//List<EMDoc> collection = new List<EMDoc>();
public EMDoc AddDoc(string file, string src, bool ocr, DateTime dt)
{
OCR = ocr;
source = src;
userDT = dt;
EMDoc doc = new EMDoc();
string text;
doc.Filename = file;
//Read the file and look for the info.
if (OCR) //if OCRing read it all
{
text = GetTextFromPDF(file, -1);
doc.OCRText = text;
}
else //else read first page
{
text = GetTextFromPDF(file, 1);
}
if (text == null)
{
text = "";
}
//Identify most commonly used terms in document
IEnumerable <Tuple <string, int> > terms = GetMostUsedWords(text);
List <string> keywords = new List <string>();
//WORK ON LATER
//foreach (var t in terms)
//{
// //Get keywords from SQLite...need to build the SQLite method and table in file and finish the code below that compares. Then put keywords most found in XML node
// List<string> kwListText = new List<string>();
// kwListText = SQLite.GetKeyWord("dbo.key_agency_ocrtext");
// foreach (var k in kwListText)
// {
// if (t.Item1 == k //look in sqlite for terms to remove)
// Console.WriteLine(t.Item1 + ", " + t.Item2);
// }
//}
var matchID = Regex.Match(file, @"\\(\d+)_ _"); //match for docs from the website
//emcity.org
//GRAMA
//Other
if (source != "emcity.org")
{
doc.Source = source;
matchID = Regex.Match(file, @"(\\)(?!.*\\)(.*)\.");
doc.Id = SQLite.InsertID("dbo." + source.ToLower(), "G", Path.GetFileName(file));
}
else //if it is from the website
{
doc.Id = matchID.Groups[1].Value;
}
doc.Date = GetDate(Path.GetFileName(file), text);
doc.Agency = GetAgency(file, text);
doc.Type = GetDocType(file, text);
doc.Title = GetTitle(file, text);
doc.Extension = Path.GetExtension(file);
//Try to fill in the agency based on the doctype or the title since it might have changed from reading the file
if (doc.Agency == "")
{
if (doc.Type == "Agreement" || doc.Type == "Resolution")
{
doc.Agency = "City Council";
}
if (doc.Agency == "")
{
doc.Agency = GetAgency(doc.Title, "");
}
}
//if you have the right fields filled out, look for more info
if (doc.Agency != "" && doc.Type != "" && text != null)
{
if (!OCR) //if not OCRing the whole doc, a bit more is needed to find council and staff members
{
text = GetTextFromPDF(file, 2);
}
if (text != null)
{
doc.Officials = GetOfficials(doc.Agency, doc.Type, text);
doc.Staff = GetStaff(doc.Agency, doc.Type, text);
doc.TypeDet = GetTypeDet(doc.Agency, doc.Type, text);
if (doc.Agency == "City Council" && doc.Type == "Minutes")
{
doc.Votes = GetVotes(doc.Agency, doc.Type, text);
}
}
}
if (doc.Project == "")
{
if (text == null)
{
text = "";
}
doc.Project = GetProject(file, text);
if (doc.Project == "")
{
doc.Project = GetProject(doc.Agency, doc.Type, file, text);
}
}
//Console.WriteLine(text);
Console.WriteLine(doc.Id);
//collection.Add(doc);
return(doc);
}
private void BGW_Organize_DoWork(object sender, DoWorkEventArgs e)
{
Log.AddMessage("Beginning organization of files...", "Information");
UpdateStatusLabel("Finding files...");
EMDocs emdocs = new EMDocs();
//List<string> eDocs = new List<string>();
if (!Directory.Exists(tbOutputOrg.Text))
{
Directory.CreateDirectory(tbOutputOrg.Text);
}
if (Directory.Exists(tbBaseFolder.Text))
{
var files = Directory.GetFiles(tbBaseFolder.Text, "*.pdf", SearchOption.AllDirectories);
int currentEDoc = 1;
if (files.Length == 0)
{
Log.AddMessage("No files found to organize...", "Information");
MessageBox.Show("No files found to organize...", "Information");
}
foreach (var f in files)
{
EMDoc eDoc = emdocs.AddDoc(f, sSource, cbOCR.Checked, (cbFillDate.Checked ? dateTimePicker1.Value : DateTime.Now));
