def __parse_relations(self, pblock, data):
for e in data:
members = []
mid = 0
for role_id, mtype, mid_delta in zip(e.roles_sid, e.types,
e.memids):
mid += mid_delta
members.append(
RelationMember(
pblock.stringtable.s[role_id].decode('utf-8'),
(Node, Way, Relation)[mtype], mid))
yield Relation(id=e.id,
version=e.info.version,
changeset=int(e.info.changeset),
timestamp=int(e.info.timestamp),
uid=e.info.uid,
tags=self.__parse_tags(e, pblock),
members=tuple(members))
def parse(self, fp):
context = iterparse(fp, events=('start', 'end'))
# common
_type = None
_id = None
_version = None
_changeset = None
_timestamp = None
_uid = None
_tags = None
# node only
_lon = None
_lat = None
# way only
_nodes = None
# relation only
_members = None
for event, elem in context:
if event == 'start':
attrs = elem.attrib
if elem.tag in ('node', 'way', 'relation'):
_id = long(attrs['id'])
_version = int(attrs['version'])
try: # GeoFabrik's (May 2018) public snapshots strip the changeset attribute out of their data
_changeset = int(attrs['changeset'])
except:
_changeset = 0
# TODO: improve timestamp parsing - dateutil too slow
_tstxt = attrs['timestamp']
_timestamp = int(
(datetime(year=int(_tstxt[0:4]),
month=int(_tstxt[5:7]),
day=int(_tstxt[8:10]),
hour=int(_tstxt[11:13]),
minute=int(_tstxt[14:16]),
second=int(_tstxt[17:19]),
tzinfo=None) -
datetime(year=1970, month=1, day=1,
tzinfo=None)).total_seconds())
try: #An object can miss an uid (when anonymous edits were possible)
_uid = int(attrs['uid'])
except:
uid = 0
_tags = {}
if elem.tag == 'node':
_type = Node
_lon = float(attrs['lon'])
_lat = float(attrs['lat'])
elif elem.tag == 'way':
_type = Way
_nodes = []
elif elem.tag == 'relation':
_type = Relation
_members = []
elif elem.tag == 'tag':
_tags[unicode(attrs['k'])] = unicode(attrs['v'])
elif elem.tag == 'nd':
_nodes.append(long(attrs['ref']))
elif elem.tag == 'member':
_members.append(
RelationMember(unicode(attrs['role']), {
'node': Node,
'way': Way,
'relation': Relation
}[attrs['type']], long(attrs['ref'])))
elif event == 'end':
if elem.tag in ('node', 'way', 'relation'):
args = [_id, _version, _changeset, _timestamp, _uid, _tags]
if elem.tag == 'node':
args.extend((_lon, _lat))
elif elem.tag == 'way':
args.append(tuple(_nodes))
elif elem.tag == 'relation':
args.append(tuple(_members))
elem.clear()
yield _type(*args)
def parse(self, fp):
d = json.load(fp)
de = d["elements"]
# common
_type = None
_id = None
_version = None
_changeset = None
_timestamp = None
_uid = None
_tags = None
# node only
_lon = None
_lat = None
# way only
_nodes = None
# relation only
_members = None
for elem in de:
tag = elem["type"]
attrs = elem
if tag in ('node', 'way', 'relation'):
_id = long(attrs['id'])
try:
_version = int(attrs['version'])
except:
_version = 0
try:
_changeset = int(attrs['changeset'])
except:
_changeset = 0
# TODO: improve timestamp parsing - dateutil too slow
try:
_tstxt = attrs['timestamp']
_timestamp = int(
(datetime(year=int(_tstxt[0:4]),
month=int(_tstxt[5:7]),
day=int(_tstxt[8:10]),
hour=int(_tstxt[11:13]),
minute=int(_tstxt[14:16]),
second=int(_tstxt[17:19]),
tzinfo=None) -
datetime(year=1970, month=1, day=1,
tzinfo=None)).total_seconds())
except:
_timestamp = 0
try: # An object can miss an uid (when anonymous edits were possible)
_uid = int(attrs['uid'])
except:
uid = 0
_tags = {}
if tag == 'node':
_type = Node
_lon = float(attrs['lon'])
_lat = float(attrs['lat'])
elif tag == 'way':
_type = Way
_nodes = []
elif tag == 'relation':
_type = Relation
_members = []
if "tags" in elem:
_tags = elem["tags"]
if "nodes" in elem:
_nodes = [long(x) for x in elem["nodes"]]
if 'members' in elem:
for m in elem["members"]:
_members.append(
RelationMember(unicode(m['role']), {
'node': Node,
'way': Way,
'relation': Relation
}[m['type']], long(m['ref'])))
args = [_id, _version, _changeset, _timestamp, _uid, _tags]
if tag == 'node':
args.extend((_lon, _lat))
elif tag == 'way':
args.append(tuple(_nodes))
elif tag == 'relation':
args.append(tuple(_members))
yield _type(*args)