public static void ResizeBlocksets(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-table-output.pdf")
.Page(1)
.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.ParsePdf <ProcessPdfText>()
.ParseBlock <RemoveTableText>()
.ParseBlock <GroupLines>()
.ParseBlock <FindInitialBlockset>()
.ParseBlock <BreakColumns>()
.Validate <RemoveFooter>().ShowErrors(p => p.Show(Color.Purple))
.Validate <RemoveHeader>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveFooter>()
.ParseBlock <RemoveHeader>()
.ParseBlock <AddTableSpace>()
.ParseBlock <ResizeBlocksets>()
.ParseBlock <OrderBlocksets>()
.Show(Color.Orange)
.ShowLine(Color.Black);
pipeline.Done();
}
public static int Process(string basename, string inputfolder, string outputfolder)
{
//PdfReaderException.DisableWarnings();
//PdfReaderException.ContinueOnException();
var pipeline = new Execution.Pipeline();
var result =
pipeline.Input($"{inputfolder}/{basename}.pdf")
//.Output($"{outputfolder}/{basename}-output.pdf")
.AllPagesExcept <CreateTextLines>(new int[] { }, page =>
page.ParsePdf <ProcessPdfValidation>()
//.Show(Color.White)
.ParseBlock <IdentifyValidationMarks>()
.PdfCheck <CheckNoBlockSetOverlap>(Color.Orange)
//.Show(Color.Blue)
).ToList();
pipeline.SaveOk($"{outputfolder}/{basename}-ok.pdf");
int errors = pipeline.SaveErrors($"{outputfolder}/errors/{basename}-errors.pdf");
pipeline.Done();
return(errors);
}
public static void ExtractPages(string basename, string outputname, IList <int> pages)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.ExtractPages($"bin/{outputname}.pdf", pages);
}
public static void ValidateResizeBlock(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-test-output.pdf")
.Page(1)
.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.ParsePdf <PreProcessImages>()
.Validate <RemoveOverlapedImages>().ShowErrors(p => p.Show(Color.Red))
.ParseBlock <RemoveOverlapedImages>()
.ParsePdf <ProcessPdfText>()
.ParseBlock <RemoveTableText>()
.ParseBlock <GroupLines>()
.ParseBlock <FindInitialBlockset>()
.ParseBlock <BreakColumns>()
//.Validate<RemoveFooter>().ShowErrors(p => p.Show(Color.Purple))
//.Validate<RemoveHeaderImage>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveFooter>()
.ParseBlock <RemoveHeaderImage>()
.ParseBlock <AddTableSpace>()
.ParseBlock <AddImageSpace>()
//.ParseBlock<BreakInlineElements>()
.Validate <ResizeBlocksets>().ShowErrors(p => p.Show(Color.Red));
pipeline.Done();
}
public static void ExtractPage(string basename, int page)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Extract($"bin/{basename}-p{page}.pdf", page, page);
}
static PipelineText <TextLine> GetTextLines(Execution.Pipeline pipeline, string basename, string inputfolder, string outputfolder)
{
string inputfile = $"{inputfolder}/{basename}.pdf";
string outputfile = $"{outputfolder}/{basename}-parser.pdf";
return(Examples.GetTextLines(pipeline, inputfile, outputfile));
}
public static void CorrectOrder(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"{_inputFolder}/{basename}.pdf")
.Output($"{pdfsDir}/09-orders-{basename}-output.pdf")
.Page(1)
.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.ParsePdf <PreProcessImages>()
.ParseBlock <RemoveOverlapedImages>()
.ParsePdf <ProcessPdfText>()
.ParseBlock <RemoveSmallFonts>()
.ParseBlock <MergeTableText>()
.ParseBlock <HighlightTextTable>()
.ParseBlock <RemoveTableText>()
.ParseBlock <GroupLines>()
.Validate <RemoveHeaderImage>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveHeaderImage>()
.ParseBlock <FindInitialBlocksetWithRewind>()
.ParseBlock <BreakColumnsLight>()
.Validate <RemoveFooter>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveFooter>()
