/// <summary>
/// Get the blocks.
/// </summary>
/// <param name="pagePaths">The words in the page.</param>
/// <param name="minimumWidth">The minimum width for a block.</param>
/// <param name="dominantFontWidthFunc">The function that determines the dominant font width.</param>
/// <param name="dominantFontHeightFunc">The function that determines the dominant font height.</param>
public IReadOnlyList <PdfRectangle> GetBlocks(IEnumerable <PdfPath> pagePaths, decimal minimumWidth,
decimal dominantFontWidth, decimal dominantFontHeight)
{
pagePaths = pagePaths.Where(p => p != null && p.Commands.Count > 0); // clean paths
XYLeafP root = new XYLeafP(pagePaths); // Create a root node.
XYNodeP node = VerticalCut(root, minimumWidth, dominantFontWidth, dominantFontHeight);
var leafs = node.GetLeafs();
if (leafs.Count > 0)
{
return(leafs.Select(l => l.BoundingBox).ToList()); // new TextBlock(l.GetLines())).ToList();
}
return(new List <PdfRectangle>());
}
private XYNodeP VerticalCut(XYLeafP leaf, decimal minimumWidth,
decimal dominantFontWidth, decimal dominantFontHeight, int level = 0)
{
if (leaf.CountWords() <= 1 || leaf.BoundingBox.Width <= minimumWidth)
{
// we stop cutting if
// - only one word remains
// - width is too small
return(leaf);
}
// order words left to right
var paths = leaf.Paths.OrderBy(w => w.BoundingBox().Left).ToArray();
// determine dominantFontWidth and dominantFontHeight
//decimal domFontWidth = dominantFontWidthFunc(paths.SelectMany(x => x.Letters)
// .Select(x => Math.Abs(x.GlyphRectangle.Width)));
//decimal domFontHeight = dominantFontHeightFunc(paths.SelectMany(x => x.Letters)
// .Select(x => Math.Abs(x.GlyphRectangle.Height)));
List <decimal[]> projectionProfile = new List <decimal[]>();
decimal[] currentProj = new decimal[2] {
paths[0].BoundingBox().Left, paths[0].BoundingBox().Right
};
int wordsCount = paths.Count();
for (int i = 1; i < wordsCount; i++)
{
if ((paths[i].BoundingBox().Left >= currentProj[0] && paths[i].BoundingBox().Left <= currentProj[1]) ||
(paths[i].BoundingBox().Right >= currentProj[0] && paths[i].BoundingBox().Right <= currentProj[1]))
{
// it is overlapping
if (paths[i].BoundingBox().Left >= currentProj[0] &&
paths[i].BoundingBox().Left <= currentProj[1] &&
paths[i].BoundingBox().Right > currentProj[1])
{
// |____|
// |____|
// |_______| <- updated
currentProj[1] = paths[i].BoundingBox().Right;
}
// we ignore the following cases:
// |____|
// |____| (not possible because of OrderBy)
//
// |____|
//|___________| (not possible because of OrderBy)
//
// |____|
// |_|
}
else
{
// no overlap
if (paths[i].BoundingBox().Left - currentProj[1] <= dominantFontWidth)
{
// if gap too small -> don't cut
// |____| |____|
currentProj[1] = paths[i].BoundingBox().Right;
}
else if (currentProj[1] - currentProj[0] < minimumWidth)
{
// still too small
currentProj[1] = paths[i].BoundingBox().Right;
}
else
{
// if gap big enough -> cut!
// |____| | |____|
if (i != wordsCount - 1) // will always add the last one after
{
projectionProfile.Add(currentProj);
currentProj = new decimal[2] {
paths[i].BoundingBox().Left, paths[i].BoundingBox().Right
};
}
}
}
if (i == wordsCount - 1)
{
projectionProfile.Add(currentProj);
}
}
var newLeafsEnums = projectionProfile
.Select(p => leaf.Paths.Where(w => w.BoundingBox().Left >= p[0] && w.BoundingBox().Right <= p[1]));
var newLeafs = newLeafsEnums.Where(e => e.Count() > 0).Select(e => new XYLeafP(e));
var newNodes = newLeafs.Select(l => HorizontalCut(l, minimumWidth,
dominantFontWidth, dominantFontHeight, level)).ToList();
var lost = leaf.Paths.Except(newLeafsEnums.SelectMany(x => x)).ToList();
if (lost.Count > 0)
{
newNodes.AddRange(lost.Select(w => new XYLeafP(w)));
}
return(new XYNodeP(newNodes));
}
private XYNodeP HorizontalCut(XYLeafP leaf, decimal minimumWidth,
decimal dominantFontWidth, decimal dominantFontHeight, int level = 0)
{
if (leaf.CountWords() <= 1)
{
// we stop cutting if
// - only one word remains
return(leaf);
}
var words = leaf.Paths.OrderBy(w => w.BoundingBox().Bottom).ToArray(); // order bottom to top
// determine dominantFontWidth and dominantFontHeight
//decimal dominantFontWidth = dominantFontWidthFunc(words.SelectMany(x => x.Letters)
// .Select(x => Math.Abs(x.GlyphRectangle.Width)));
//decimal dominantFontHeight = dominantFontHeightFunc(words.SelectMany(x => x.Letters)
// .Select(x => Math.Abs(x.GlyphRectangle.Height)));
List <decimal[]> projectionProfile = new List <decimal[]>();
decimal[] currentProj = new decimal[2] {
words[0].BoundingBox().Bottom, words[0].BoundingBox().Top
};
int wordsCount = words.Count();
for (int i = 1; i < wordsCount; i++)
{
if ((words[i].BoundingBox().Bottom >= currentProj[0] && words[i].BoundingBox().Bottom <= currentProj[1]) ||
(words[i].BoundingBox().Top >= currentProj[0] && words[i].BoundingBox().Top <= currentProj[1]))
{
// it is overlapping
if (words[i].BoundingBox().Bottom >= currentProj[0] &&
words[i].BoundingBox().Bottom <= currentProj[1] &&
words[i].BoundingBox().Top > currentProj[1])
{
currentProj[1] = words[i].BoundingBox().Top;
}
}
else
{
// no overlap
if (words[i].BoundingBox().Bottom - currentProj[1] <= dominantFontHeight)
{
// if gap too small -> don't cut
// |____| |____|
currentProj[1] = words[i].BoundingBox().Top;
}
else
{
// if gap big enough -> cut!
// |____| | |____|
if (i != wordsCount - 1) // will always add the last one after
{
projectionProfile.Add(currentProj);
currentProj = new decimal[2] {
words[i].BoundingBox().Bottom, words[i].BoundingBox().Top
};
}
}
}
if (i == wordsCount - 1)
{
projectionProfile.Add(currentProj);
}
}
if (projectionProfile.Count == 1)
{
if (level >= 1)
{
return(leaf);
}
else
{
level++;
}
}
var newLeafsEnums = projectionProfile.Select(p =>
leaf.Paths.Where(w => w.BoundingBox().Bottom >= p[0] && w.BoundingBox().Top <= p[1]));
var newLeafs = newLeafsEnums.Where(e => e.Count() > 0).Select(e => new XYLeafP(e));
var newNodes = newLeafs.Select(l => VerticalCut(l, minimumWidth,
dominantFontWidth, dominantFontHeight, level)).ToList();
var lost = leaf.Paths.Except(newLeafsEnums.SelectMany(x => x)).ToList();
if (lost.Count > 0)
{
newNodes.AddRange(lost.Select(w => new XYLeafP(w)));
}
return(new XYNodeP(newNodes));
}
