def parseDomNodeLabel(self, domnode, defaultCls=None):
"""
Parse and set the graph node label and return its class index
raise a ValueError if the label is missing while bOther was not True, or if the label is neither a valid one nor an ignored one
"""
sLabel = self.sDefaultLabel
try:
try:
sXmlLabel = PageXml.getCustomAttr(domnode, self.sCustAttr_STRUCTURE, self.sCustAttr2_TYPE)
except PageXmlException as e:
if self.bOther:
return self.sDefaultLabel #absence of label but bOther was True (I guess)
else:
raise e
try:
sLabel = self.dXmlLabel2Label[sXmlLabel]
except KeyError:
#not a label of interest
try:
self.checkIsIgnored(sXmlLabel)
#if self.lsXmlIgnoredLabel and sXmlLabel not in self.lsXmlIgnoredLabel:
except:
raise ValueError("Invalid label '%s' in node %s"%(sXmlLabel, etree.tostring(domnode)))
except KeyError:
#no label at all
if not self.sDefaultLabel: raise ValueError("Missing label in node %s"%etree.tostring(domnode))
return sLabel
def _convertPageAnnotation(self, pnum, page, domNdPage):
"""
"""
for nd in self._iter_TextRegionNodeTop2Bottom(domNdPage, page):
try:
sResoNum = None
lbl = PageXml.getCustomAttr(nd, "structure", "type")
if lbl in ["heading"]:
semLabel = self.dAnnotMapping[lbl]
#heading may indicate a new resolution!
if self.prevResolutionNumber == None:
sgmLabel = self._getCurrentSegmentationLabel(
) #for instance 2 consecutive headings
else:
sgmLabel = self._switchSegmentationLabel()
self.prevResolutionNumber = None #so that next number does not switch Heigh/Ho label
elif lbl in ["header", "page-number", "marginalia"]:
#continuation of a resolution
semLabel = self.dAnnotMapping[lbl]
sgmLabel = self._getCurrentSegmentationLabel()
else:
o = self.creResolutionHumanLabel.match(lbl)
if not o:
raise ValueError("%s is not a valid human annotation" %
lbl)
semLabel = self.dAnnotMapping[o.group(
1)] #"" for the resolution number
#Here we have a resolution number!
sResoNum = o.group(2)
if not sResoNum:
raise ValueError(
"%s is not a valid human annotation - missing resolution number"
% lbl)
#now switch between heigh and ho !! :))
if self.prevResolutionNumber != None and self.prevResolutionNumber != sResoNum:
#we got a new number, so switching segmentation label!
sgmLabel = self._switchSegmentationLabel()
else:
#either same number or switching already done due to a heading
sgmLabel = self._getCurrentSegmentationLabel()
self.prevResolutionNumber = sResoNum
except PageXmlException:
semLabel = self.sOther
sgmLabel = self._getCurrentSegmentationLabel()
nd.set(self.sSemAttr, semLabel)
nd.set(self.sSgmAttr, sgmLabel)
if sResoNum:
nd.set(self.sNumAttr, sResoNum
) #only when the number is part of the humanannotation!
def _convertPageAnnotation(self, pnum, page, domNdPage):
"""
"""
#change: on each page we start by Heigh
bRestartAtEachPageWithHeigh = True
if bRestartAtEachPageWithHeigh: self._initSegmentationLabel()
for nd in self._iter_TextRegionNodeTop2Bottom(domNdPage, page):
try:
lbl = PageXml.getCustomAttr(nd, "structure", "type")
except PageXmlException:
nd.set(self.sSemAttr, self.sOther)
nd.set(self.sSgmAttr, self.sOther)
continue #this node has no annotation whatsoever
if lbl in ["heading", "header", "page-number", "marginalia"]:
semLabel = lbl
sgmLabel = self.sOther #those elements are not part of a resolution
sResoNum = None
else:
o = self.creResolutionHumanLabel.match(lbl)
if not o:
raise ValueError("%s is not a valid human annotation" %
lbl)
semLabel = o.group(1) #"" for the resolution number
#now decide on the segmentation label
sResoNum = o.group(2)
if not sResoNum:
raise ValueError(
"%s is not a valid human annotation - missing resolution number"
% lbl)
#now switch between heigh and ho !! :))
if self.prevResolutionNumber == sResoNum:
sgmLabel = self.prevSgmLbl
else:
sgmLabel = self._getNextSegmentationLabel(self.prevSgmLbl)
assert bRestartAtEachPageWithHeigh or sResoNum not in self.lSeenResoNum, "ERROR: the ordering of the block has not preserved resolution number contiguity"
self.lSeenResoNum.append(sResoNum)
self.prevResolutionNumber, self.prevSgmLbl = sResoNum, sgmLabel
#always have a semantic label
sNewSemLbl = self.dAnnotMapping[semLabel]
assert sNewSemLbl
nd.set(self.sSemAttr, sNewSemLbl) #DU annotation
#resolution parts also have a segmentation label and a resolution number
assert sgmLabel
nd.set(self.sSgmAttr, sgmLabel) #DU annotation
if sResoNum:
nd.set(self.sNumAttr, sResoNum)