UpdateStatusLabel("Indexing source " + currentEDoc + " of " + files.Length + ":\n" + eDoc.Id + " - " + eDoc.Title);
//eDocs.Add(em.Id + " - " + em.Title);
//Filename
//string saveFile = eDoc.Id + "_" + eDoc.Date.ToShortDateString() + "_" + eDoc.Agency + "_" + eDoc.Type + "_" + eDoc.Title;
string ag = eDoc.Agency;
string type = eDoc.Type;
int year = eDoc.Date.Year;
int month = eDoc.Date.Month;
int day = eDoc.Date.Day;
string docDir = "";
string relDocDir = "";
string xmlDir = "";
string ocrDir = "";
string relOcrDir = "";
docDir = tbOutputOrg.Text + "\\docs\\";//"\\" + ag + "\\" + type + "\\" + year + "\\";
relDocDir = "\\docs\\";
xmlDir = tbOutputOrg.Text + "\\xmls\\";
ocrDir = tbOutputOrg.Text + "\\ocr\\";
relOcrDir = "\\ocr\\";
string shortDate = eDoc.Date.ToShortDateString();
if (eDoc.Agency == "" || eDoc.Type == "" || year == 0 || eDoc.Title == "" || shortDate == "1/1/0001")
{
docDir = tbOutputOrg.Text + "\\_incomplete\\";
relDocDir = "\\_incomplete\\";
xmlDir = tbOutputOrg.Text + "\\_incomplete\\";
ocrDir = tbOutputOrg.Text + "\\_incomplete\\";
relOcrDir = "\\_incomplete\\";
}
UpdateStatusLabel("Saving output " + currentEDoc + " of " + files.Length + ":\n" + eDoc.Id + " - " + eDoc.Title);
if (!Directory.Exists(docDir))
{
Directory.CreateDirectory(docDir);
}
if (!Directory.Exists(xmlDir))
{
Directory.CreateDirectory(xmlDir);
}
if (cbOCR.Checked)
{
if (!Directory.Exists(ocrDir))
{
Directory.CreateDirectory(ocrDir);
}
}
//Path to save to
File.Copy(eDoc.Filename, docDir + Path.GetFileNameWithoutExtension(eDoc.Filename) + eDoc.Extension, true); //tbOutputOrg.Text + "\\" + ag + "\\" + year + "\\" + month + "\\" + RemoveProbChars(saveFile) + eDoc.Extension, true);
UpdateStatusLabel("Building index " + currentEDoc + " of " + files.Length + ":\n" + eDoc.Id + " - " + eDoc.Title);
XmlDocument xml = new XmlDocument(); // Create the XML Declaration, and append it to XML eDoc
XmlDeclaration dec = xml.CreateXmlDeclaration("1.0", "UTF-8", null);
xml.AppendChild(dec); // Create the root element
XmlElement root = xml.CreateElement("EMDoc");
xml.AppendChild(root);
XmlElement id = xml.CreateElement("WebsiteID");
id.InnerText = eDoc.Id.ToString();
root.AppendChild(id);
XmlElement mime = xml.CreateElement("Mime");
mime.InnerText = eDoc.Extension;
root.AppendChild(mime);
XmlElement source = xml.CreateElement("Source");
source.InnerText = eDoc.Source;
root.AppendChild(source);
XmlElement sourceURL = xml.CreateElement("SourceURL");
if (sSource == "emcity.org")
{
if (eDoc.Extension.ToLower() == ".pdf")
{
sourceURL.InnerText = @"http://www.eaglemountaincity.org/Home/ShowDocument?id=" + eDoc.Id.ToString();
}
else
{
sourceURL.InnerText = @"http://www.eaglemountaincity.org/Home/ShowImage?id=" + eDoc.Id.ToString();
}
}
root.AppendChild(sourceURL);
XmlElement path = xml.CreateElement("Path");
path.InnerText = relDocDir + Path.GetFileNameWithoutExtension(eDoc.Filename) + eDoc.Extension;
//path.InnerText = tbOutputFolder.Text + "\\" + eDoc.Id + "_" + eDoc.Date.ToShortDateString() + "_" + eDoc.Title + eDoc.Extension;
root.AppendChild(path);
XmlElement filename = xml.CreateElement("Filename");
filename.InnerText = Path.GetFileNameWithoutExtension(eDoc.Filename) + eDoc.Extension;
root.AppendChild(filename);
if (cbOCR.Checked)
{
XmlElement ocr = xml.CreateElement("OCR");
ocr.InnerText = relOcrDir + Path.GetFileNameWithoutExtension(eDoc.Filename) + ".txt";
root.AppendChild(ocr);
}
XmlElement flds = xml.CreateElement("Fields");
root.AppendChild(flds);
XmlElement docName = xml.CreateElement("Title");
