def _iter_GraphNode(self, doc, domNdPage, page):
"""
Get the DOM, the DOM page node, the page object
iterator on the DOM, that returns nodes (of class Block)
"""
#--- XPATH contexts
assert self.sxpNode, "CONFIG ERROR: need an xpath expression to enumerate elements corresponding to graph nodes"
lNdBlock = domNdPage.xpath(self.sxpNode, namespaces=self.dNS) #all relevant nodes of the page
for ndBlock in lNdBlock:
domid = ndBlock.get("id")
sText = self._get_GraphNodeText(doc, domNdPage, ndBlock)
if sText == None:
sText = ""
traceln("Warning: no text in node %s"%domid)
#raise ValueError, "No text in node: %s"%ndBlock
#now we need to infer the bounding box of that object
lXY = PageXml.getPointList(ndBlock) #the polygon
if lXY == []:
continue
plg = Polygon(lXY)
try:
x1,y1, x2,y2 = plg.fitRectangle()
except ZeroDivisionError:
# traceln("Warning: ignoring invalid polygon id=%s page=%s"%(ndBlock.prop("id"), page.pnum))
# continue
# if True:
# #we reduce a bit this rectangle, to ovoid overlap
# w,h = x2-x1, y2-y1
# dx = max(w * 0.066, min(20, w/3)) #we make sure that at least 1/"rd of te width will remain!
# dy = max(h * 0.066, min(20, w/3))
# x1,y1, x2,y2 = [ int(round(v)) for v in [x1+dx,y1+dy, x2-dx,y2-dy] ]
x1,y1,x2,y2 = plg.getBoundingBox()
except ValueError:
x1,y1,x2,y2 = plg.getBoundingBox()
#we reduce a bit this rectangle, to ovoid overlap
if not(self.BBoxDeltaFun is None):
w,h = x2-x1, y2-y1
dx = self.BBoxDeltaFun(w)
dy = self.BBoxDeltaFun(h)
x1,y1, x2,y2 = [ int(round(v)) for v in [x1+dx,y1+dy, x2-dx,y2-dy] ]
#TODO
orientation = 0 #no meaning for PageXml
classIndex = 0 #is computed later on
#and create a Block
blk = Block(page, (x1, y1, x2-x1, y2-y1), sText, orientation, classIndex, self, ndBlock, domid=domid)
yield blk
raise StopIteration()
def convertTR2Sep(filename):
"""
"""
print (filename)
tagname='TextRegion'
xml = etree.parse(filename)
ltextsep = xml.getroot().findall(f".//pc:{tagname}[@custom]", {"pc":"http://schema.primaresearch.org/PAGE/gts/pagecontent/2013-07-15"})
for x in ltextsep:
if "separator" in x.get('custom'):
x.tag = 'SeparatorRegion'
#now we need to convert that object to a line
lXY = PageXml.getPointList(x) #the polygon
assert lXY, "Separator without Coord??"
