def __parse_document(self, tree):
document = BioCDocument()
document.id = tree.findtext('id')
document.infons = self.__parse_infons(tree)
for child in tree.findall('passage'):
document.add_passage(self.__parse_passage(child))
for child in tree.findall('annotation'):
document.add_annotation(self.__parse_annotation(child))
for child in tree.findall('relation'):
document.add_relation(self.__parse_relation(child))
return document
def parse_doc(obj: Dict) -> BioCDocument:
"""
Deserialize a dict obj to a BioCDocument object
"""
doc = BioCDocument()
doc.id = obj['id']
doc.infons = obj['infons']
for passage in obj['passages']:
doc.add_passage(parse_passage(passage))
if 'annotations' in obj:
for annotation in obj['annotations']:
doc.add_annotation(parse_annotation(annotation))
for relation in obj['relations']:
doc.add_relation(parse_relation(relation))
return doc
def __parse_document(self, tree):
document = BioCDocument()
document.id = tree.attrib['id']
document.infons = self.__parse_infons(tree)
if tree.find('text') is not None:
document.text = tree.findtext('text')
for child in tree.findall('passage'):
document.add_passage(self.__parse_passage(child))
for child in tree.findall('sentence'):
document.add_sentence(self.__parse_sentence(child))
for child in tree.findall('annotation'):
document.add_annotation(self.__parse_annotation(child))
for child in tree.findall('relation'):
document.add_relation(self.__parse_relation(child))
return document
def __read(self):
while self.__has_next():
event, elem = self.__next_event()
if self.__state == 0:
if event == 'start':
if elem.tag == 'collection':
self.__state = 1
self.__collection = BioCCollection()
# collection
elif self.__state == 1:
if event == 'start':
if elem.tag == 'document':
self.__document = BioCDocument()
self.__document.id = elem.get('id')
self.__state = 2
elif event == 'end':
if elem.tag == 'source':
self.__collection.source = elem.text
elif elem.tag == 'date':
self.__collection.date = elem.text
elif elem.tag == 'key':
self.__collection.key = elem.text
elif elem.tag == 'version':
self.__collection.version = elem.text
elif elem.tag == 'infon':
self.__collection.infons[elem.get('key')] = elem.text
elif elem.tag == 'collection':
self.__state = 0
self.__document = None
self.__passage = None
self.__sentence = None
# document
elif self.__state == 2:
if event == 'start':
if elem.tag == 'passage':
self.__passage = BioCPassage()
self.__passage.offset = int(elem.get('offset'))
self.__state = 3
elif elem.tag == 'annotation':
self.__document.add_annotation(
self.__read_annotation(elem))
elif elem.tag == 'relation':
self.__document.add_relation(
self.__read_relation(elem))
elif event == 'end':
if elem.tag == 'infon':
self.__document.infons[elem.get('key')] = elem.text
elif elem.tag == 'document':
self.__state = 1
return
# passage
elif self.__state == 3:
if event == 'start':
if elem.tag == 'sentence':
self.__sentence = BioCSentence()
self.__sentence.offset = int(elem.get('offset'))
self.__state = 4
elif elem.tag == 'annotation':
self.__passage.add_annotation(
self.__read_annotation(elem))
elif elem.tag == 'relation':
self.__passage.add_relation(self.__read_relation(elem))
elif event == 'end':
if elem.tag == 'text':
self.__passage.text = elem.text
elif elem.tag == 'infon':
self.__passage.infons[elem.get('key')] = elem.text
elif elem.tag == 'passage':
self.__state = 2
if self.__passage is not None:
self.__document.add_passage(self.__passage)
# sentence
elif self.__state == 4:
if event == 'start':
if elem.tag == 'annotation':
self.__sentence.add_annotation(
self.__read_annotation(elem))
elif elem.tag == 'relation':
self.__sentence.add_relation(
self.__read_relation(elem))
elif event == 'end':
if elem.tag == 'text':
self.__sentence.text = elem.text
elif elem.tag == 'infon':
self.__sentence.infons[elem.get('key')] = elem.text
elif elem.tag == 'sentence':
self.__state = 3
if self.__sentence is not None:
self.__passage.add_sentence(self.__sentence)
class BioCXMLDocumentReader:
"""
Reader for the BioC XML format, one document per iteration.
"""
def __init__(self, source: Union[str, BinaryIO]):
# if not isinstance(file, str):
# file = str(file)
self.file = source
self.__context = iter(
etree.iterparse(self.file, events=('start', 'end')))
self.__state = 0
self.__event = None
self.__elem = None
self.__read()
def __iter__(self):
return self
def __next__(self):
"""
Reads one BioC document from the XML file.
