def processRegions(ndPage, bVerbose=False):
"""
Delete empty regions
resize no empty regions
"""
lDel = []
lndRegions = ndPage.xpath(".//pg:TextRegion", namespaces=dNS)
for ndRegion in lndRegions:
lTL = ndRegion.xpath(".//pg:TextLine", namespaces=dNS)
if lTL == []:
# to be deleted
lDel.append(ndRegion)
else:
#resize it
oHull = ShapeLoader.convex_hull(lTL, bShapelyObject=True)
PageXml.getChildByName(ndRegion, 'Coords')[0].set(
"points", ShapeLoader.getCoordsString(oHull, bFailSafe=True))
# contour = cascaded_union([p if p.is_valid else p.convex_hull for p in lTL ])
# o = contour.minimum_rotated_rectangle
# ndRegion.getChildByName('Coords').set("points", ShapeLoader.getCoordsString(o, bFailSafe=True))
# delete empty regions
[ndRegion.getparent().remove(ndRegion) for ndRegion in lDel]
if bVerbose:
traceln(" - %d regions deleted" % (len(lDel)))
traceln(" - %d regions updated" % (len(lndRegions) - len(lDel)))
def computeShape(cls, ndPage, setID, bConvexHull=False):
"""
compute a shape for this cluster, as the minimum rotated rectangle of its content
or optionally as the convex hull
"""
# let's find the nodes and compute the shape
lNode = [
ndPage.xpath(".//*[@id='%s']" % _id, namespaces=dNS)[0]
for _id in setID
]
return ShapeLoader.convex_hull(lNode, bShapelyObject=True) \
if bConvexHull \
else ShapeLoader.minimum_rotated_rectangle(lNode, bShapelyObject=True)
def makeTableNode(self):
"""
Make a DOM tree for this table
"""
lK = self._dCellNd.keys()
lRow = list(set(_row for _row, _col in lK))
lRow.sort()
lCol = list(set(_col for _row, _col in lK))
lCol.sort()
ndTable = PageXml.createPageXmlNode("TableRegion")
ndTable.set("id", "p%s_%s" % (self.pagenum, self.tablenum))
ndTable.tail = "\n"
lCellShape = []
lNdCell = []
for row in lRow:
for col in lCol:
lNdText = self._dCellNd[(row, col)]
# <TableCell row="0" col="1" rowSpan="1" colSpan="1" id="TableCell_1500971530732_2485">
# <Coords points="480,42 485,323 878,323 874,38"/>
if lNdText:
ndCell = PageXml.createPageXmlNode("TableCell")
ndCell.set(
"id", "p%s_t%s_r%s_c%s" %
(self.pagenum, self.tablenum, row, col))
# shape of the cell
oHull = ShapeLoader.convex_hull(lNdText,
bShapelyObject=True)
lCellShape.append(
oHull) # keep those to compute table contour
# Coords sub-element
ndCoords = PageXml.createPageXmlNode("Coords")
ndCoords.set(
"points",
ShapeLoader.getCoordsString(oHull, bFailSafe=True))
ndCoords.tail = "\n"
ndCell.append(ndCoords)
# row="0" col="0" rowSpan="1" colSpan="1" leftBorderVisible="false" rightBorderVisible="false" topBorderVisible="false" bottomBorderVisible="false"
ndCell.set("row", str(row))
ndCell.set("rowSpan", "1")
ndCell.set("col", str(col))
ndCell.set("colSpan", "1")
ndCell.tail = "\n"
#add corner
cornerNode = PageXml.createPageXmlNode("CornerPts")
cornerNode.text = "0 1 2 3"
ndCell.append(cornerNode)
for nd in lNdText:
ndCell.append(nd)
lNdCell.append(ndCell)
# Table geometry
ndCoords = PageXml.createPageXmlNode("Coords")
contour = cascaded_union(
[p if p.is_valid else p.convex_hull for p in lCellShape])
o = contour.minimum_rotated_rectangle
ndCoords.set("points", ShapeLoader.getCoordsString(o, bFailSafe=True))
ndCoords.tail = "\n"
ndTable.append(ndCoords)
for nd in lNdCell:
ndTable.append(nd)
return ndTable