public ParatextTextCorpus(ITokenizer <string, int> wordTokenizer, string projectDir)
{
string settingsFileName = Path.Combine(projectDir, "Settings.xml");
var settingsDoc = XDocument.Load(settingsFileName);
var codePage = (int?)settingsDoc.Root.Element("Encoding") ?? 65001;
EncodingInfo encodingInfo = Encoding.GetEncodings().FirstOrDefault(ei => ei.CodePage == codePage);
if (encodingInfo == null)
{
throw new InvalidOperationException("The Paratext project contains an unknown encoding.");
}
Encoding encoding = encodingInfo.GetEncoding();
var scrVersType = (int?)settingsDoc.Root.Element("Versification") ?? (int)ScrVersType.English;
Versification = new ScrVers((ScrVersType)scrVersType);
var stylesheetName = (string)settingsDoc.Root.Element("StyleSheet") ?? "usfm.sty";
string stylesheetFileName = Path.Combine(projectDir, stylesheetName);
var stylesheet = new UsfmStylesheet(stylesheetFileName);
foreach (string sfmFileName in Directory.EnumerateFiles(projectDir, "*.SFM"))
{
AddText(new UsfmFileText(wordTokenizer, stylesheet, encoding, sfmFileName, Versification));
}
}
protected UsfmTextBase(ITokenizer <string, int> wordTokenizer, string id, UsfmStylesheet stylesheet,
Encoding encoding, ScrVers versification)
: base(wordTokenizer, id)
{
_parser = new UsfmParser(stylesheet);
_encoding = encoding;
Versification = versification ?? ScrVers.English;
}
public UsfmFileTextCorpus(ITokenizer <string, int> wordTokenizer, string stylesheetFileName, Encoding encoding,
string projectPath, ScrVers versification = null)
{
Versification = versification ?? ScrVers.English;
var stylesheet = new UsfmStylesheet(stylesheetFileName);
foreach (string sfmFileName in Directory.EnumerateFiles(projectPath, "*.SFM"))
{
AddText(new UsfmFileText(wordTokenizer, stylesheet, encoding, sfmFileName, Versification));
}
}
public UsfmParser(UsfmStylesheet stylesheet)
{
_stylesheet = stylesheet;
}
public UsfmFileText(ITokenizer <string, int> wordTokenizer, UsfmStylesheet stylesheet,
Encoding encoding, string fileName, ScrVers versification = null)
: base(wordTokenizer, GetId(fileName, encoding), stylesheet, encoding, versification)
{
_fileName = fileName;
}