.ParseBlock <AddTableSpace>()
.ParseBlock <AddImageSpace>()
.ParseBlock <BreakInlineElements>()
.ParseBlock <ResizeBlocksets>()
.Validate <ResizeBlocksets>().ShowErrors(p => p.Show(Color.Red))
.ParseBlock <OrderBlocksets>()
.Show(Color.Orange)
.ShowLine(Color.Black);
pipeline.Done();
}
public static void ShowHeaderFooter(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-header-footer-output.pdf")
.AllPages(page =>
{
page.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.Show(Color.Red)
.ParsePdf <PreProcessImages>()
.Show(Color.Green)
.ParseBlock <RemoveOverlapedImages>()
.ParsePdf <ProcessPdfText>()
.ParseBlock <RemoveSmallFonts>()
.ParseBlock <MergeTableText>()
.ParseBlock <HighlightTextTable>()
.ParseBlock <RemoveTableText>()
.ParseBlock <GroupLines>()
.Validate <RemoveHeaderImage>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveHeaderImage>()
.ParseBlock <FindInitialBlocksetWithRewind>()
.ParseBlock <BreakColumnsLight>()
//.ParseBlock<BreakColumns>()
.Validate <RemoveFooter>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveFooter>();
});
pipeline.Done();
}
static void ExtractPages2(string basename, string outputname, IList <int> pages)
{
using (var pipeline = new Execution.Pipeline())
{
pipeline.Input($"{basename}.pdf")
.ExtractPages($"{outputname}.pdf", pages);
}
}
public static void MultipageCore(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-page-output.pdf")
.AllPages(p => ProcessPage(p));
pipeline.Done();
}
public static IEnumerable <TextLine> GetEnumerableLines(string basename)
{
var pipeline = new Execution.Pipeline();
var result =
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-test-output.pdf")
.StreamConvert <CreateTextLines>(ProcessPage);
return(result);
}
public static PipelineText <TextLine> GetTextLinesWithPipeline(string basename, out Execution.Pipeline pipeline)
{
pipeline = new Execution.Pipeline();
var result =
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-test-output.pdf")
.AllPages <CreateTextLines>(ProcessPage);
return(result);
}
public static PipelineText <TextLine> GetTextLines(Execution.Pipeline pipeline, string inputname, string outputname)
{
var result =
pipeline.Input(inputname)
.Output(outputname)
.AllPages <CreateTextLines>(page =>
page.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>() // 1
.ParsePdf <PreProcessImages>()
.ParseBlock <BasicFirstPageStats>() // 2
.ParseBlock <RemoveOverlapedImages>() // 3
.ParsePdf <ProcessPdfText>() // 4
.ParseBlock <RemoveSmallFonts>() // 5
.ParseBlock <MergeTableText>() // 6
.ParseBlock <HighlightTextTable>() // 7
.ParseBlock <RemoveTableText>() // 8
.ParseBlock <ReplaceCharacters>() // 9
.ParseBlock <GroupLines>() // 10
.ParseBlock <RemoveTableDotChar>() // 11
.Show(Color.Yellow)
.Validate <RemoveHeaderImage>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveHeaderImage>() // 12
.ParseBlock <FindInitialBlocksetWithRewind>() // 13
.Show(Color.Gray)
.ParseBlock <BreakColumnsLight>() // 14
.ParseBlock <AddTableSpace>() // 15
.ParseBlock <RemoveTableOverImage>() // 16
.ParseBlock <RemoveImageTexts>() // 17
.ParseBlock <AddImageSpace>() // 18
.Validate <RemoveFooter>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveFooter>() // 19
.ParseBlock <AddTableHorizontalLines>() // 20
.ParseBlock <RemoveBackgroundNonText>() // 21
.ParseBlock <BreakColumnsRewrite>() // 22
.ParseBlock <BreakInlineElements>() // 23
.ParseBlock <ResizeBlocksets>() // 24
.ParseBlock <ResizeBlocksetMagins>() // 25
.ParseBlock <OrderBlocksets>() // 26
.ParseBlock <OrganizePageLayout>() // 27
.ParseBlock <MergeSequentialLayout>() // 28
.ParseBlock <ResizeSequentialLayout>() // 29
.Show(Color.Orange)
.ShowLine(Color.Black)
