entries.First(z => Path.GetExtension(z.Name).StartsWith(".htm"));
// copy resources so we can leave the zip safely
var resourceFiles = entries.Except(new[] { docFile }).Select(r => {
var name = r.FullName;
var imageStream = new MemoryStream();
r.Open().CopyTo(imageStream);
var content = imageStream.ToArray();
return(new {
Name = name,
Content = content
});
}).ToList();
Html2XmlUtil.imageAsBase64 = true;
Html2XmlUtil.TreatExternalData += (sender, args) => {
var images = resourceFiles.Where(f => f.Name.EndsWith(args.FileName)).ToList();
if (!images.Any())
{
return;
}
if (images.Count() > 1)
{
images = resourceFiles.Where(f => f.Name == Path.Combine(args.FilePath, args.FileName)).ToList();
}
var image = images.First();
try {
using (var imageStream = new MemoryStream(image.Content)) {
var img = System.Drawing.Image.FromStream(imageStream);
using (var pngImgMs = new MemoryStream()) {
img.Save(pngImgMs, ImageFormat.Png);
pngImgMs.Position = 0;
args.Data = pngImgMs.ToArray();
}
}
} catch (Exception) {
args.Data = null;
}
};
var result = new MemoryStream();
docFile.Open().CopyTo(result);
if (result.Length == 0)
{
throw new ArgumentOutOfRangeException();
}
string html;
var bytes = result.ToArray();
html = Encoding.UTF8.GetString(bytes); // unclear, even ISO docs do well in UTF8 IsUtf8(bytes) ? Encoding.UTF8.GetString(bytes) : Encoding.GetEncoding(1252).GetString(bytes);
var xDoc = Html2XmlUtil.CleanUpHtmlWithResources(html); // make clean XHTML with embedded images
return(xDoc);
}
} catch (Exception ex) {
throw;
}
}
}
public static Opus ImportSingleHtml(byte[] content, Import mapping, string name)
{
Opus opus = new Opus();
var html = Encoding.UTF8.GetString(content);
try {
// convert prepared HTML into internal <Content> XML (backup and restore format)
var parameters = new System.Collections.Specialized.NameValueCollection();
if (mapping != null)
{
mapping.CharacterStyles.ForEach(c => parameters.Add(c.Key, c.Value));
mapping.ParagraphStyles.ForEach(c => parameters.Add(c.Key, c.Value));
mapping.NumberingStyles.ForEach(c => parameters.Add(c.Key, c.Value));
}
var xml = Html2XmlUtil.HtmlToOpusXsltParser(html, parameters);
using (var xmlStream = new MemoryStream(Encoding.UTF8.GetBytes(xml))) {
var xDoc = XDocument.Load(xmlStream);
// use restore to create import, all content is in Opus, then
RestoreOpusFromFile(opus, xDoc, null);
// register the import information with the uploaded and converted HTML
}
}
catch (Exception ex) {
}
return(opus);
}
private static void Main(string[] args)
{
string path, html;
var defKey = "y";
XDocument xml;
do
{
Console.Write("Import Word HTML (H), Word (W), XML (X), Cleanup (C) HTML? ");
XDocument fixTableXslt;
XDocument fixListingXslt;
XDocument config = XDocument.Load("config.xml");
switch (Console.ReadKey(false).KeyChar.ToString().ToLower())
{
case "t":
fixTableXslt = Html2XmlUtil.GenerateFixTableXslt(config);
break;
case "h":
path = GetPath("Word HTML");
// generate the transform files from config on-the-fly
fixTableXslt = Html2XmlUtil.GenerateFixTableXslt(config);
fixListingXslt = Html2XmlUtil.GenerateFixListingXslt(config);
ImportSingleHtml(path, fixTableXslt, fixListingXslt);
break;
case "w":
Console.WriteLine("");
path = GetPath("Word DOCX");
break;
case "x":
Console.WriteLine("");
path = GetPath("texxtoor XML");
break;
