public void Append(string content)
{
if (Html.Length > 0)
{
Html.Append("\r\n[...] ");
}
var tag = new HtmlTagSpan(new Span(Html.Length, Html.Length + content.Length - 1));
HtmlTags.Add(tag);
Html.Append(content);
}
private void Append(int pageIdx,
Span span,
StringBuilder str)
{
PdfPage page = Document.Pages[pageIdx];
PdfText pdfText = page.Text;
// Get all text objects
TextObject GetTextObject(PdfTextObject textObj)
{
var bbox = textObj.GetCharRect(0);
var absIdx = pdfText.GetCharIndexAtPos(bbox.left,
bbox.top,
1,
1);
return(new TextObject(textObj, absIdx));
}
var textObjects = page.PageObjects
.Where(o => o.ObjectType == PageObjectTypes.PDFPAGE_TEXT)
.Select(o => GetTextObject((PdfTextObject)o))
.OrderBy(t => t.StartIndex)
.ToList();
// Some PDF documents are improperly formatted and miss PdfTextObjects -- Fill the gaps
int lastEndIdx = textObjects.FirstOrDefault()?.EndIndex ?? 0;
for (int i = 1; i < textObjects.Count; i++)
{
var textObj = textObjects[i];
if (textObj.StartIndex <= lastEndIdx + 1) // This shouldn't be < -- But allow it nevertheless
{
lastEndIdx = textObj.EndIndex;
continue;
}
var gapTextObj = new TextObject(textObjects[i - 1],
lastEndIdx + 1,
textObj.StartIndex - lastEndIdx - 1);
textObjects.Insert(i++, gapTextObj);
lastEndIdx = textObj.EndIndex;
}
// Build the HTML tags
int shift = str.Length;
foreach (var textObj in textObjects)
{
// Check overlap
var objSpan = new Span(textObj.StartIndex,
textObj.StartIndex + textObj.Length - 1);
if (objSpan.Overlaps(span,
out var overlap) == false)
{
continue;
}
// Look behind for line return, and extend span for inclusion -- Unlike PdfTextObjects, GetText includes \r\n
int lookbackIdx = textObj.StartIndex - 2;
int tagStartIdxExtendBehind = 0;
if (lookbackIdx >= span.StartIdx &&
pdfText.GetText(lookbackIdx,
2) is "\r\n")
{
tagStartIdxExtendBehind = -2;
}
// Generate text object tag
var relStartIdx = shift + overlap.StartIdx - span.StartIdx;
var tag = new HtmlTagSpan(new Span(relStartIdx + tagStartIdxExtendBehind,
relStartIdx + overlap.Length - 1))
.WithStyle(s => SetTextStyle(s, textObj));
HtmlTags.Add(tag);
// Generate extract tag
if (OverlapsWithExtract(pageIdx,
overlap,
out var extractOverlaps))
{
foreach (var extractOverlap in extractOverlaps)
{
int extractStartIdx = shift + extractOverlap.StartIdx - span.StartIdx;
var extractSpan = new Span(extractStartIdx,
extractStartIdx + extractOverlap.Length - 1);
var extractTag = new HtmlTagSpan(extractSpan, 100);
extractTag.WithStyle(s => s.WithBackgroundColorColor(extractOverlap.Object));
HtmlTags.Add(extractTag);
}
}
}
str.Append(pdfText.GetText(span.StartIdx,
span.Length));
}
