public static Dictionary <int, string> Extract(string pdfFileName)
{
if (!File.Exists(pdfFileName))
{
throw new FileNotFoundException("pdfFileName");
}
var result = new Dictionary <int, string>();
PDDocument pdfDocument = PDDocument.load(pdfFileName);
var pdfStripper = new PDFTextStripper();
pdfStripper.setPageSeparator(Environment.NewLine + Environment.NewLine);
for (int i = 1; i <= pdfDocument.getNumberOfPages(); i++)
{
pdfStripper.setStartPage(i);
pdfStripper.setEndPage(i);
result.Add(i, GetText(pdfStripper, pdfDocument));
}
pdfDocument.close();
return(result);
}
public static Dictionary <int, string> Extract(string pdfFileName)
{
if (!File.Exists(pdfFileName))
{
throw new FileNotFoundException("pdfFileName");
}
var result = new Dictionary <int, string>();
PDDocument pdfDocument = PDDocument.load(pdfFileName);
var pdfStripper = new PDFTextStripper();
pdfStripper.setPageSeparator(Environment.NewLine + Environment.NewLine);
for (int i = 1; i <= pdfDocument.getNumberOfPages(); i++)
{
pdfStripper.setStartPage(i);
pdfStripper.setEndPage(i);
//ExtractText(pdfStripper, pdfDocument,
// string.Format(@"c:\Users\tri.hoang\Desktop\temp\epub-belastingblad\2014-08\pdf\page_{0}.txt", i.ToString().PadLeft(5, '0')));
result.Add(i, GetText(pdfStripper, pdfDocument));
}
pdfDocument.close();
return(result);
}
/// <summary>
/// Get a thumbnail of the document, if possible
/// </summary>
/// <param name="sizeX">The maximum X size of the thumbnail</param>
/// <param name="sizeY">The maximum y size of the thumbnail</param>
/// <param name="forceFullSize">True if the thumbnail should be exatly XxY pixels and False if the thumbnail
/// should fit inside a XxY box but should maintain its aspect ratio</param>
/// <returns>A JPEG byte thumbnail or null if the thumbnail can´t be generated</returns>
public override byte[] GetThumbnail(int sizeX, int sizeY, bool forceFullSize)
{
// If we have no bytes then we can't do anything.
if (Bytes == null || Bytes.Length == 0)
{
return(null);
}
try
{
org.pdfbox.pdfviewer.PageDrawer pagedrawer = new
org.pdfbox.pdfviewer.PageDrawer();
java.io.ByteArrayInputStream byteStream = new java.io.ByteArrayInputStream(Bytes);
PDDocument doc = PDDocument.load(byteStream);
int count = doc.getNumberOfPages();
java.util.List pages = doc.getDocumentCatalog().getAllPages();
if (pages.size() > 0)
{
PDPage page = pagedrawer.getPage();
java.awt.image.BufferedImage image = page.convertToImage();
java.io.ByteArrayOutputStream os = new java.io.ByteArrayOutputStream();
ImageIO.write(image, "jpg", os);
byte[] data = os.toByteArray();
return(data);
}
}
catch (Exception e)
{
log.Error("Failed to get the thumbnail from the PDF file " + Name, e);
}
return(null);
}
private void CheckAllPages(PDDocument doc)
{
int num = 0;
try
{
this.NumberOfPagesDict = doc.getNumberOfPages();
foreach (PDPage page in doc.getPages())
{
if (page.getMediaBox() == null)
{
this.ErrorMessage = string.Format("Page number {0} has no media box", num);
this.IsValid = false;
}
if (page.getResources() == null)
{
this.ErrorMessage = string.Format("Page number {0}, has no page resources", num);
this.IsValid = false;
}
num++;
}
if (this.NumberOfPagesDict != num)
{
this.ErrorMessage = string.Format("Page Number Mismatch between dictionary and actual document", new object[0]);
this.IsValid = false;
}
}
catch (Exception exception1)
{
Exception exception = exception1;
this.ErrorMessage = string.Format("PDF analysis failed on page number {0},\nWith exception {1}", num, exception.Message);
this.IsValid = false;
}
}
internal static bool CreateHocrFileFromPDF(PDDocument document, string outputfile, bool useWords)
{
bool flag;
try
{
PDFHocr pDFHocr = new PDFHocr()
{
getHOCRByWords = useWords
};
pDFHocr.setSortByPosition(true);
pDFHocr.setStartPage(0);
pDFHocr.setEndPage(document.getNumberOfPages());
PDFHelper.DisplayTrialPopupIfNecessary();
