"""Functions to convert tabular data to popular python graph structures
"""
import os
[docs]def to_igraph(vdf, edf, pmax, comms=None, peak=None, save_dir=None, ext='gml'):
"""Convert Data from (Geo)DataFrames to python-igraph's Graph class
Each commodity gets its own graph.
Weights are derived from built capacity.
Parameters
----------
vdf : [Geo]DataFrame
Holding Vertex Data id=Vertex
and Commodity Sources as columns
edf : [Geo]DataFrame
Holding (V1,V2) Multi-indexed Edge data
To be sure, that all edges are created.
pmax : DataFrame
Commodities as columns with max capacity per edge
returned by rivus.get_constants()
comms : iterable, optional
Names of the commodities from which we
build the graphs. (Each as separate graph.) If omitted, the columns
of pmax will be used.
peak : DataFrame, optional
Commodities as columns with demands
in t_peak time-step. Calculated in main.rivus
save_dir : path string, optional
Path to a dir to save graphs as `ext`
Path preferably constructed using the os.path module.
If dir does not exit yet, it will be created.
ext : str, optional
Description
ext (string) file extension, supported by igraph.save()
If not one of the following, the default 'gml' will be applied.
'adjacency', 'dimacs', 'dot', 'graphviz', 'edgelist', 'edges',
'edge', 'gml', 'graphml', 'graphmlz', 'gw', 'leda', 'lgl', 'lgr',
'ncol', 'net', 'pajek', 'pickle', 'picklez', 'svg'
Returns
-------
list
List of igraph.Graph objects in order of ``comms``.
Graphs are undirected and weighted.
Example
-------
::
_, pmax, _, _ = get_constants(prob)
graphs = to_igraph(vertex, edge, pmax, ['Gas', 'Heat'])
"""
import igraph as ig
ext_list = ['adjacency', 'dimacs', 'dot', 'graphviz', 'edgelist', 'edges',
'edge', 'gml', 'graphml', 'graphmlz', 'gw', 'leda', 'lgl',
'lgr', 'ncol', 'net', 'pajek', 'pickle', 'picklez', 'svg']
if ext not in ext_list:
ext = 'gml'
comms = pmax.columns.values if comms is None else comms
if len(edf) != len(pmax):
# To have all rows, also the ones, where nothing was built.
pmax = edf.join(pmax).fillna(0)
graphs = []
for comm in comms:
# Graph can be created from vertices, because no isolated node is
# in the input graph.
g = ig.Graph(pmax[comm].index.values.tolist())
g['Name'] = '{} capacity graph'.format(comm.upper())
g['Commodity'] = comm
g.vs['Label'] = vdf.index.values.tolist()
g.vs[comm] = vdf[comm].tolist() if comm in vdf else [0, ] * len(vdf)
g.es['Label'] = list(map(lambda v1v2: '{}-{}'.format(*v1v2),
edf.index.values))
g.es[comm] = pmax[comm].tolist()
cap_max = pmax[comm].max()
if cap_max == 0:
weights = [0] * len(pmax)
else:
weights = pmax[comm] / cap_max
weights = weights.tolist()
g.es['Weight'] = weights # Camel case for Gephi
g.es['weight'] = weights
if peak is not None:
g.es[comm + '-peak'] = peak[comm].tolist()
# For possible GeoLayout (e.g. in Gephi)
g.vs['Longitude'] = vdf.geometry.map(lambda p: p.x).tolist()
g.vs['Latitude'] = vdf.geometry.map(lambda p: p.y).tolist()
# Delete demand aedges, what are not built in optimum
g.delete_edges(g.es.select(weight_eq=0))
graphs.append(g)
if save_dir:
if not os.path.isdir(save_dir):
os.makedirs(save_dir)
for graph in graphs:
gpath = os.path.join(save_dir, '{}.{}'.format(graph['Name']), ext)
with open(gpath, 'w') as fhandle:
graph.save(fhandle, ext)
return graphs
[docs]def to_nx(vdf, edf, pmax, comms=None, save_dir=None):
"""Convert to networkx graph representation
Parameters
----------
vdf : [Geo]DataFrame
Holding Vertex Data id=Vertex
and Commodity Sources as columns
edf : [Geo]DataFrame
Holding (V1,V2) Multi-indexed Edge data
To be sure, that all edges are created.
pmax : DataFrame
Commodities as columns with max capacity per edge
returned by rivus.get_constants()
comms : iterable, optional
Names of the commodities from which we
build the graphs. (Each as separate graph.) If omitted, the columns
of pmax will be used.
save_dir : path string, optional
Path to a dir to save graphs as GML.
Path preferably constructed using the `os.path` module
If dir does not exit yet, it will be created.
Returns
-------
list
nx_graph objects in accordance with input `comms` or all commodities
found in pmax.columns
Example
-------
::
_, pmax, _, _ = get_constants(prob)
graphs = to_nx(vertex, edge, pmax, ['Gas', 'Heat'])
Note
----
``nx.from_pandas_dataframe()`` was also investigated for conversion,
but it is a bit slower and does not improve code quality in my opinion.
"""
import networkx as nx
comms = pmax.columns.values if comms is None else comms
if len(edf) != len(pmax):
# To have all rows, also the ones, where nothing was built.
_pmax = edf.join(pmax).fillna(0)
else:
_pmax = pmax.copy()
graphs = []
for comm in comms:
g = nx.Graph(Name='{} capacity graph'.format(comm.upper()),
Commodity=comm)
g.add_nodes_from(vdf.index.values.tolist())
for x, row in vdf.iterrows():
g.node[x]['Label'] = x
g.node[x][comm] = row[comm] if comm in row else 0
g.node[x]['Longitude'] = row.geometry.x
g.node[x]['Latitude'] = row.geometry.y
cap_max = _pmax[comm].max()
if cap_max != 0:
for v1v2, row in _pmax.iterrows():
if row[comm] == 0:
# do not connect verticies in the graph,
# if the edge is not built in optimum
continue
this_weight = row[comm] / cap_max
this_label = '{}-{}'.format(*v1v2)
g.add_edge(*v1v2, Label=this_label, Commodity=row[comm],
Weight=this_weight)
graphs.append(g)
if save_dir:
ext = 'gml' # TODO networkx has different functions for each format...
if not os.path.isdir(save_dir):
os.makedirs(save_dir)
for graph in graphs:
gpath = os.path.join(save_dir, '{}.{}'.format(graph['Name']), ext)
nx.write_gml(graph, gpath)
return graphs