Returns:
BioCDocument: the BioC document
"""
if self.__document is None:
raise StopIteration
else:
document = self.__document
self.__read()
return document
def __read(self):
while self.__has_next():
event, elem = self.__next_event()
if self.__state == 0:
if event == 'start':
if elem.tag == 'collection':
self.__state = 1
self.__collection = BioCCollection()
# collection
elif self.__state == 1:
if event == 'start':
if elem.tag == 'document':
self.__document = BioCDocument()
self.__document.id = elem.get('id')
self.__state = 2
elif event == 'end':
if elem.tag == 'source':
self.__collection.source = elem.text
elif elem.tag == 'date':
self.__collection.date = elem.text
elif elem.tag == 'key':
self.__collection.key = elem.text
elif elem.tag == 'version':
self.__collection.version = elem.text
elif elem.tag == 'infon':
self.__collection.infons[elem.get('key')] = elem.text
elif elem.tag == 'collection':
self.__state = 0
self.__document = None
self.__passage = None
self.__sentence = None
# document
elif self.__state == 2:
if event == 'start':
if elem.tag == 'passage':
self.__passage = BioCPassage()
self.__passage.offset = int(elem.get('offset'))
self.__state = 3
elif elem.tag == 'annotation':
self.__document.add_annotation(
self.__read_annotation(elem))
elif elem.tag == 'relation':
self.__document.add_relation(
self.__read_relation(elem))
elif event == 'end':
if elem.tag == 'infon':
self.__document.infons[elem.get('key')] = elem.text
elif elem.tag == 'document':
self.__state = 1
return
# passage
elif self.__state == 3:
if event == 'start':
if elem.tag == 'sentence':
self.__sentence = BioCSentence()
self.__sentence.offset = int(elem.get('offset'))
self.__state = 4
elif elem.tag == 'annotation':
self.__passage.add_annotation(
self.__read_annotation(elem))
elif elem.tag == 'relation':
self.__passage.add_relation(self.__read_relation(elem))
elif event == 'end':
if elem.tag == 'text':
self.__passage.text = elem.text
elif elem.tag == 'infon':
self.__passage.infons[elem.get('key')] = elem.text
elif elem.tag == 'passage':
self.__state = 2
if self.__passage is not None:
self.__document.add_passage(self.__passage)
# sentence
elif self.__state == 4:
if event == 'start':
if elem.tag == 'annotation':
self.__sentence.add_annotation(
self.__read_annotation(elem))
elif elem.tag == 'relation':
self.__sentence.add_relation(
self.__read_relation(elem))
elif event == 'end':
if elem.tag == 'text':
self.__sentence.text = elem.text
elif elem.tag == 'infon':
self.__sentence.infons[elem.get('key')] = elem.text
elif elem.tag == 'sentence':
self.__state = 3
if self.__sentence is not None:
self.__passage.add_sentence(self.__sentence)
def __read_annotation(self, start_elem):
ann = BioCAnnotation()
ann.id = start_elem.get('id')
while self.__has_next():
event, elem = self.__next_event()
if event == 'start':
pass
elif event == 'end':
if elem.tag == 'text':
ann.text = elem.text
elif elem.tag == 'infon':
ann.infons[elem.get('key')] = elem.text
elif elem.tag == 'location':
ann.add_location(
BioCLocation(int(elem.get('offset')),
int(elem.get('length'))))
elif elem.tag == 'annotation':
return ann
raise RuntimeError("should not reach here") # pragma: no cover
def __read_relation(self, start_elem):
rel = BioCRelation()
rel.id = start_elem.get('id')
while self.__has_next():
event, elem = self.__next_event()
if event == 'start':
pass
elif event == 'end':
if elem.tag == 'infon':
rel.infons[elem.get('key')] = elem.text
elif elem.tag == 'node':
rel.add_node(BioCNode(elem.get('refid'), elem.get('role')))
if elem.tag == 'relation':
return rel
raise RuntimeError("should not reach here") # pragma: no cover
def __has_next(self):
try:
self.__event, self.__elem = next(self.__context)
return True
except StopIteration:
self.__event = None
self.__elem = None
return False
def __next_event(self):
return self.__event, self.__elem
def get_collection_info(self) -> BioCCollection:
"""
Reads the collection information: encoding, version, DTD, source, date, key, infons, etc.
Returns:
the BioC collection that contains only information
"""
return self.__collection