if (PDFHelper.AddStamp)
{
pDFHocr.setEndPage(3);
}
pDFHocr.writeText(document, new OutputStreamWriter(new ByteArrayOutputStream()));
if ((pDFHocr.lineList == null ? false : pDFHocr.lineList.Count > 0))
{
HocrPageModel hocrPageModel = new HocrPageModel();
hocrPageModel.Lines.AddRange(pDFHocr.SortLineList(pDFHocr.lineList));
pDFHocr.pageList.Add(hocrPageModel);
pDFHocr.lineList.Clear();
}
pDFHocr.GetHocrFromPageList(pDFHocr.pageList, outputfile);
flag = true;
}
catch (Exception exception)
{
flag = false;
}
return(flag);
}
internal static List <HocrPageModel> GetPageWordDetails(PDDocument document)
{
List <HocrPageModel> hocrPageModels;
try
{
PDFHocr pDFHocr = new PDFHocr();
pDFHocr.setSortByPosition(true);
pDFHocr.setStartPage(0);
pDFHocr.setEndPage(document.getNumberOfPages());
Writer outputStreamWriter = new OutputStreamWriter(new ByteArrayOutputStream());
PDFHelper.DisplayTrialPopupIfNecessary();
if (PDFHelper.AddStamp)
{
pDFHocr.setEndPage(3);
}
pDFHocr.writeText(document, outputStreamWriter);
if ((pDFHocr.lineList == null ? false : pDFHocr.lineList.Count > 0))
{
HocrPageModel hocrPageModel = new HocrPageModel();
hocrPageModel.Lines.AddRange(pDFHocr.SortLineList(pDFHocr.lineList));
pDFHocr.pageList.Add(hocrPageModel);
pDFHocr.lineList.Clear();
}
hocrPageModels = pDFHocr.pageList;
}
catch (Exception exception)
{
hocrPageModels = null;
}
return(hocrPageModels);
}
internal static bool AddBookmarkTooutline(PDFBookmarkItem bookmarentry, PDDocument document, PDOutlineItem outline)
{
bool flag;
try
{
if (bookmarentry.BookMarkPage <= document.getNumberOfPages())
{
PDPage page = document.getPage(bookmarentry.BookMarkPage - 1);
PDPageFitWidthDestination pDPageFitWidthDestination = new PDPageFitWidthDestination();
pDPageFitWidthDestination.setPage(page);
outline.setDestination(pDPageFitWidthDestination);
outline.setTitle(bookmarentry.BookmarkTitle);
}
if ((bookmarentry.BookmarkItems == null ? false : bookmarentry.BookmarkItems.Count > 0))
{
foreach (PDFBookmarkItem bookmarkItem in bookmarentry.BookmarkItems)
{
PDOutlineItem pDOutlineItem = new PDOutlineItem();
PDFHelper.AddBookmarkTooutline(bookmarkItem, document, pDOutlineItem);
outline.addLast(pDOutlineItem);
}
}
flag = true;
}
catch (Exception exception)
{
flag = false;
}
return(flag);
}
public static PDFInfo GetPDFDoucmentInformation(PDDocument document)
{
PDFInfo i = new PDFInfo();
PDDocumentInformation info = document.getDocumentInformation();
i.Author = info.getAuthor();
if (info.getCreationDate() != null)
{
DateTime dt = Utilities.Utils.GetDateFromJava(info.getCreationDate());
i.CreationDate = dt.ToLongDateString() + " " + dt.ToLongTimeString();
}
i.Creator = info.getCreator();
i.Keywords = info.getKeywords();
if (info.getModificationDate() != null)
{
DateTime dt = Utilities.Utils.GetDateFromJava(info.getModificationDate());
i.ModificationDate = dt.ToLongDateString() + " " + dt.ToLongTimeString();
}
i.Producer = info.getProducer();
i.Subject = info.getSubject();
i.Title = info.getTitle();
i.Trapped = info.getTrapped();
i.NumberOfPages = document.getNumberOfPages();
return i;
}
private void Run()
{
PDDocument doc = PDDocument.load(_path);
for (int i = 1; i <= doc.getNumberOfPages(); i++)
{
ExtractText et = new ExtractText();
et.Extract(doc, i);
TextPosition.Add(i, et._nodes);
Pages.Add(i, et._text);
}
if (Pages == null)
{
return;
}
for (int i = 1; i <= Pages.Keys.Count; i++)
{
Content += Pages[i];
}
}
public void Process(int pageNum, File file)
{
try
{
PDDocument document = PDDocument.load(file);
// PDFTextStripper stripper = new TextPositonExtracter();
if (pageNum == -1)
{
this.setStartPage(1);
this.setEndPage(document.getNumberOfPages());
}
else
{
this.setStartPage(pageNum);
this.setEndPage(pageNum);
}
Writer dumpy = new OutputStreamWriter(new ByteArrayOutputStream());
this.writeText(document, dumpy);
}
catch
{
}
}
