def read_pajek(path, encoding='UTF-8', project=False, auto_table=False):
"""Reimplemented method for reading Pajek files; written in
C++ for maximum performance.
:param path: File or file name to write.
:type path: string
:param encoding: Encoding of input text file, default 'UTF-8'.
:type encoding: string
:param project: Determines whether the input file is a Pajek project file,
possibly containing multiple networks and other data. If :obj:`True`,
a list of networks is returned instead of just a network. Default is
:obj:`False`.
:type project: boolean.
Return the network (or a list of networks if project=:obj:`True`) of type
:obj:`Orange.network.Graph` or :obj:`Orange.network.DiGraph`.
Examples
>>> G=Orange.network.nx.path_graph(4)
>>> Orange.network.readwrite.write_pajek(G, "test.net")
>>> G=Orange.network.readwrite.read_pajek("test.net")
To create a Graph instead of a MultiGraph use
>>> G1=Orange.network.Graph(G)
References
See http://vlado.fmf.uni-lj.si/pub/networks/pajek/doc/draweps.htm
for format information.
"""
path = _check_network_dir(path)
G = _wrap(rwpajek.read_pajek(path))
# Additionally read values into Table; needed to get G nodes properly sorted
# (Consult OWNxFile.readDataFile(), orangeom.GraphLayout.readPajek(), and the Pajek format spec)
import shlex, numpy as np
rows, metas, remapping = [], [], {}
with open(path) as f:
for line in f:
if line.lower().startswith('*vertices'):
nvertices = int(line.split()[1])
break
# Read vertices lines
for line in f:
parts = shlex.split(line)[:4]
if len(parts) == 1:
i = label = parts[0]
elif len(parts) == 2:
i, label = parts
metas.append((label,))
elif len(parts) == 4:
i, label, x, y = parts
x, y = float(x), float(y)
rows.append((x, y))
metas.append((label,))
i = int(i) - 1 # -1 because pajek is 1-indexed
remapping[label] = i
nvertices -= 1
if not nvertices: break
from Orange.data import Domain, Table, ContinuousVariable, StringVariable
# Construct x-y-label table (added in OWNxFile.readDataFile())
table = None
if rows:
domain = Domain([ContinuousVariable('x'), ContinuousVariable('y')],
metas=[StringVariable('label')])
table = Table.from_numpy(domain, np.array(rows, dtype=float),
metas=np.array(metas, dtype=str))
elif metas:
domain = Domain([], metas=[StringVariable('label')])
table = Table.from_numpy(domain, np.zeros((len(metas), 0)),
metas=np.array(metas, dtype=str))
if table is not None and auto_table:
G.set_items(table)
# Relabel nodes to integers, sorted by appearance
for node in G.node:
G.node[node]['label'] = node
nx.relabel_nodes(G, remapping, copy=False)
assert not table or len(table) == G.number_of_nodes(), 'There was a bug in NetworkX. Please update to git if need be'
return G
def read_pajek(path, encoding='UTF-8', project=False, auto_table=False):
"""Reimplemented method for reading Pajek files; written in
C++ for maximum performance.
:param path: File or file name to write.
:type path: string
:param encoding: Encoding of input text file, default 'UTF-8'.
:type encoding: string
:param project: Determines whether the input file is a Pajek project file,
possibly containing multiple networks and other data. If :obj:`True`,
a list of networks is returned instead of just a network. Default is
:obj:`False`.
:type project: boolean.
Return the network (or a list of networks if project=:obj:`True`) of type
:obj:`Orange.network.Graph` or :obj:`Orange.network.DiGraph`.
Examples
>>> G = orangecontrib..network.nx.path_graph(4)
>>> orangecontrib..network.readwrite.write_pajek(G, "test.net")
>>> G = orangecontrib.network.readwrite.read_pajek("test.net")
To create a Graph instead of a MultiGraph use
>>> G1 = orangecontrib.network.Graph(G)
References
See http://vlado.fmf.uni-lj.si/pub/networks/pajek/doc/draweps.htm
for format information.