docName.InnerText = eDoc.Title;
flds.AppendChild(docName);
XmlElement date = xml.CreateElement("Date");
date.InnerText = eDoc.Date.ToShortDateString();
flds.AppendChild(date);
XmlElement yr = xml.CreateElement("Year");
yr.InnerText = year.ToString();
flds.AppendChild(yr);
XmlElement agency = xml.CreateElement("Agency");
agency.InnerText = eDoc.Agency;
flds.AppendChild(agency);
XmlElement docType = xml.CreateElement("DocType");
docType.InnerText = eDoc.Type;
flds.AppendChild(docType);
XmlElement docTypeDet = xml.CreateElement("DocTypeDetail");
docTypeDet.InnerText = eDoc.TypeDet;
flds.AppendChild(docTypeDet);
if (sProj != "")
{
if (eDoc.Project == "")
{
XmlElement project = xml.CreateElement("Project");
project.InnerText = sProj;
flds.AppendChild(project);
}
{
XmlElement project = xml.CreateElement("Project");
project.InnerText = eDoc.Project;
flds.AppendChild(project);
}
}
else
{
XmlElement project = xml.CreateElement("Project");
project.InnerText = eDoc.Project;
flds.AppendChild(project);
}
XmlElement officials = xml.CreateElement("Officials");
officials.InnerText = eDoc.Officials;
flds.AppendChild(officials);
XmlElement staff = xml.CreateElement("Staff");
staff.InnerText = eDoc.Staff;
flds.AppendChild(staff);
XmlElement votes = xml.CreateElement("Votes");
votes.InnerText = eDoc.Votes;
flds.AppendChild(votes);
xml.Save(xmlDir + Path.GetFileNameWithoutExtension(eDoc.Filename) + ".xml"); //tbOutputOrg.Text + "\\" + ag + "\\" + year + "\\" + month + "\\" + RemoveProbChars(saveFile) + ".xml");
if (cbOCR.Checked)
{
if (eDoc.OCRText != "")
{
UpdateStatusLabel("Saving OCR " + currentEDoc + " of " + files.Length + ":\n" + eDoc.Id + " - " + eDoc.Title);
using (StreamWriter file = new System.IO.StreamWriter(ocrDir + Path.GetFileNameWithoutExtension(eDoc.Filename) + ".txt", false))
{
file.Write(eDoc.OCRText);
}
}
}
currentEDoc++;
}
}
else
{
Console.WriteLine("Base Directory Doesn't Exist");
}
//int currentEDoc = 1;
//Log.AddMessage(eDocs.Count + " files found to organize", "Information");
//foreach (EMDoc eDoc in eDocs)
//{
// UpdateStatusLabel("Saving files and indexes " + currentEDoc + " of " + emdocs.CollectionCount() + ":\n" + eDoc.Id + " - " + eDoc.Title);
// //Filename
// //string saveFile = eDoc.Id + "_" + eDoc.Date.ToShortDateString() + "_" + eDoc.Agency + "_" + eDoc.Type + "_" + eDoc.Title;
// string ag = eDoc.Agency;
// string type = eDoc.Type;
// int year = eDoc.Date.Year;
// int month = eDoc.Date.Month;
// int day = eDoc.Date.Day;
// string docDir = "";
// string relDocDir = "";
// string xmlDir = "";
// string ocrDir = "";
// string relOcrDir = "";
// docDir = tbOutputOrg.Text + "\\docs\\";//"\\" + ag + "\\" + type + "\\" + year + "\\";
// relDocDir = "\\docs\\";
// xmlDir = tbOutputOrg.Text + "\\xmls\\";
// ocrDir = tbOutputOrg.Text + "\\ocr\\";
// relOcrDir = "\\ocr\\";
// string shortDate = eDoc.Date.ToShortDateString();
// if (eDoc.Agency == "" || eDoc.Type == "" || year == 0 || eDoc.Title == "" || shortDate == "1/1/0001")
// {
// docDir = tbOutputOrg.Text + "\\_incomplete\\";
// relDocDir = "\\_incomplete\\";
// xmlDir = tbOutputOrg.Text + "\\_incomplete\\";
// ocrDir = tbOutputOrg.Text + "\\_incomplete\\";
// relOcrDir = "\\_incomplete\\";
// }
// if (!Directory.Exists(docDir))
// {
// Directory.CreateDirectory(docDir);
// }
// if (!Directory.Exists(xmlDir))
// {
// Directory.CreateDirectory(xmlDir);
// }
// if (!Directory.Exists(ocrDir))
// {
// Directory.CreateDirectory(ocrDir);
// }
// //Path to save to