case "c":
Console.WriteLine("");
path = GetPath("Word HTML (clean only)");
break;
default:
Console.WriteLine("");
Console.Write("Unknown key. ");
break;
}
Console.Write("Repeat (y) or End (n)?");
defKey = Console.ReadKey(false).KeyChar.ToString().ToLower();
Console.WriteLine("");
} while (defKey == "y");
Console.WriteLine("done");
// TODO: Check results by creating a preview in plain HTML
Console.ReadLine();
}
private static void ImportSingleHtml(string path, XDocument tableFix, XDocument listingFix)
{
var html = File.ReadAllText(path, Encoding.UTF8);
var name = Path.GetFileNameWithoutExtension(path);
Console.WriteLine("Convert start");
try {
// convert prepared HTML into internal <Content> XML (backup and restore format)
var xml = Html2XmlUtil.HtmlToOpusXsltParser(name, html, tableFix, listingFix);
using (var xmlStream = new MemoryStream(Encoding.UTF8.GetBytes(xml))) {
Console.WriteLine("Parser");
var xDoc = XDocument.Load(xmlStream);
}
}
catch (Exception ex) {
var cc = Console.ForegroundColor;
Console.ForegroundColor = ConsoleColor.Red;
Console.WriteLine(ex.Message);
Console.ForegroundColor = cc;
}
}
private void fehlerBehandelnToolStripMenuItem_Click(object sender, EventArgs e)
{
// convert into opus XML and manage conversion errors
var mapping = GetDefaultMapping();
var html = File.ReadAllText(htmlFileName, Encoding.Default);
try {
// Make XHTML
var xhtmlDocument = new Texxtoor.BaseLibrary.Core.HtmlAgility.Pack.HtmlDocument();
xhtmlDocument.OptionOutputAsXhtml = true;
xhtmlDocument.OptionFixNestedTags = true;
xhtmlDocument.OptionOutputAsXml = true;
xhtmlDocument.LoadHtml(html);
var ms = new MemoryStream();
var tw = new XmlTextWriter(ms, Encoding.Default);
xhtmlDocument.Save(tw);
// change encoding, as our xslt module needs utf-8
var enc = Encoding.GetEncoding("iso-8859-1");
var text = enc.GetString(ms.ToArray());
var xhtml = Encoding.UTF8.GetString(Encoding.UTF8.GetBytes(text));
// transform
var xml = Html2XmlUtil.HtmlToOpusXsltParser(xhtml, mapping);
// TODO: check XML here
// TODO: if XML is okay store on disc for further processing
var xmlFileName = Path.Combine(Path.GetDirectoryName(htmlFileName) ?? String.Empty, String.Format("{0}.xml", Path.GetFileNameWithoutExtension(htmlFileName)));
if (File.Exists(xmlFileName))
{
AddTextToProtocol("XML exists, deleting ");
File.Delete(xmlFileName);
AddTextToProtocol("XML deleted ");
}
AddTextToProtocol("Attempt to write XML");
File.WriteAllText(xmlFileName, xml, Encoding.UTF8);
AddTextToProtocol("XML written to disc at {0}", xmlFileName);
} catch (Exception ex) {
AddTextToProtocol("** Fehler beim Konvertieren in Texxtoor-XML: {0}", ex.Message);
}
}
private void fehlerBehandelnToolStripMenuItem_Click(object sender, EventArgs e)
{
// convert into opus XML and manage conversion errors
var mapping = GetDefaultMapping();
var html = File.ReadAllText(htmlFileName, Encoding.Default);
try {
AddTextToProtocol("Make XHTML");
// Make XHTML
var xhtmlDocument = new Texxtoor.BaseLibrary.Core.HtmlAgility.Pack.HtmlDocument();
xhtmlDocument.OptionOutputAsXhtml = true;
xhtmlDocument.OptionFixNestedTags = true;
xhtmlDocument.OptionOutputAsXml = true;
xhtmlDocument.LoadHtml(html);
var ms = new MemoryStream();
var tw = new XmlTextWriter(ms, Encoding.Default);
xhtmlDocument.Save(tw);
// change encoding, as our xslt module needs utf-8