"""
path = _check_network_dir(path)
G = _wrap(rwpajek.read_pajek(path))
# Additionally read values into Table; needed to get G nodes properly sorted
# (Consult OWNxFile.readDataFile(), orangeom.GraphLayout.readPajek(), and the Pajek format spec)
import shlex, numpy as np
rows, metas, remapping = [], [], {}
with open(path, encoding='utf-8') as f:
for line in f:
if line.lower().startswith('*vertices'):
nvertices = int(line.split()[1])
break
# Read vertices lines
for line in f:
parts = shlex.split(line)[:4]
if len(parts) == 1:
i = label = parts[0]
elif len(parts) == 2:
i, label = parts
metas.append((label, ))
elif len(parts) == 4:
i, label, x, y = parts
# The format specification was never set in stone, it seems
try:
x, y = float(x), float(y)
except ValueError:
metas.append((label, x, y))
else:
rows.append((x, y))
metas.append((label, ))
i = int(i) - 1 # -1 because pajek is 1-indexed
remapping[label] = i
nvertices -= 1
if not nvertices: break
from Orange.data import Domain, ContinuousVariable, StringVariable
# Construct x-y-label table (added in OWNxFile.readDataFile())
table = None
vars = [ContinuousVariable('x'), ContinuousVariable('y')] if rows else []
meta_vars = [
StringVariable('label ' + str(i))
for i in range(len(metas[0]) if metas else 0)
]
if rows or metas:
domain = Domain(vars, metas=meta_vars)
table = Table.from_numpy(domain,
np.array(rows, dtype=float).reshape(
len(metas),
len(rows[0]) if rows else 0),
metas=np.array(metas, dtype=str))
if table is not None and auto_table:
G.set_items(table)
# Relabel nodes to integers, sorted by appearance
for node in G.node:
G.node[node]['label'] = node
nx.relabel_nodes(G, remapping, copy=False)
if table is not None and len(table) != G.number_of_nodes():
raise PajekBug(
"There is a bug in your version of NetworkX reading Pajek files. "
"Please update your NetworkX installation.")
return G
def load(graph_path):
return pajek.read_pajek(graph_path)
def read_pajek(path, encoding='UTF-8', project=False, auto_table=False):
"""Reimplemented method for reading Pajek files; written in
C++ for maximum performance.
:param path: File or file name to write.
:type path: string
:param encoding: Encoding of input text file, default 'UTF-8'.
:type encoding: string
:param project: Determines whether the input file is a Pajek project file,
possibly containing multiple networks and other data. If :obj:`True`,
a list of networks is returned instead of just a network. Default is
:obj:`False`.
:type project: boolean.
Return the network (or a list of networks if project=:obj:`True`) of type
:obj:`Orange.network.Graph` or :obj:`Orange.network.DiGraph`.
Examples
>>> G=Orange.network.nx.path_graph(4)
>>> Orange.network.readwrite.write_pajek(G, "test.net")
>>> G=Orange.network.readwrite.read_pajek("test.net")
To create a Graph instead of a MultiGraph use
>>> G1=Orange.network.Graph(G)
References
See http://vlado.fmf.uni-lj.si/pub/networks/pajek/doc/draweps.htm
for format information.
"""
path = _check_network_dir(path)
G = _wrap(rwpajek.read_pajek(path))
# Additionally read values into Table; needed to get G nodes properly sorted
# (Consult OWNxFile.readDataFile(), orangeom.GraphLayout.readPajek(), and the Pajek format spec)
import shlex, numpy as np
rows, remapping = [], {}
with open(path) as f:
for line in f:
if line.lower().startswith('*vertices'):
nvertices = int(line.split()[1])
break
# Read vertices lines
for line in f:
i, label, x, y = shlex.split(line)[:4]
i, x, y = int(i) - 1, float(x), float(
y) # -1 because pajek is 1-indexed
remapping[label] = i
rows.append((x, y))
nvertices -= 1
if not nvertices: break
# Construct x-y table (added in OWNxFile.readDataFile())
domain = Orange.data.Domain([
Orange.data.ContinuousVariable('x'),
Orange.data.ContinuousVariable('y')
])
table = Orange.data.Table.from_numpy(domain, np.array(rows, dtype=float))
G.set_items(table)
# Relabel nodes to integers, sorted by appearance
for node in G.node:
G.node[node]['label'] = node
nx.relabel_nodes(G, remapping, copy=False)
assert len(table) == G.number_of_nodes(
), 'There was a bug in NetworkX. Please update to git if need be'
return G