.ParseBlock <CheckOverlap>() // 30
.Validate <CheckOverlap>().ShowErrors(p => p.Show(Color.Red))
.Validate <ValidatePositiveCoordinates>().ShowErrors(p => p.Show(Color.Red))
.PrintWarnings()
);
return(result);
}
public static void MarkAllComponents(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-tmp-output.pdf")
.Page(1)
.ParsePdf <ProcessPdfText>()
.Show(Color.Yellow);
pipeline.Done();
}
public static void Blocks(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"{_inputFolder}/{basename}.pdf")
.Output($"{pdfsDir}/01-blocks-{basename}-output.pdf")
.Page(1)
.ParsePdf <ProcessPdfText>()
.Show(Color.Orange);
pipeline.Done();
}
public static void ProcessImages(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-img-output.pdf")
.Page(1)
.ParsePdf <PreProcessImages>()
.Show(Color.Red);
pipeline.Done();
}
public static PipelineText <TextStructure> GetTextParagraphs(string basename)
{
var pipeline = new Execution.Pipeline();
var result =
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-test-output.pdf")
.AllPages <CreateTextLines>(ProcessPage)
.ConvertText <CreateStructures, TextStructure>();
return(result);
}
public static void ShowRenderPath(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-tmp-output.pdf")
.Page(1)
.ParsePdf <PreProcessRenderPath>()
.ShowLine(Color.Green);
pipeline.Done();
}
public static void GroupLines(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf").Page(1)
.Output($"bin/{basename}-tmp-output.pdf")
.ParsePdf <ProcessPdfText>()
.ParseBlock <GroupLines>()
.Show(Color.Orange);
pipeline.Done();
}
public static void FollowText(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-follow-text-output.pdf")
.AllPages(page => page
.ParsePdf <ProcessPdfText>()
.ShowLine(Color.Orange)
);
pipeline.Done();
}
public static void BlockLines(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"{_inputFolder}/{basename}.pdf")
.Output($"{pdfsDir}/02-blockline-{basename}-output.pdf")
.Page(1)
.ParsePdf <ProcessPdfText>()
.ParseBlock <GroupLines>()
.Show(Color.Red);
pipeline.Done();
}
public static void FindIds(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"{_inputFolder}/{basename}.pdf")
.Output($"{pdfsDir}/10-ids-{basename}-output.pdf")
.AllPages(page =>
{
page.ParsePdf <ProcessPdfText>()
.Validate <RemoveSmallFonts>().ShowErrors(p => p.ShowText(Color.Green));
});
pipeline.Done();
}
public static void ShowTables(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-table-output.pdf")
.Page(1)
.ParsePdf <PreProcessTables>()
.Show(Color.Yellow)
.ParseBlock <IdentifyTables>()
.Show(Color.Green);
pipeline.Done();
}
public static void ShowTables(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-tables.pdf")
.AllPages(page => page.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.Show(Color.Green)
.Validate <CheckOverlap>().ShowErrors(b => b.Show(Color.Red))
);
pipeline.Done();
}
public static void FollowText(IVirtualFS virtualFS, string basename)
{
VirtualFS.ConfigureFileSystem(virtualFS);
var pipeline = new Execution.Pipeline();
pipeline.Input($"{basename}.pdf")
.Output($"{basename}-follow-text-output.pdf")
.AllPages(page => page
.ParsePdf <ProcessPdfText>()
.ShowLine(Color.Orange)
);
pipeline.Done();
}
public static void FollowLine(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"{_inputFolder}/{basename}.pdf")
.Output($"{pdfsDir}/04-followline-{basename}-output.pdf")
.Page(1)
.ParsePdf <ProcessPdfText>()
.ParseBlock <GroupLines>()
.ShowLine(Color.Green)
.ParseBlock <FindInitialBlockset>()
.Show(Color.Orange);
pipeline.Done();
}
public static void RemoveOverlapedImages(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-tmp-output.pdf")
.Page(1)
//.ParsePdf<PreProcessTables>()