var enc = Encoding.GetEncoding("iso-8859-1");
var text = enc.GetString(ms.ToArray());
var xhtml = Encoding.UTF8.GetString(Encoding.UTF8.GetBytes(text));
// convert to UTF-8 for XSLT parsing
var doc = System.Xml.Linq.XDocument.Parse(xhtml);
doc.Declaration.Encoding = "UTF-8";
var sb = new StringBuilder();
var targetDocWriter = new StringWriterUtf8(sb);
// replace entites with regular spaces, double call for weird &nbsp;'s
var leaves = from elm in doc.Descendants()
where !elm.Elements().Any()
select elm;
foreach (var leaf in leaves)
{
var v = HttpUtility.HtmlDecode(leaf.Value);
v = HttpUtility.HtmlDecode(v);
if (v.Contains(((char)133).ToString()))
{
v = v.Replace(((char)133).ToString(), "...");
}
leaf.Value = v;
}
doc.Save(targetDocWriter);
xhtml = sb.ToString();
// transform
var xml = Html2XmlUtil.HtmlToOpusXsltParser(xhtml, mapping);
// TODO: check XML here
using (var xmlStream = new MemoryStream(Encoding.UTF8.GetBytes(xml))) {
using (var xmlReader = new XmlTextReader(xmlStream)) {
doc = XDocument.Load(xmlReader);
}
}
AddTextToProtocol("Fix XML");
// 1. Fix headers: <Element type="section" Name="1 Name">1 Name</Element> ===> Remove "1 "
var headers = doc.Descendants("Element").Where(elm => elm.Attribute("Type") != null && elm.Attribute("Type").Value == "Section");
foreach (var header in headers)
{
if (header.FirstNode.NodeType == XmlNodeType.Text)
{
var value = ((XText)header.FirstNode).Value.Trim();
var match = Regex.Match(value, @"^\d{1,3}(?:\.\d{1,3}){0,5}\s+(.*)");
if (match.Success)
{
((XText)header.FirstNode).Value = match.Groups[1].Value;
}
}
}
// 2. Fix bullet points <li>. Text</li> ===> Remove ". "
var bullets = doc.Descendants("li");
foreach (var bullet in bullets)
{
var match = Regex.Match(bullet.Value, @"[·]\s+(.*)");
if (match.Success)
{
bullet.Value = match.Groups[1].Value;
}
}
AddTextToProtocol("Embed Images");
// 3. Embed images (Element Type="Image")
var images = doc.Descendants("Element").Where(elm => elm.Attribute("Type") != null && elm.Attribute("Type").Value == "Image");
foreach (var image in images)
{
// try reading file
if (image.Attribute("Path") == null)
{
continue;
}
// we take PNG only, hence a conversion is appropriate
var path = Path.Combine(imgFolder, Path.GetFileName(image.Attribute("Path").Value));
if (!File.Exists(path))
{
continue;
}
var img = Image.FromFile(path);
using (var imgMs = new MemoryStream()) {
img.Save(imgMs, ImageFormat.Png);
var base64 = Convert.ToBase64String(imgMs.ToArray());
// apply as embedded base64
image.Add(new XAttribute("Method", "Base64"));
image.Value = base64;
}
}
sb.Clear();
targetDocWriter = new StringWriterUtf8(sb);
doc.Save(targetDocWriter);
xml = sb.ToString();
// TODO: if XML is okay store on disc for further processing
var xmlFileName = Path.Combine(Path.GetDirectoryName(htmlFileName) ?? String.Empty, String.Format("{0}.xml", Path.GetFileNameWithoutExtension(htmlFileName)));
if (File.Exists(xmlFileName))
{
AddTextToProtocol("XML exists, deleting ");
File.Delete(xmlFileName);
AddTextToProtocol("XML deleted ");
}
AddTextToProtocol("Attempt to write XML");
File.WriteAllText(xmlFileName, xml, Encoding.UTF8);
AddTextToProtocol("XML written to disc at {0}", xmlFileName);
} catch (Exception ex) {
AddTextToProtocol("** Fehler beim Konvertieren in Texxtoor-XML: {0}", ex.Message);
}
}
