|
1
|
|
|
# pylint: disable=line-too-long,invalid-name,protected-access,missing-module-docstring |
|
2
|
|
|
def _region_class(self, x): # pylint: disable=unused-argument |
|
3
|
|
|
return x.__class__.__name__.replace('RegionType', '') |
|
4
|
|
|
|
|
5
|
|
|
def _get_recursive_regions(self, regions, level, classes=None): |
|
6
|
|
|
if level == 1: |
|
7
|
|
|
# stop recursion, filter classes |
|
8
|
|
|
if classes: |
|
9
|
|
|
return [r for r in regions if self._region_class(r) in classes] |
|
10
|
|
|
if regions and regions[0].__class__.__name__ == 'PageType': |
|
11
|
|
|
regions = regions[1:] |
|
12
|
|
|
return regions |
|
13
|
|
|
# find more regions recursively |
|
14
|
|
|
more_regions = [] |
|
15
|
|
|
for region in regions: |
|
16
|
|
|
more_regions.append([]) |
|
17
|
|
|
for class_ in ['Advert', 'Chart', 'Chem', 'Custom', 'Graphic', 'Image', |
|
18
|
|
|
'LineDrawing', 'Map', 'Maths', 'Music', 'Noise', |
|
19
|
|
|
'Separator', 'Table', 'Text', 'Unknown']: |
|
20
|
|
|
if class_ == 'Map' and not isinstance(region, PageType): # pylint: disable=undefined-variable |
|
|
|
|
|
|
21
|
|
|
# 'Map' is not recursive in 2019 schema |
|
22
|
|
|
continue |
|
23
|
|
|
more_regions[-1] += getattr(region, 'get_{}Region'.format(class_))() |
|
24
|
|
|
if not any(more_regions): |
|
25
|
|
|
return self._get_recursive_regions(regions, 1, classes) |
|
26
|
|
|
ret = [] |
|
27
|
|
|
for r, more in zip(regions, more_regions): |
|
28
|
|
|
ret.append(r) |
|
29
|
|
|
ret += self._get_recursive_regions(more, level - 1 if level else 0, classes) |
|
30
|
|
|
return self._get_recursive_regions(ret, 1, classes) |
|
31
|
|
|
|
|
32
|
|
|
def _get_recursive_reading_order(self, rogroup): |
|
33
|
|
|
if isinstance(rogroup, (OrderedGroupType, OrderedGroupIndexedType)): # pylint: disable=undefined-variable |
|
|
|
|
|
|
34
|
|
|
elements = rogroup.get_AllIndexed() |
|
35
|
|
|
if isinstance(rogroup, (UnorderedGroupType, UnorderedGroupIndexedType)): # pylint: disable=undefined-variable |
|
|
|
|
|
|
36
|
|
|
elements = (rogroup.get_RegionRef() + rogroup.get_OrderedGroup() + rogroup.get_UnorderedGroup()) |
|
37
|
|
|
regionrefs = list() |
|
38
|
|
|
for elem in elements: |
|
|
|
|
|
|
39
|
|
|
regionrefs.append(elem.get_regionRef()) |
|
40
|
|
|
if not isinstance(elem, (RegionRefType, RegionRefIndexedType)): # pylint: disable=undefined-variable |
|
|
|
|
|
|
41
|
|
|
regionrefs.extend(self._get_recursive_reading_order(elem)) |
|
42
|
|
|
return regionrefs |
|
43
|
|
|
|
|
44
|
|
|
def get_AllRegions(self, classes=None, order='document', depth=0): |
|
45
|
|
|
""" |
|
46
|
|
|
Get all the *Region element or only those provided by ``classes``. |
|
47
|
|
|
Returned in document order unless ``order`` is ``reading-order`` |
|
48
|
|
|
Arguments: |
|
49
|
|
|
classes (list) Classes of regions that shall be returned, e.g. ``['Text', 'Image']`` |
|
50
|
|
|
order ("document"|"reading-order"|"reading-order-only") Whether to |
|
51
|
|
|
return regions sorted by document order (``document``, default) or by |
|
52
|
|
|
reading order with regions not in the reading order at the end of the |
|
53
|
|
|
returned list (``reading-order``) or regions not in the reading order |
|
54
|
|
|
omitted (``reading-order-only``) |
|
55
|
|
|
depth (int) Recursive depth to look for regions at, set to `0` for all regions at any depth. Default: 0 |
|
56
|
|
|
|
|
57
|
|
|
For example, to get all text anywhere on the page in reading order, use: |
|
58
|
|
|
:: |
|
59
|
|
|
'\\n'.join(line.get_TextEquiv()[0].Unicode |
|
60
|
|
|
for region in page.get_AllRegions(classes=['Text'], depth=0, order='reading-order') |
|
61
|
|
|
for line in region.get_TextLine()) |
|
62
|
|
|
""" |
|
63
|
|
|
if order not in ['document', 'reading-order', 'reading-order-only']: |
|
64
|
|
|
raise Exception("Argument 'order' must be either 'document', 'reading-order' or 'reading-order-only', not '{}'".format(order)) |
|
65
|
|
|
if depth < 0: |
|
66
|
|
|
raise Exception("Argument 'depth' must be an integer greater-or-equal 0, not '{}'".format(depth)) |
|
67
|
|
|
ret = self._get_recursive_regions([self], depth + 1 if depth else 0, classes) |
|
68
|
|
|
if order.startswith('reading-order'): |
|
69
|
|
|
reading_order = self.get_ReadingOrder() |
|
70
|
|
|
if reading_order: |
|
71
|
|
|
reading_order = reading_order.get_OrderedGroup() or reading_order.get_UnorderedGroup() |
|
72
|
|
|
if reading_order: |
|
73
|
|
|
reading_order = self._get_recursive_reading_order(reading_order) |
|
74
|
|
|
if reading_order: |
|
75
|
|
|
id2region = {region.id: region for region in ret} |
|
76
|
|
|
in_reading_order = [id2region[region_id] for region_id in reading_order if region_id in id2region] |
|
77
|
|
|
# print("ret: {} / in_ro: {} / not-in-ro: {}".format( |
|
78
|
|
|
# len(ret), |
|
79
|
|
|
# len([id2region[region_id] for region_id in reading_order if region_id in id2region]), |
|
80
|
|
|
# len([r for r in ret if r not in in_reading_order]) |
|
81
|
|
|
# )) |
|
82
|
|
|
if order == 'reading-order-only': |
|
83
|
|
|
ret = in_reading_order |
|
84
|
|
|
else: |
|
85
|
|
|
ret = in_reading_order + [r for r in ret if r not in in_reading_order] |
|
86
|
|
|
return ret |
|
87
|
|
|
|