plg = Polygon(lXY)
try:
x1,y1, x2,y2 = plg.fitRectangle()
except ValueError:
print("Warning: Coords might be bad, taking bounding box: ", lXY)
x1,y1,x2,y2 = plg.getBoundingBox()
# try:
# x1,y1, x2,y2 = plg.fitRectangle()
# except ZeroDivisionError:
# x1,y1,x2,y2 = plg.getBoundingBox()
# except ValueError:
# x1,y1,x2,y2 = plg.getBoundingBox()
if abs(x2-x1) > abs(y2-y1): # horizontal
y1 = (y1+y2)/2
y2 = y1
else:
x1 = (x1+x2)/2
x2=x1
ndCoord = x.xpath(".//pc:Coords", namespaces={"pc":PageXml.NS_PAGE_XML})[0]
PageXml.setPoints(ndCoord, [(x1,y1), (x2,y2)])
return xml
def resizeCell(self, cell, ns):
"""
replace the cell region by a BB for textlines: better for transcriber
"""
xpath = "./a:%s" % ("TextLine")
lTextLines = cell.xpath(xpath, namespaces={'a': PageXml.NS_PAGE_XML})
if lTextLines == []:
return True
## get minx,maxy for the cell
lXY = PageXml.getPointList(cell) #the polygon
plg = Polygon(lXY)
x1, y1, x2, y2 = plg.fitRectangle()
x1, y1, x2, y2 = plg.getBoundingBox()
cellX1 = x1
cellX2 = x2
cellY1 = y1
cellY2 = y2
## get all the textlines of the cell
minx, miny, maxx, maxy = 9e9, 9e9, 0, 0
for line in lTextLines:
lXY = PageXml.getPointList(line) #the polygon
# in case empty cell
if lXY == []:
continue
plg = Polygon(lXY)
try:
x1, y1, x2, y2 = plg.fitRectangle()
except ZeroDivisionError:
continue
x1, y1, x2, y2 = plg.getBoundingBox()
minx = min(minx, x1)
miny = min(miny, y1)
maxx = max(maxx, x2)
maxy = max(maxy, y2)
"""
finally: simply use the BB of the textlines + padding
"""
# # new request: height= max(cell,text)
# ## new new request: (12/10/2017): min (cell, text)!
# HCell = cellY2 - cellY1
# HBBText = maxy - miny
#
miny -= self.vpadding # vertical padding (top)
maxy += self.vpadding # vertical padding (bottom)
#
# # Height computation
# ## if HBBText <= self.HeightTH * HCell: take HBBText as height for TextREgion
# if HBBText > self.HeightTH * HCell:
# miny = max(miny,cellY1)
# maxy = min(maxy,cellY2)
# # else : don't touch miny, maxy : use only Text for computing Height
#
# # Width computation
minx -= self.hpadding # horizontal padding
maxx += self.hpadding # horizontal padding
# minx = min(cellX1,minx)
# maxx = max(cellX2, maxx)
# print cellX2, maxx
corner = cell[0]
# print minx,miny,maxx,miny,maxx,maxy,minx,maxy
corner.set(
'points', "%d,%d %d,%d %d,%d %d,%d" %
(minx, miny, maxx, miny, maxx, maxy, minx, maxy))
def _iter_GraphNode(self, doc, sFilename, page=None):
"""
Get the json dict
iterator on the DOM, that returns nodes (of class Block)
"""
# --- XPATH contexts
lNdBlock = doc['GlynchResults']['Areas']
#page_w, page_h = self.getPageWidthandHeight(sFilename)
page = Page(1, 1, 1, 1)
for ndBlock in lNdBlock:
try:
sText = ndBlock['label']
except KeyError:
sText = ""
traceln("WARNING: no 'label' in : %s" % ndBlock)
if sText == None:
sText = ""
traceln("Warning: no text in node")
# raise ValueError, "No text in node: %s"%ndBlock
# now we need to infer the bounding box of that object
lXY = self.getPointList(ndBlock) # the polygon
if lXY == []:
continue
plg = Polygon(lXY)
try:
x1, y1, x2, y2 = plg.fitRectangle(
bPreserveWidth=self.bPreserveWidth)
except ZeroDivisionError:
x1, y1, x2, y2 = plg.getBoundingBox()
except ValueError:
x1, y1, x2, y2 = plg.getBoundingBox()
# we reduce a bit this rectangle, to ovoid overlap
if not (self.BBoxDeltaFun is None):
