void DetectRowBounds(PageLayout cbi, Blob[] blobs)
{
if (blobs.Length == 0) { return; }
if (cbi.Bounds == Rectangle.Empty) { return; }
List<LayoutElement> rows = new List<LayoutElement>();
LayoutElement currentRow = null;
// Attempt drawing lines between the rows.
for (int y = cbi.Bounds.Top; y < cbi.Bounds.Bottom; y++)
{
Rectangle rowRect = new Rectangle(cbi.Bounds.Left, y, cbi.Bounds.Width, 1);
var blobsInRow = blobs.Where(b => b.Rectangle.IntersectsWith(rowRect));
if (blobsInRow.FirstOrDefault() == null)
{
// Empty row detected. Commit current row (if any)
TryAddRow(rows, currentRow);
currentRow = null;
}
else
{
// Start new row if needed
if (currentRow == null)
{
currentRow = new LayoutElement();
currentRow.Type = LayoutElementType.Row;
}
currentRow.Children.AddRange(blobsInRow.Select(x => LayoutElement.NewWord(cbi.PageSize, x.Rectangle)));
// Advance to test the next empty space
// TODO: beware of off-by-1
//y = currentRow.Bounds.Bottom - 1;
}
}
// Add row at the end
TryAddRow(rows, currentRow);
FindAndRemoveHeaderAndFooter(cbi, rows);
cbi.Children = rows;
cbi.SetBoundsFromNodes(true);
}
PageLayout DetectLayout(DW<PDFWrapper> doc, PDFPage page)
{
Size pageSize = new Size(doc.o.PageWidth, doc.o.PageHeight);
PageLayout layout = new PageLayout(pageSize);
// Get text
// TODO: check how text is split in multicolumn case -- is this the method with correct options (flow, not physical)
layout.Text = page.Text;
var words = new List<LayoutElement>();
var nonEmptyWords = page.WordList
.Where(x => !x.Bounds.IsEmpty && !x.Word.IsEmpty())
.Select(x => LayoutElement.NewWord(pageSize, x.Bounds, x.Word));
words.AddRange(nonEmptyWords);
// Detect rows and columns
var rows = words.Split(StartsNewRow).Select(ws => LayoutElement.NewRow(ws, LayoutElementType.Row));
var cols = rows.Split(StartsNewColumn).Select(rs => LayoutElement.NewRow(rs, LayoutElementType.Column));
// TODO: detect header/footer
layout.Children.AddRange(cols);
// Strange bug -- if doing the following, first word is missing and last word is blank.
// However, with LINQ query above it's fine
//List<PDFTextWord> ws = new List<PDFTextWord>();
//ws.AddRange(page.WordList);
if (layout.Children.Count > 0)
{
layout.SetBoundsFromNodes(true);
/*
// expand by width a bit (to prevent cutting off words which
// may not be recognized properly.
int expandWidth = (0.05 * layout.Bounds.Width).Round();
RectangleF expBounds = layout.UnitBounds;
expBounds.X -= expandWidth / 2;
expBounds.Width += expandWidth;
if (expBounds.X <= 1 && expBounds.Width <= 1)
{
layout.UnitBounds = expBounds;
}
*/
}
// error checking
if (layout.Bounds.X < 0 || layout.Bounds.Y < 0 ||
layout.Bounds.Width <= 0 || layout.Bounds.Height <= 0)
{
logger.Error("Wrong bounds: " + layout.Bounds + " images: " + page.ImagesCount);
float height = page.ImagesCount > 0 ? 1 : 0.1f;
layout.UnitBounds = new RectangleF(0,0, 1, height);
}
// TODO: detect rows
// TODO: detect header/footer (if any)
return layout;
}