def find_clusters(para_blocks):
def uniq_path(blocks):
r = []
prev = None
for b in blocks:
if prev == None or b.path != prev:
r.append(b.path)
prev = b.path
return r
def find_common(seqs):
s0 = None
for s1 in seqs:
if s0 == None:
s0 = s1
else:
s0 = [ s0[i0] for (i0,i1) in find_lcs(s0, s1) ]
return s0
# obtain the common paths.
common_paths = find_common([ uniq_path(blocks) for blocks in para_blocks ])
# clusters = [ ( doc1_blocks1, doc2_blocks1, ..., docm_blocks1 ),
# ...
# ( doc1_blocksn, doc2_blocksn, ..., docm_blocksn ) ]
clusters = zip(*[ retrieve_blocks(common_paths, blocks) for blocks in para_blocks ])
# compare each cluster of text blocks.
layout = []
for blockgroups in clusters:
if blockgroups[0]:
layout.append(LayoutSectionCluster(len(layout), blockgroups))
return layout
def match_blocks(self, blocks0, strict=True):
diffs = [ d for (d,m,p) in self.pattern ]
mains = [ m for (d,m,p) in self.pattern ]
paths = [ p for (d,m,p) in self.pattern ]
layout = []
for (diffscore,mainscore,blocks1) in zip(diffs, mains, retrieve_blocks(paths, blocks0)):
if strict and not blocks1:
return None
layout.append(LayoutSection(len(layout), diffscore, mainscore, blocks1))
return layout