if type(self.BBoxDeltaFun) is tuple and len(
self.BBoxDeltaFun) == 2:
xFun, yFun = self.BBoxDeltaFun
if xFun is not None:
dx = xFun(x2 - x1)
x1, x2 = int(round(x1 + dx)), int(round(x2 - dx))
if yFun is not None:
dy = yFun(y2 - y1)
y1, y2 = int(round(y1 + dy)), int(round(y2 - dy))
else:
# historical code
w, h = x2 - x1, y2 - y1
dx = self.BBoxDeltaFun(w)
dy = self.BBoxDeltaFun(h)
x1, y1, x2, y2 = [
int(round(v))
for v in [x1 + dx, y1 + dy, x2 - dx, y2 - dy]
]
# TODO
orientation = 0 # no meaning for PageXml
classIndex = 0 # is computed later on
# and create a Block
blk = Block(page, (x1, y1, x2 - x1, y2 - y1),
sText,
orientation,
classIndex,
self,
ndBlock,
domid=None)
blk.setShape(geom.Polygon(lXY))
yield blk
return
def _iter_GraphNode(self, doc, domNdPage, page):
"""
Get the DOM, the DOM page node, the page object
iterator on the DOM, that returns nodes (of class Block)
"""
#--- XPATH contexts
assert self.sxpNode, "CONFIG ERROR: need an xpath expression to enumerate elements corresponding to graph nodes"
lNdBlock = domNdPage.xpath(
self.sxpNode, namespaces=self.dNS) #all relevant nodes of the page
for ndBlock in lNdBlock:
domid = ndBlock.get("id")
sText = self._get_GraphNodeText(doc, domNdPage, ndBlock)
if sText == None:
sText = ""
NodeType_PageXml.nbNoTextWarning += 1
if NodeType_PageXml.nbNoTextWarning < 33:
traceln("Warning: no text in node %s" % domid)
elif NodeType_PageXml.nbNoTextWarning == 33:
traceln(
"Warning: no text in node %s - *** %d repetition : I STOP WARNING ***"
% (domid, NodeType_PageXml.nbNoTextWarning))
#raise ValueError, "No text in node: %s"%ndBlock
#now we need to infer the bounding box of that object
lXY = PageXml.getPointList(ndBlock) #the polygon
if lXY == []:
continue
plg = Polygon(lXY)
try:
x1, y1, x2, y2 = plg.fitRectangle(
bPreserveWidth=self.bPreserveWidth)
except ZeroDivisionError:
# traceln("Warning: ignoring invalid polygon id=%s page=%s"%(ndBlock.prop("id"), page.pnum))
# continue
# if True:
# #we reduce a bit this rectangle, to ovoid overlap
# w,h = x2-x1, y2-y1
# dx = max(w * 0.066, min(20, w/3)) #we make sure that at least 1/"rd of te width will remain!
# dy = max(h * 0.066, min(20, w/3))
# x1,y1, x2,y2 = [ int(round(v)) for v in [x1+dx,y1+dy, x2-dx,y2-dy] ]
x1, y1, x2, y2 = plg.getBoundingBox()
except ValueError:
x1, y1, x2, y2 = plg.getBoundingBox()
#we reduce a bit this rectangle, to ovoid overlap
if not (self.BBoxDeltaFun is None):
if type(self.BBoxDeltaFun) is tuple and len(
self.BBoxDeltaFun) == 2:
xFun, yFun = self.BBoxDeltaFun
if xFun is not None:
dx = xFun(x2 - x1)
x1, x2 = int(round(x1 + dx)), int(round(x2 - dx))
if yFun is not None:
dy = yFun(y2 - y1)
y1, y2 = int(round(y1 + dy)), int(round(y2 - dy))
else:
# historical code
w, h = x2 - x1, y2 - y1
dx = self.BBoxDeltaFun(w)
dy = self.BBoxDeltaFun(h)
x1, y1, x2, y2 = [
int(round(v))
for v in [x1 + dx, y1 + dy, x2 - dx, y2 - dy]
]
# store the rectangle"
ndBlock.set(
"DU_points", " ".join([
"%d,%d" % (int(x), int(y))
for x, y in [(x1, y1), (x2, y1), (x2, y2), (x1, y2)]
]))
#TODO
orientation = 0 #no meaning for PageXml
classIndex = 0 #is computed later on
#and create a Block
blk = Block(page, (x1, y1, x2 - x1, y2 - y1),
sText,
orientation,
classIndex,
self,
ndBlock,
domid=domid)
yield blk
return