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 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 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 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 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();
}
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 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 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 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 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 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 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}-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 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 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();
}
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 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 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();
}
public static void ProcessPage1(string basename, string inputfolder, string outputfolder)
{
//PdfReaderException.DisableWarnings();
//PdfReaderException.ContinueOnException();
var pipeline = new Execution.Pipeline();
var result =
pipeline.Input($"{inputfolder}/{basename}")
.Output($"{outputfolder}/{basename}-invalid.pdf")
.Page(1)
.ParsePdf <ProcessPdfValidation>()
.Show(Color.White)
.ParseBlock <IdentifyValidationMarks>()
.ParseBlock <CheckNoBlockSetOverlap>()
.Show(Color.Blue);
pipeline.Done();
}
public static void AddImageSpace(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-img-output.pdf")
.Page(1)
.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.ParsePdf <PreProcessImages>()
.ParsePdf <ProcessPdfText>()
.ParseBlock <RemoveTableText>()
.ParseBlock <GroupLines>()
.ParseBlock <FindInitialBlockset>()
.ParseBlock <AddTableSpace>()
.ParseBlock <AddImageSpace>()
.Show(Color.Orange);
pipeline.Done();
}
public static void BreakColumn(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"{_inputFolder}/{basename}.pdf")
.Output($"{pdfsDir}/05-breakcolumn-{basename}-output.pdf")
.Page(1)
.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.ParsePdf <PreProcessImages>()
.ParseBlock <RemoveOverlapedImages>()
.ParsePdf <ProcessPdfText>()
.ParseBlock <GroupLines>()
.Validate <RemoveHeaderImage>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveHeaderImage>()
.ParseBlock <FindInitialBlocksetWithRewind>()
.ParseBlock <BreakColumnsLight>()
.Show(Color.Orange);
pipeline.Done();
}
public static void Process(string basename, string inputfolder, string outputfolder)
{
BasicFirstPageStats.Reset();
PdfReaderException.ContinueOnException();
var pipeline = new Execution.Pipeline();
var artigos = GetTextLines(pipeline, basename, inputfolder, outputfolder)
.ConvertText <CreateTextLineIndex, TextLine>()
.ConvertText <PreCreateStructures, TextLine2>()
.ConvertText <CreateStructures2, TextStructure>()
.ConvertText <PreCreateTextSegments, TextStructureAgg>()
.ConvertText <AggregateStructures, TextStructure>()
.ShowPdf <ShowStructureCentral>($"{outputfolder}/{basename}-show-central.pdf")
.ConvertText <CreateTextSegments, TextSegment>()
.ConvertText <CreateTreeSegments, TextSegment>()
.Log <AnalyzeTreeStructure>($"{outputfolder}/{basename}-tree.txt")
.ToList();
pipeline.ExtractOutput <ShowParserWarnings>($"{outputfolder}/errors/{basename}-parser-errors.pdf");
pipeline.Done();
}
//public static void ValidateBreakColumns(string basename)
//{
// var pipeline = new Execution.Pipeline();
// pipeline.Input($"bin/{basename}.pdf").Page(1)
// .Output($"bin/{basename}-tmp-output.pdf")
// .ParsePdf<ProcessPdfText>()
// .ParseBlock<GroupLines>()
// .ParseBlock<FindInitialBlockset>()
// .Validate<BreakColumns>()
// .ShowErrors(p => p.Show(Color.Purple));
// pipeline.Done();
//}
//public static void BreakColumns(string basename)
//{
// var pipeline = new Execution.Pipeline();
// pipeline.Input($"bin/{basename}.pdf").Page(1)
// .Output($"bin/{basename}-tmp-output.pdf")
// .ParsePdf<ProcessPdfText>()
// .ParseBlock<GroupLines>()
// .ParseBlock<FindInitialBlockset>()
// .Validate<BreakColumns>().ShowErrors(p => p.Show(Color.LightGray))
// .ParseBlock<BreakColumns>()
// .Show(Color.Green)
// .Validate<BreakColumns>().ShowErrors(p => p.Show(Color.Red));
// pipeline.Done();
//}
//public static void RemoveHeaderFooter(string basename)
//{
// var pipeline = new Execution.Pipeline();
// pipeline.Input($"bin/{basename}.pdf").Page(1)
// .Output($"bin/{basename}-tmp-output.pdf")
// .ParsePdf<ProcessPdfText>()
// .ParseBlock<GroupLines>()
// .ParseBlock<FindInitialBlockset>()
// .ParseBlock<BreakColumns>()
// .Validate<RemoveFooter>().ShowErrors(p => p.Show(Color.Purple))
// .Validate<RemoveHeader>().ShowErrors(p => p.Show(Color.Purple))
// .ParseBlock<RemoveFooter>()
// .ParseBlock<RemoveHeader>()
// .Show(Color.Yellow);
// pipeline.Done();
//}
public static void MergeBlockLines(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf").Page(1)
.Output($"bin/{basename}-tmp-output.pdf")
.ParsePdf <ProcessPdfText>()
.ParseBlock <GroupLines>()
.ParseBlock <FindInitialBlockset>()
//.ParseBlock<TestSplitBlocksets>()
//.Show(Color.Red)
.ShowLine(Color.Gray)
.ParseBlock <MergeBlockLines>()
.Show(Color.Green)
//.ParseBlock<BreakColumns>()
.Validate <RemoveFooter>().ShowErrors(p => p.Show(Color.Purple))
.Validate <RemoveHeader>().ShowErrors(p => p.Show(Color.Purple))
.ParseBlock <RemoveFooter>()
.ParseBlock <RemoveHeader>();
//.Show(Color.Yellow);
pipeline.Done();
}
public static void RemoveHeaderImage(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-img-output.pdf")
.Page(1)
.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.ParsePdf <PreProcessImages>()
.ParsePdf <ProcessPdfText>()
.ParseBlock <RemoveTableText>()
.ParseBlock <GroupLines>()
.ParseBlock <FindInitialBlockset>()
.Validate <RemoveFooter>().ShowErrors(p => p.Show(Color.Orange))
//.Validate<RemoveHeader>().ShowErrors(p => p.Show(Color.Orange))
.Validate <RemoveHeaderImage>().ShowErrors(p => p.Show(Color.Orange))
.ParseBlock <RemoveFooter>()
//.ParseBlock<RemoveHeader>()
.ParseBlock <RemoveHeaderImage>()
.Show(Color.Yellow);
pipeline.Done();
}
public static void DetectInvisibleTable(string basename)
{
var pipeline = new Execution.Pipeline();
pipeline.Input($"bin/{basename}.pdf")
.Output($"bin/{basename}-table-output.pdf")
.Page(1)
.ParsePdf <PreProcessTables>()
.ParseBlock <IdentifyTables>()
.ParsePdf <PreProcessImages>()
.Validate <RemoveOverlapedImages>().ShowErrors(p => p.Show(Color.Red))
.ParseBlock <RemoveOverlapedImages>()
.ParsePdf <ProcessPdfText>()
.ParseBlock <RemoveTableText>()
.ParseBlock <GroupLines>()
.ParseBlock <FindInitialBlockset>()
.Show(Color.Yellow)
.ParseBlock <DetectImplicitTable>()
.Show(Color.Green)
.Validate <DetectImplicitTable>().ShowErrors(p => p.Show(Color.Red))
.ShowLine(Color.Black);
pipeline.Done();
}