// File.Copy(eDoc.Filename, docDir + Path.GetFileNameWithoutExtension(eDoc.Filename) + eDoc.Extension, true); //tbOutputOrg.Text + "\\" + ag + "\\" + year + "\\" + month + "\\" + RemoveProbChars(saveFile) + eDoc.Extension, true);
// XmlDocument xml = new XmlDocument();// Create the XML Declaration, and append it to XML eDoc
// XmlDeclaration dec = xml.CreateXmlDeclaration("1.0", "UTF-8", null);
// xml.AppendChild(dec);// Create the root element
// XmlElement root = xml.CreateElement("EMDoc");
// xml.AppendChild(root);
// XmlElement id = xml.CreateElement("WebsiteID");
// id.InnerText = eDoc.Id.ToString();
// root.AppendChild(id);
// XmlElement mime = xml.CreateElement("Mime");
// mime.InnerText = eDoc.Extension;
// root.AppendChild(mime);
// XmlElement source = xml.CreateElement("Source");
// source.InnerText = eDoc.Source;
// root.AppendChild(source);
// XmlElement sourceURL = xml.CreateElement("SourceURL");
// if (eDoc.Extension.ToLower() == ".pdf")
// {
// sourceURL.InnerText = @"http://www.eaglemountaincity.org/Home/ShowDocument?id=" + eDoc.Id.ToString();
// }
// else
// {
// sourceURL.InnerText = @"http://www.eaglemountaincity.org/Home/ShowImage?id=" + eDoc.Id.ToString();
// }
// root.AppendChild(sourceURL);
// XmlElement path = xml.CreateElement("Path");
// path.InnerText = relDocDir + Path.GetFileNameWithoutExtension(eDoc.Filename) + eDoc.Extension;
// //path.InnerText = tbOutputFolder.Text + "\\" + eDoc.Id + "_" + eDoc.Date.ToShortDateString() + "_" + eDoc.Title + eDoc.Extension;
// root.AppendChild(path);
// XmlElement filename = xml.CreateElement("Filename");
// filename.InnerText = Path.GetFileNameWithoutExtension(eDoc.Filename) + eDoc.Extension;
// root.AppendChild(filename);
// if (cbOCR.Checked)
// {
// XmlElement ocr = xml.CreateElement("OCR");
// ocr.InnerText = relOcrDir + Path.GetFileNameWithoutExtension(eDoc.Filename) + ".txt";
// root.AppendChild(ocr);
// }
// XmlElement flds = xml.CreateElement("Fields");
// root.AppendChild(flds);
// XmlElement docName = xml.CreateElement("Title");
// docName.InnerText = eDoc.Title;
// flds.AppendChild(docName);
// XmlElement date = xml.CreateElement("Date");
// date.InnerText = eDoc.Date.ToShortDateString();
// flds.AppendChild(date);
// XmlElement yr = xml.CreateElement("Year");
// yr.InnerText = year.ToString();
// flds.AppendChild(yr);
// XmlElement agency = xml.CreateElement("Agency");
// agency.InnerText = eDoc.Agency;
// flds.AppendChild(agency);
// XmlElement docType = xml.CreateElement("DocType");
// docType.InnerText = eDoc.Type;
// flds.AppendChild(docType);
// XmlElement docTypeDet = xml.CreateElement("DocTypeDetail");
// docTypeDet.InnerText = eDoc.TypeDet;
// flds.AppendChild(docTypeDet);
// XmlElement project = xml.CreateElement("Project");
// project.InnerText = eDoc.Project;
// flds.AppendChild(project);
// XmlElement officials = xml.CreateElement("Officials");
// officials.InnerText = eDoc.Officials;
// flds.AppendChild(officials);
// XmlElement staff = xml.CreateElement("Staff");
// staff.InnerText = eDoc.Staff;
// flds.AppendChild(staff);
// XmlElement votes = xml.CreateElement("Votes");
// votes.InnerText = eDoc.Votes;
// flds.AppendChild(votes);
// xml.Save(xmlDir + Path.GetFileNameWithoutExtension(eDoc.Filename) + ".xml"); //tbOutputOrg.Text + "\\" + ag + "\\" + year + "\\" + month + "\\" + RemoveProbChars(saveFile) + ".xml");
// if (cbOCR.Checked)
// {
// if (eDoc.OCRText != "")
// {
// using (StreamWriter file = new System.IO.StreamWriter(ocrDir + Path.GetFileNameWithoutExtension(eDoc.Filename) + ".txt", false))
// {
// file.Write(eDoc.OCRText);
// }
// }
// }
// currentEDoc++;
//}
}
