public override ExtractionResult ExtractText(IDoc doc, ITextExtractorSettings settings)
{
var result = new ExtractionResult(settings.MaxExtractionSize);
var textAbsorber = new TextAbsorber();
using (var pdfDocument = new Document(doc.Stream))
{
pdfDocument.Pages.Accept(textAbsorber);
}
result.Append(textAbsorber.Text);
return(result);
}
public ExtractionResult ExtractTextFromDocument(string inputFile, ITextExtractorSettings settings)
{
var retVal = new ExtractionResult(settings.MaxExtractionSize);
var outputFile = Path.Combine(Path.GetTempPath(), Guid.NewGuid().ToString("N") + ".txt");
var engine = enginesPool.GetEngine();
bool isRecycleRequired = false;
try
{
IFRDocument fineReaderDocument = null;
try
{
fineReaderDocument = LoadFineReaderDocument(inputFile, settings.TextExtractionProfile, engine, settings.Context);
SubscribeExtractionEvents((FRDocument)fineReaderDocument);
fineReaderDocument.Process();
// Leerseiten überspringen bei einseitigen Dokumenten.
if (fineReaderDocument.Pages.Count == 1)
{
// IsEmpty nutzt die Einstellungen die über das Profil festgelegt wurden
if (fineReaderDocument.Pages[0].IsEmpty())
{
Log.Information("The page {inputFile} was detected as empty.", inputFile);
return(retVal);
}
}
fineReaderDocument.Export(outputFile, FileExportFormatEnum.FEF_TextUnicodeDefaults, null);
// Read the contents of the exported file
using (var sr = new StreamReader(outputFile))
{
while (sr.Peek() >= 0)
{
retVal.Append(sr.ReadLine());
if (retVal.LimitExceeded)
{
break;
}
}
}
// Alles OK
return(retVal);
}
finally
{
if (fineReaderDocument != null)
{
UnsubscribeExtractionEvents((FRDocument)fineReaderDocument);
fineReaderDocument.Close();
if (Marshal.IsComObject(fineReaderDocument))
{
Marshal.ReleaseComObject(fineReaderDocument);
}
}
}
}
catch (Exception ex)
{
Log.Error(ex, $"Abbyy Textextraction failed: {ex.Message}");
isRecycleRequired = enginesPool.ShouldRestartEngine(ex);
retVal.HasError = true;
retVal.ErrorMessage = ex.Message;
// Push an error message to indicate that the item has failed
bus.Publish <AbbyyProgressEvent>(new
{
__TimeToLive = TimeSpan.FromSeconds(3),
File = sourceFile,
Process = ProcessType.TextExtraction,
EventType = AbbyyEventType.AbbyyOnProgressEvent,
HasFailed = true
});
}
finally
{
// Push the final message to indicate that the item is finished
bus.Publish <AbbyyProgressEvent>(new
{
__TimeToLive = TimeSpan.FromSeconds(3),
File = sourceFile,
IsComplete = true,
Percentage = 100,
Process = ProcessType.TextExtraction,
EventType = AbbyyEventType.AbbyyOnProgressEvent
});
enginesPool.ReleaseEngine(engine, isRecycleRequired);
if (File.Exists(outputFile))
{
File.Delete(outputFile);
}
}
return(retVal);
}
public ExtractionResult ExtractTextFromDocument(string inputFile, ITextExtractorSettings settings)
{
var retVal = new ExtractionResult(settings.MaxExtractionSize);
var outputFile = Path.Combine(Path.GetTempPath(), Guid.NewGuid().ToString("N") + ".txt");
var engine = enginesPool.GetEngine();
bool isRecycleRequired = false;
try
{
IFRDocument fineReaderDocument = null;
try
{
fineReaderDocument = LoadFineReaderDocument(inputFile, settings.TextExtractionProfile, engine);
var documentProcessingParams = CreateProcessingParams(engine);
fineReaderDocument.Process(documentProcessingParams);
// Leerseiten überspringen bei einseitigen Dokumenten.
if (fineReaderDocument.Pages.Count == 1)
{
// IsEmpty nutzt die Einstellungen die über das Profil festgelegt wurden
if (fineReaderDocument.Pages[0].IsEmpty())
{
return(retVal);
}
}
fineReaderDocument.Export(outputFile, FileExportFormatEnum.FEF_TextUnicodeDefaults, null);
// Read the contents of the exported file
using (var sr = new StreamReader(outputFile))
{
while (sr.Peek() >= 0)
{
retVal.Append(sr.ReadLine());
if (retVal.LimitExceeded)
{
break;
}
}
}
// Alles OK
return(retVal);
}
finally
{
if (fineReaderDocument != null)
{
fineReaderDocument.Close();
if (Marshal.IsComObject(fineReaderDocument))
{
Marshal.ReleaseComObject(fineReaderDocument);
}
}
}
}
catch (Exception ex)
{
Log.Error(ex, $"Abbyy Textextraction failed: {ex.Message}");
isRecycleRequired = enginesPool.ShouldRestartEngine(ex);
retVal.HasError = true;
retVal.ErrorMessage = ex.Message;
}
finally
{
enginesPool.ReleaseEngine(engine, isRecycleRequired);
if (File.Exists(outputFile))
{
File.Delete(outputFile);
}
}
return(retVal);
}
public AllowedExtensionFilter(ITextExtractorSettings settings)
{
this.settings = settings;
}
public abstract ExtractionResult ExtractText(IDoc doc, ITextExtractorSettings settings);
