public Task <ParsingResult> Parse(ParsingRequest request)
{
if (request == null)
{
throw new ArgumentNullException(nameof(request));
}
logger.LogDebug("Parsing [{0}]", request.File.FullName);
var document = new RawDocument();
using (var documentProcessor = new PdfDocumentProcessor())
{
documentProcessor.LoadDocument(request.File.FullName);
var pages = request.MaxPages > documentProcessor.Document.Pages.Count ? documentProcessor.Document.Pages.Count : request.MaxPages;
var pagesList = new List <RawPage>();
document.Pages = new RawPage[pages];
for (var i = 1; i <= pages; i++)
{
var page = new RawPage();
var data = GetImage(request, documentProcessor, i);
page.Blocks = ocrImageParser.Parse(data).ToArray();
document.Pages[i - 1] = page;
}
}
return(Task.FromResult(new ParsingResult(document, request, ParsingType.OCR)));
}
public RawDocument GenerateResult(int maxPages)
{
ExtractBlocks();
var result = new RawDocument();
result.Pages = pages.Take(maxPages).ToArray();
return(result);
}
public MarkerWindow()
: base(Gtk.WindowType.Toplevel)
{
Build ();
rdocument = null;
ndocument = null;
html_document = null;
AddColumns ();
AddButtons ();
}
public void ConstructorInterfase()
{
int count = 0;
Type etype = Type.GetType ("Scielo.PDF2Text.RawDocument");
foreach (PDFPoppler doc in test_docs) {
RawDocument rdoc = new RawDocument (doc);
Assert.IsInstanceOfType (etype, rdoc, "CI" + count);
count++;
}
}
public void ConstructorString()
{
RawDocument rdoc0 = new RawDocument ("", "atm");
RawDocument rdoc1 = new RawDocument ("Hola Mundo", "atm");
RawDocument rdoc2 = new RawDocument (" ad ", "atm");
Type etype = Type.GetType ("Scielo.PDF2Text.RawDocument");
Assert.IsInstanceOfType (etype, rdoc0, "CI01");
Assert.IsInstanceOfType (etype, rdoc1, "CI01");
Assert.IsInstanceOfType (etype, rdoc2, "CI01");
}
public Normalizer(RawDocument document, string format)
{
// Construimos un StyleReader para obtener las regexp.
StyleReader style = new StyleReader (format);
this.format = format;
rules = style.GetRules ();
// Si el estilo tiene mas de una columna se rompe y se convierte a una
// sola columna.
if (style.GetNumColumns () > 1)
document.BreakColumns ();
EncodeText (document.GetText ());
}
public Task <Document[]> Extract(string domain, RawDocument rawDocument)
{
logger.LogDebug("Parsing");
SingleRequestData[] requests = new SingleRequestData[rawDocument.Pages.Length];
for (int i = 0; i < rawDocument.Pages.Length; i++)
{
string text = rawDocument.Pages[i].Blocks.Select(x => x.Text).AccumulateItems(" ");
SingleRequestData request = new SingleRequestData
{
Text = text,
Id = i.ToString()
};
requests[i] = request;
}
return(GetSentiment(domain, requests));
}
public Task <ParsingResult> Parse(ParsingRequest request)
{
if (request == null)
{
throw new ArgumentNullException(nameof(request));
}
logger.LogDebug("Parsing [{0}]", request.File.FullName);
var sourceImage = Image.FromFile(request.File.FullName);
using (var byteStream = new MemoryStream())
{
sourceImage.Save(byteStream, ImageFormat.Tiff);
var data = byteStream.ToArray();
var document = new RawDocument();
document.Pages = new[] { new RawPage() };
document.Pages[0].Blocks = ocrImageParser.Parse(data).Take(request.MaxPages).ToArray();
return(Task.FromResult(new ParsingResult(document, request, ParsingType.OCR)));
