def to_tables7(self):
lines = []
for t in filter(lambda x: x not in self.multi_col_boxes, self.texts):
for line in t:
placed = False
for l in lines:
if (l['contain'].voverlap(line) / l['contain'].height >
0.9) and (l['contain'].voverlap(line) /
l['contain'].height < 1.1):
l['texts'].append(line)
l['contain'].set_bbox(
(min(l['contain'].x0, line.x0), l['contain'].y0,
max(l['contain'].x1, line.x1), l['contain'].y1))
placed = True
if not placed:
lines.append({
'contain': LTComponent(line.bbox),
'texts': [line]
})
lines.sort(key=lambda x: x['contain'].y0)
for line in lines:
line['cells'] = [None] * len(self.finalcols2)
for columntext in line['texts']:
for i, column in enumerate(self.finalcols2):
if column['contain'].hoverlap(columntext):
if line['cells'][i] is None:
line['cells'][i] = columntext
break
else:
# seems like the parser library sometimes duplicates text, possible bug
pass
self.lines = lines
def __init__(self, annoObj, uri, pos, pageid):
self.origObjs = [annoObj]
self.gotoLoc = uri
self.assocText = [""]
self.assocTextIn = 0
self.positions = [[LTComponent(pos), pageid]]
self.destPage = None
self.unparseCite = ""
self.finalCiteStr = ""
self.papObj = None
self.papLink = ""
self.author = ""
self.year = ""
def to_tables3(self):
colgroups = []
for c in self.columns:
colg = None
for d in colgroups:
if c is not d:
if all(c['contain'].is_hoverlap(e['contain'])
for e in d['cols']):
d['cols'].append(c)
colg = d
break
if not colg:
colgroups.append({
'contain': LTComponent(c['contain'].bbox),
'cols': [c]
})
logger.info('{x} colgroups'.format(x=len(colgroups)))
self.colgroups = colgroups
def given_plane_with_one_object(object_size=50, gridsize=50):
bounding_box = (0, 0, 100, 100)
plane = Plane(bounding_box, gridsize)
obj = LTComponent((0, 0, object_size, object_size))
plane.add(obj)
return plane, obj
def to_tables2(self):
columns = []
# texts.sort(key=lambda x: x.width)
for e in self.layout:
if isinstance(e, LTTextBoxHorizontal):
logger.info('Finding a column for box {i}'.format(i=e.index))
##im2 = im.copy()
##d = ImageDraw.Draw(im2)
col = None
for c in columns:
if (e.x1 < c['contain'].x1) and (e.x0 > c['contain'].x0):
if (e.width / c['contain'].width) < 0.8:
logger.info(
'Item too small, column may be several columns wide'
)
else:
logger.info('Item totally contained in column')
logger.info(
'{ex1} < {cx1} and {ex0} > {cx0}'.format(
ex1=e.x1,
cx1=c['contain'].x1,
ex0=e.x0,
cx0=c['contain'].x0))
col = c
col['boxes'].append(e)
col['contain'].set_bbox(
(c['contain'].x0, min(c['contain'].y0,
e.y0), c['contain'].x1,
max(c['contain'].y1, e.y1)))
##d.rectangle([int(c['contain'].bbox[0] * scale_factor), h - int(c['contain'].bbox[3] * scale_factor), int(c['contain'].bbox[2] * scale_factor), h - int(c['contain'].bbox[1] * scale_factor)], fill=None, outline=c['color'], width=5 * scale_factor)
break
elif ((c['contain'].hoverlap(e) / c['contain'].width) >
0.9) and ((c['contain'].hoverlap(e) /
c['contain'].width) < 1.1):
logger.info('Item is within 10% of current col width')
logger.info(
'Overlap of {hdist}, column width of {width}'.
format(hdist=c['contain'].hoverlap(e),
width=c['contain'].width))
col = c
col['boxes'].append(e)
col['contain'].set_bbox(
(min(c['contain'].x0,
e.x0), min(c['contain'].y0,
e.y0), max(c['contain'].x1, e.x1),
max(c['contain'].y1, e.y1)))
##d.rectangle([int(c['contain'].bbox[0] * scale_factor), h - int(c['contain'].bbox[3] * scale_factor), int(c['contain'].bbox[2] * scale_factor), h - int(c['contain'].bbox[1] * scale_factor)], fill=None, outline=c['color'], width=5 * scale_factor)
break
if not col:
logger.info('Creating new column')
col = {
'contain':
LTComponent(e.bbox),
'boxes':
list(e),
'color':
random.choice(list(PIL.ImageColor.colormap.keys()))
}
columns.append(col)
columns.sort(key=lambda x: x['contain'].width)
##d.rectangle([int(col['contain'].bbox[0] * scale_factor), h - int(col['contain'].bbox[3] * scale_factor), int(col['contain'].bbox[2] * scale_factor), h - int(col['contain'].bbox[1] * scale_factor)], fill=None, outline=col['color'], width=5 * scale_factor)
##im2.save('page0.{x}.png'.format(x=e.index), "PNG")
self.columns = columns
def addPosition(self, pos, pageid):
self.positions.append([LTComponent(pos), pageid])
def _create_cells(network):
"""
Creates cells from the network and returns then
as LTComponents.
"""
squares_taken = defaultdict(set)
cells = set()
def city_distance(point, point_prime):
return abs(point.x - point_prime.x) + abs(point.y - point_prime.y)
def is_perpendicular(v1_x, v1_y, v2_x, v2_y):
return v1_x*v2_x + v1_y*v2_y == 0
for point in sorted(network, key=lambda p: (p.x, p.y)):
for l1 in sorted(network.links[point],
key=lambda p: city_distance(p, point)):
valid_links = [
link for link in network.links[point] if link != l1 and
is_perpendicular(link.x - point.x, link.y - point.y,
l1.x - point.x, l1.y - point.y)]
for l2 in sorted(valid_links,
key=lambda p: city_distance(p, point)):
inter = network.links[l2].intersection(network.links[l1])
intersection = list(inter)
# remove initial point
intersection.remove(point)
if len(intersection) == 0:
continue
# sort by areas: smallest area first
area = lambda p: (p.x - point.x)*(p.y - point.y)
intersection.sort(key=area)
# square is formed by [point, l1, l2, last_point], in this
# order.
points = [point, l1, l2, intersection[0]]
# compute middle position of the square
middle_x = sum(point.x for point in points)/4.
middle_y = sum(point.y for point in points)/4.
# check if any point already has one of its squares
# (at most 4) used.
is_taken = False
square = range(4)
for i in range(4):
# compute the position of the point in relation to the
# middle corresponding to one of the following squares
# position: [(1,1), (-1,1), (1,-1), (-1,-1)]
vx = middle_x - points[i].x
vy = middle_y - points[i].y
square[i] = (int(vx/abs(vx)), int(vy/abs(vy)))
belongs = square[i] in squares_taken[points[i]]
is_taken = is_taken or belongs
if not is_taken:
cell = LTComponent((point.x, point.y,
intersection[0].x, intersection[0].y))
cells.add(cell)
for i in range(4):
squares_taken[points[i]].add(square[i])
break
return cells