def parseDocNodeLabel(self, graph_node, defaultCls=None):
"""
Parse and set the graph node label and return its class index
We rely on the standard self.sLabelAttr
raise a ValueError if the label is missing while bOther was not True
, or if the label is neither a valid one nor an ignored one
"""
ndParent = graph_node.node.getparent()
try:
sLabel = "%s_%s" % ( self.sLabelAttr,
PageXml.getCustomAttr(ndParent, 'structure','type')
)
except :
sLabel='type_None'
return sLabel
def test_getsetCustomAttr():
sXml = b"""
<TextRegion type="page-number" id="p1_region_1471502505726_2" custom="readingOrder {index:9;} structure {type:page-number;}">
<Coords points="972,43 1039,43 1039,104 972,104"/>
</TextRegion>
"""
doc = etree.parse(BytesIO(sXml))
nd = doc.getroot()
assert PageXml.getCustomAttr(nd, "readingOrder", "index") == '9'
assert PageXml.setCustomAttr(nd, "readingOrder", "index", 99) == 99
assert PageXml.getCustomAttr(nd, "readingOrder", "index") == '99'
assert PageXml.getCustomAttr(nd, "readingOrder") == {'index': '99'}
assert PageXml.setCustomAttr(nd, "readingOrder", "toto", "zou") == "zou"
assert PageXml.getCustomAttr(nd, "readingOrder", "toto") == 'zou'
with pytest.raises(PageXmlException):
PageXml.getCustomAttr(nd, "readingOrder", "axiste_pas")
with pytest.raises(PageXmlException):
PageXml.getCustomAttr(nd, "axiste_pas_non_plus", "axiste_pas")
def parsePage(self, doc, ctxtNd, name):
for tag in self.lTag:
lNdTag = ctxtNd.xpath(".//pg:%s"%tag, namespaces=self.dNS)
for nd in lNdTag:
self.seenDocTag(name, tag)
if self.sCustom != None:
if self.sCustom == "":
try:
lbl = PageXml.getCustomAttr(nd, "structure", "type")
except:
lbl = ''
else:
lbl = nd.get(self.sCustom)
else:
lbl = nd.get("type")
if lbl:
for cre, sRepl in self.ltCRES: lbl = cre.sub(sRepl, lbl) #pattern processing
self.seenTagLabel(tag, lbl)
def _convertPageAnnotation(self, pnum, page, domNdPage):
"""
"""
for nd in self._iter_TextRegionNodeTop2Bottom(domNdPage, page):
sResoNum = None
bCurrentIsAStart = None
try:
lbl = PageXml.getCustomAttr(nd, "structure", "type")
if lbl == "heading":
semLabel = self.dAnnotMapping[lbl]
#heading indicate the start of a new resolution, unless the previous is already a start!
if self._prevIsB:
bCurrentIsAStart = False
else:
bCurrentIsAStart = True
self._prevNum = False #to prevent starting again when find the resolution number
elif lbl in ["header", "page-number", "marginalia"]:
semLabel = self.dAnnotMapping[lbl]
#continuation of a resolution, except at very beginning (first node)
if self._prevNd == None:
bCurrentIsAStart = True
else:
bCurrentIsAStart = False
else:
o = self.creResolutionHumanLabel.match(lbl)
if not o:
if False: # strict
raise ValueError("%s is not a valid human annotation" % lbl)
else:
# relaxed
print(" ** WARNING ** strange annotation on node id=%s : '%s'"%(nd.get("id"), lbl))
semLabel = self.dAnnotMapping[None]
#Here we have a resolution number!
sResoNum = self._prevNum
else:
semLabel = self.dAnnotMapping[o.group(1)] #"" for the resolution number
#Here we have a resolution number!
sResoNum = o.group(2)
if not sResoNum: raise ValueError("%s is not a valid human annotation - missing resolution number" % lbl)
if self._prevNum != False and self._prevNum != sResoNum:
#we got a new number, so switching segmentation label!
bCurrentIsAStart = True
else:
#either same number or switching already done due to a heading
bCurrentIsAStart = False
self._prevNum = sResoNum
except PageXmlException:
semLabel = self.sOther
bCurrentIsAStart = False
#Now tagging!!
#Semantic (easy)
nd.set(self.sSemAttr, semLabel)
# BIES, tough...
if bCurrentIsAStart:
if self._prevIsB:
#make previous a singleton!
if self._prevNd: self._prevNd.set(self.sSgmAttr, self.S)
else:
#make previous a End
if self._prevNd: self._prevNd.set(self.sSgmAttr, self.E)
self._prevIsB = True #for next cycle!
else:
if self._prevIsB:
#confirm previous a a B
if self._prevNd: self._prevNd.set(self.sSgmAttr, self.B)
else:
#confirm previous as a I
if self._prevNd: self._prevNd.set(self.sSgmAttr, self.I)
self._prevIsB = False #for next cycle!
if sResoNum: nd.set(self.sNumAttr, sResoNum) #only when the number is part of the humanannotation!
self._prevNd = nd #for next cycle!
# end for
if self._prevIsB:
#make previous a singleton!
if self._prevNd: self._prevNd.set(self.sSgmAttr, self.S)
else:
#make previous a End
if self._prevNd: self._prevNd.set(self.sSgmAttr, self.E)
return
