def setRightEdge(self,modertm):
for r in modertm:
if misc.isEqual(self.x1,r,Xequalness):
self.rightedge = 1
return
if misc.isEqual(self.x1,self.rtmargin,Xequalness):
self.rightedge = 2
else:
self.rightedge = 3
def calcIndent(self, i, MOLS):
global Yequalness, edgemargin
currentEdge=self.lines[i].getx0()
if i != 0:
prevEdge=self.lines[i-1].getx0()
else:
prevEdge=currentEdge
if i == len(self.lines)-1:
nextEdge = prevEdge# if last line of page
else:
nextEdge = self.lines[i+1].getx0()
# the following four lines adds a bias to each edge value if any of them are negative
edges = [prevEdge,currentEdge,nextEdge]
edges.sort()
if edges[0] < 0:
bias = abs(edges[0])
prevEdge = prevEdge + bias
currentEdge = prevEdge + bias
nextEdge = prevEdge + bias
# if the adjacent two lines are aligned *AND* the current line is indented
if (self.lines[i].getLS() == "?" \
or misc.isEqual(prevEdge,nextEdge,Xequalness) \
or self.lines[i].getLS() > MOLS*10) \
and currentEdge > ((1+edgemargin)*nextEdge):
return abs(nextEdge-currentEdge) # Then the line is indented
else:
return 0 # otherwise not
def calcMargins(self, MOLS, MOLEN):
# generate a list of x1,y1 pairs. From this a frequency graph can be
# generated, and from that, the number of columns can be established
x1list = []
minLineLen = MOLEN/MaxColumns
for line in self.lines:
# only consider this line if it is wide enough to possibly be a whole line,
# using the approximation that there will be at most, MaxColumns on one page.
if line.getlength() > minLineLen or misc.isEqual(line.getlength(), minLineLen,Xequalness):
x1list.append( (line.getx1(),line.gety()) )
## x{0,1}list contain (x,y) pairs for each line on the page ##
# compress each list down to frequency,value pairs
self.rtMargins = self.findFreq( x1list )
self.rtMargins.sort( lambda a,b: int(a[1]-b[1]) ) # just to be quite sure
## {rt,lt}Margins NOW contain (f,x-av,y-av,y-min,y-max) ##
# We assume there will be no more than Maxcolumns (real) columns in the page.
# More can be apparent due to tables, non-text objects, etc, and this
# serves to eliminate some of these.
if len(self.rtMargins) > MaxColumns:
del self.rtMargins[:-MaxColumns]
if len(self.rtMargins) == 0:
return # no columns were found, so can't apply them
# lines with no right margin set will be handled in Page.predict_para
# all the extra information is there for historical reasons
# we need a copy of the margins to give to setRTmargin that doesn't have all that in it
tmpRTmargins = []
for m in self.rtMargins:
tmpRTmargins.append(m[1])
# now each line must be set with its margin as appropriate
for l in self.lines:
l.setRTmargin( tmpRTmargins )
def findFreq( self, list ):
if len(list) == 0: # can't operate on nothing
msg( "findFreq: WARNING: called with an empty input list!!" )
return []
output = [] # output list
f = 0 # frequency of val
val = list[0] # start with the first element
valcum = (0,0) # cumulation of vals
min = max = val[1] # minimum and maximum y values
# since val's aren't strictly the same, we'll average the val's to get a
# more accurate view of what val actually is
for l in list:
if misc.isEqual(val[0],l[0],Xequalness):
f = f + 1 # we've found another val
valcum = (valcum[0] + l[0], valcum[1] + l[1])
if l[1] < min:
min = l[1]
elif l[1] > max:
max = l[1]
else:
output.append( (f, round(valcum[0]/f,1),round(valcum[1]/f,1), min, max ) )
# reset
val = l # next val
f = 1 # and we've already found one
valcum = l
min = max = val[1]
output.append( (f, round(valcum[0]/f,1),round(valcum[1]/f,1), min, max) )
output.sort() # sort on frequency (pri) and x (sec)
# output now contains the list of consecutive frequencies.
# I found the following few lines of code very difficult to comment,
# so please excuse the bad explanation.
#
# 'output contains a list of (consecutive frequency,value) pairs. That
# means you can have several records that have the same value, but since
# they weren't consecutive in the input list, they were not combined.
#
# for example, [(1,10),(2,8),(1,10),(1,9),(5,10)]
#
# in that example, 10 is the value with a sufficiently high frequency,
# and so all values less than 10 must be deleted and all (x,10) records
# must be combined.
# this code finds the first record whose f is >= MinConsecFreq
i = 0 # where to cut
while i < len(output) and output[i][0] < MinConsecFreq:
i = i + 1 # find first i where output[i] >= MinConsecFreq
# if i points off the list, no eligible candidates were found.
# When this happens, we take the largest frequency found, which will
# be the last record in the list.
if i == len(output):
i = i - 1
# this code backtracks to ensure that any occurrances of the val
# associated with the i'th record also gets included. Without this,
# the frequency reported for the val[i] might be too small because
# some small groups (less than MinConsecFreq) were discounted
# because their f's were too small. Sort of. Try and make sense
# of that.
j = 0
while j < i:
if misc.isEqual(output[j][1],output[i][1],Xequalness):
j = j + 1
else:
del output[j]
i = i - 1
# output now contains only eligible frequencies which must now be combined
# ie, [(1,10),(2,10),(1,11)] ==> [(3,10),(1,11)]
#
output.sort( lambda a,b: int(a[1] - b[1]) ) # sort on x value
i = 0
while i <= len(output) - 2: # go from 0 to second to last index
if misc.isEqual( output[i][1], output[i+1][1],Xequalness ): #abs(output[i][1] - output[i+1][1]) <= 2:
output[i] = self.mergeRecords( output[i], output[i+1], lambda a,b: round((a+b)/2) )
del output[i+1]
else:
i = i + 1
return output
def calcValues(self, MOLS, MOLEN, modeRightMargins, lastPageLastLine):
global Xequalness
global Yequalness
for i in range(len(self.lines)):
self.lines[i].setRightEdge(modeRightMargins)
AVLS_CL=self.lines[i].localAverage=self.calcLocalAverage(i,MOLS)
NL_AVLS=self.lines[i].getNumberLocalAverage() # NL_AVLS -- Number of lines in the local average calc
LS_CL=self.lines[i].getLS() # LS_CL -- linespace of the current line
if i < len(self.lines)-1:
LS_NL=self.lines[i+1].getLS()
else:
LS_NL=10*MOLS
if LS_CL == '?':
self.lines[i].LS_CL_missing = YES
if lastPageLastLine != None:
if misc.isEqual(lastPageLastLine.getlength(),MOLEN,Xequalness): self.lines[i].lastpagelastline_eq_MOLEN = YES
if misc.isEqual(lastPageLastLine.getx1(),lastPageLastLine.getRTmargin(),Xequalness): self.lines[i].RTedge_lastpagelastline_eq_RTmargin = YES
else:
# We are only interested in the absolute space between lines,
# and no longer direction, and it is important that we make sure,
# we do not have any negative values in the linespaces from now on.
LS_CL=abs(LS_CL)
LN_PL=self.lines[i-1].getlength() # LN_PL -- Length of the previous line
LN_CL=self.lines[i].getlength() # LN_CL -- Length of the previous line
self.lines[i].diffLS_AVLS = LS_CL-AVLS_CL
self.lines[i].diffModeLineSpace = LS_CL-MOLS
self.lines[i].diffModeLength = LN_CL-MOLEN
if self.calcIndent(i, MOLS) != 0:
self.lines[i].isIndented = YES
if i > 0:
self.lines[i].lastindent = abs(self.lines[i].getx0()-self.lines[i-1].getx0())
if i < len(self.lines)-1:
self.lines[i].nextindent = abs(self.lines[i].getx0()-self.lines[i+1].getx0())
if misc.isEqual(LS_CL,MOLS,Yequalness):
self.lines[i].LS_CL_eq_MOLS = YES
if misc.isEqual(LS_CL,AVLS_CL,Yequalness):
self.lines[i].LS_CL_eq_AVLS = YES
if misc.isEqual(LS_CL,LS_NL,Yequalness):
self.lines[i].LS_CL_eq_LS_NL = YES
if LS_CL == AVLS_CL:
self.lines[i].LS_CL_exactly_AVLS = YES
if LS_CL < AVLS_CL:
self.lines[i].LS_CL_lt_AVLS = YES
if LS_CL > AVLS_CL:
self.lines[i].LS_CL_gt_AVLS = YES
if LS_CL > MOLS * 10:
self.lines[i].isColumnbreak = YES
if NL_AVLS == 1:
self.lines[i].NL_AVLS_eq_1 = YES
if misc.isEqual(LS_NL,MOLS,Yequalness):
self.lines[i].LS_NL_eq_MOLS = YES
if misc.isEqual(LN_PL,MOLEN,Xequalness):
self.lines[i].LN_PL_eq_MOLEN = YES
if misc.isEqual(LN_CL,MOLEN,Xequalness):
self.lines[i].LN_CL_eq_MOLEN = YES
if misc.isEqual(self.lines[i-1].getx1(),self.lines[i-1].getRTmargin(),Xequalness):
self.lines[i].RTedge_PL_eq_RTmargin = YES
if misc.isEqual(self.lines[i ].getx1(),self.lines[i ].getRTmargin(),Xequalness):
self.lines[i].RTedge_CL_eq_RTmargin = YES
def setRTmargin(self, rtMargins):
for m in range(len(rtMargins)):
if m == len(rtMargins)-1 or misc.isEqual(self.getx1(), rtMargins[m], Xequalness) or self.getx1() < rtMargins[m]:
self.rtmargin = rtMargins[m]
break
self.rtwhitespace = round(self.rtmargin-self.x1,1)