}
}
public Task <ParsingResult> Parse(ParsingRequest request)
{
if (request == null)
{
throw new ArgumentNullException(nameof(request));
}
logger.LogDebug("Parsing [{0}]", request.File.FullName);
var document = new RawDocument();
bool containsText = false;
using (var documentProcessor = new PdfDocumentProcessor())
{
documentProcessor.LoadDocument(request.File.FullName);
var pages = request.MaxPages > documentProcessor.Document.Pages.Count ? documentProcessor.Document.Pages.Count : request.MaxPages;
document.Pages = new RawPage[pages];
for (var i = 1; i <= pages; i++)
{
var page = new RawPage
{
Blocks = new[] { new TextBlockItem() }
};
page.Blocks[0].Text = documentProcessor.GetPageText(i);
if (!string.IsNullOrWhiteSpace(page.Blocks[0].Text))
{
containsText = true;
}
document.Pages[i - 1] = page;
}
}
if (!containsText)
{
logger.LogInformation("Failed to find text in: [{0}]", request.File.FullName);
return(Task.FromResult(ParsingResult.ConstructError(request)));
}
return(Task.FromResult(new ParsingResult(document, request, ParsingType.Extract)));
}
public void GetText()
{
string rawtext;
int count = 0;
foreach (PDFPoppler doc in test_docs) {
RawDocument rdoc = new RawDocument (doc);
rawtext = rdoc.GetText ();
Assert.AreEqual (raw_docs[count], rawtext, "GT" + count);
count += 1;
}
}
public void WriteDocument()
{
string result, temp_dir;
int count = 0;
foreach (PDFPoppler doc in test_docs) {
RawDocument rdoc = new RawDocument (doc);
temp_dir = Path.GetTempPath ();
rdoc.WriteDocument (temp_dir, "temp01", "txt");
result = Test.ReadFile (Path.Combine (temp_dir, "temp01.txt"));
Assert.AreEqual (raw_docs[count], result, "WD" + count);
count++;
}
}
public void Normalize()
{
NormDocument ndoc;
int count = 0;
Type etype = Type.GetType ("Scielo.PDF2Text.RawDocument");
Type etype1 = Type.GetType ("Scielo.PDF2Text.NormDocument");
foreach (PDFPoppler doc in test_docs) {
RawDocument rdoc = new RawDocument (doc);
ndoc= rdoc.Normalize (styles[count]);
Assert.IsInstanceOfType (etype, rdoc, "NM" + count);
Assert.IsInstanceOfType (etype1, ndoc, "NM" + count);
count++;
}
}
public void SetUp()
{
instance = CreateRawDocument();
}
public ParsingResult(RawDocument document, ParsingRequest request, ParsingType?processedAs)
{
Document = document ?? throw new ArgumentNullException(nameof(document));
Request = request ?? throw new ArgumentNullException(nameof(request));
ProcessedAs = processedAs;
}
private void OnOpenActivated(object sender, System.EventArgs e)
{
OpenPDFDialog dialog = new OpenPDFDialog ();
if (dialog.Run () == (int) ResponseType.Ok) {
Uri uri = new Uri (dialog.Document);
PDFPoppler reader = new PDFPoppler (uri);
//Extracting images from document
reader.GetNonText ();
//Extracting text from document
rdocument = reader.CreateRawDocument ();
textview.Buffer.Text = rdocument.GetText ();
Markup.Sensitive = true;
Normalize.Sensitive = true;
store.Clear ();
// Logger.ClearList ();
}
dialog.Destroy ();
}
public void GetText()
{
string normtext;
int count = 0;
NormDocument ndoc;
foreach (PDFPoppler doc in test_docs) {
RawDocument rdoc = new RawDocument (doc);
ndoc = rdoc.Normalize (styles [count]);
normtext = ndoc.GetText ();
Assert.AreEqual (norm_docs[count], normtext, "GT" + count);
count += 1;
}
}