// .ParseBlock<IdentifyTables>()
.ParsePdf <PreProcessImages>()
.Validate <RemoveOverlapedImages>().ShowErrors(p => p.Show(Color.Red))
.ParseBlock <RemoveOverlapedImages>()
.Show(Color.Green);
pipeline.Done();
}
static PipelineText <TextLine> GetTextLines(string basename, string inputfolder, string outputfolder, out Execution.Pipeline pipeline)
{
pipeline = new Execution.Pipeline();
var result =
pipeline.Input($"{inputfolder}/{basename}.pdf")
.Output($"{outputfolder}/{basename}-output.pdf")
.AllPagesExcept <CreateTextLines>(new int[] { }, page =>
page.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.ParsePdf <PreProcessImages>()
.ParseBlock <RemoveOverlapedImages>()
.ParsePdf <ProcessPdfText>()
.Validate <RemoveSmallFonts>().ShowErrors(p => p.ShowText(Color.Green))
.ParseBlock <RemoveSmallFonts>()
.ParseBlock <MergeTableText>()
.ParseBlock <HighlightTextTable>()
.ParseBlock <RemoveTableText>()
.ParseBlock <ReplaceCharacters>()
.ParseBlock <GroupLines>()
.ParseBlock <RemoveTableDotChar>()
.Show(Color.Yellow)
.Validate <RemoveHeaderImage>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveHeaderImage>()
.ParseBlock <FindInitialBlocksetWithRewind>()
.Show(Color.Gray)
.ParseBlock <BreakColumnsLight>()
//.ParseBlock<BreakColumns>()
.ParseBlock <AddTableSpace>()
.ParseBlock <RemoveTableOverImage>()
.ParseBlock <RemoveImageTexts>()
.ParseBlock <AddImageSpace>()
.Validate <RemoveFooter>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveFooter>()
.ParseBlock <BreakInlineElements>()
.ParseBlock <ResizeBlocksets>()
.Validate <ResizeBlocksets>().ShowErrors(p => p.Show(Color.Gray))
.ParseBlock <OrderBlocksets>()
.Show(Color.Orange)
.ShowLine(Color.Black)
.ParseBlock <OrganizePageLayout>()
.ParseBlock <CheckOverlap>()
.Validate <ValidatePositiveCoordinates>().ShowErrors(p => p.Show(Color.Red))
);
return(result);
}
public static void RemoveTables(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-table-output.pdf")
.Page(1)
.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.Show(Color.Green)
.ParsePdf <ProcessPdfText>()
.ParseBlock <RemoveTableText>()
.ParseBlock <GroupLines>()
.Show(Color.Red);
pipeline.Done();
}
static PipelineText <TextLine> GetTextLinesWithPipelineBlockset(string basename, out Execution.Pipeline pipeline)
{
pipeline = new Execution.Pipeline();
var result =
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-test-output.pdf")
.AllPagesExcept <CreateTextLines>(new int[] { }, page =>
page.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.ParsePdf <PreProcessImages>()
.Validate <RemoveOverlapedImages>().ShowErrors(p => p.Show(Color.Red))
.ParseBlock <RemoveOverlapedImages>()
.ParsePdf <ProcessPdfText>()
.Validate <RemoveSmallFonts>().ShowErrors(p => p.ShowText(Color.Green))
.ParseBlock <RemoveSmallFonts>()
//.Validate<MergeTableText>().ShowErrors(p => p.Show(Color.Blue))
.ParseBlock <MergeTableText>()
//.Validate<HighlightTextTable>().ShowErrors(p => p.Show(Color.Green))
.ParseBlock <HighlightTextTable>()
.ParseBlock <RemoveTableText>()
.ParseBlock <GroupLines>()
.Show(Color.Yellow)
.Validate <RemoveHeaderImage>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveHeaderImage>()
.ParseBlock <FindInitialBlocksetWithRewind>()
.Show(Color.Gray)
.ParseBlock <BreakColumnsLight>()
//.ParseBlock<BreakColumns>()
.Validate <RemoveFooter>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveFooter>()
.ParseBlock <AddTableSpace>()
.ParseBlock <AddImageSpace>()
.ParseBlock <BreakInlineElements>()
.ParseBlock <ResizeBlocksets>()
.Validate <ResizeBlocksets>().ShowErrors(p => p.Show(Color.Red))
.ParseBlock <OrderBlocksets>()
.Show(Color.Orange)
.ShowLine(Color.Black)
);
return(result);
}
