My company, Atalasoft, has just released some PDF manipulation tools that run on .NET. There is a text extract class that you can use to find the text and determine how you will split your document and a very high level document class that makes the splitting trivial. Suppose you have a Stream to your source PDF and an increasingly ordered List that describes the starting page of each split, then the code to generate your split files looks like this:
public void SplitPdf(Stream stm, List<int> pageStarts, string outputDirectory)
{
PdfDocument mainDoc = new PdfDocument(stm);
int lastPage = mainDoc.Pages.Count - 1;
for (int i=0; i < pageStarts.Count; i++) {
int startPage = pageStarts[i];
int endPage= (i < pageStarts.Count - 1) ?
pageStarts[i + 1] - 1 :
lastPage;
if (startPage > endPage) throw new ArgumentException("list is not ordered properly", "pageStarts");
PdfDocument splitDoc = new PdfDocument();
for (j = startPage; j <= endPage; j++)
splitDoc.Pages.Add(mainDoc.Pages[j];
string outputPath = Path.Combine(outputDirectory,
string.Format("{0:D3}.pdf", i + 1));
splitDoc.Save(outputPath);
}
if you generalize this into a page range struct:
public struct PageRange {
public int StartPage;
public int EndPage;
}
where StartPage and EndPage inclusively describe a range of pages, then the code is simpler:
public void SplitPdf(Stream stm, List<PageRange> ranges, string outputDirectory)
{
PdfDocument mainDoc = new PdfDocument(stm);
int outputDocCount = 1;
foreach (PageRange range in ranges) {
int startPage = Math.Min(range.StartPage, range.EndPage); // assume not in order
int endPage = Math.Max(range.StartPage, range.EndPage);
PdfDocument splitDoc = new PdfDocument();
for (int i=startPage; i <= endPage; i++)
splitDoc.Pages.Add(mainDoc.Pages[i]);
string outputPath = Path.Combine(outputDirectory,
string.Format("{0:D3}.pdf", outputDocCount));
splitDoc.Save(outputPath);
outputDocCount++;
}
}