def _main_lozenge_graph(args):
"""Display any lozenge shaped subgraphs
"""
corpus = read_corpus(args, preselected={'stage': ['discourse', 'units']})
output_dir = get_output_dir(args)
keys = [k for k in corpus if k.stage == 'discourse']
loz_count = Counter()
loz_edges = Counter()
for key in sorted(keys):
gra = stacgraph.Graph.from_doc(corpus, key)
if args.strip_cdus:
gra = gra.without_cdus(sloppy=True)
interesting = set()
for node in gra.nodes():
mloz = _maybe_lozenge(gra, node)
if mloz is not None:
l_n, l_e = mloz
loz_count[key] += 1
loz_edges[key] += len(l_e)
interesting |= l_n
gra = gra.copy(interesting)
dot_gra = stacgraph.DotGraph(gra)
if dot_gra.get_nodes():
write_dot_graph(key, output_dir, dot_gra, run_graphviz=args.draw)
for key in sorted(loz_count):
print(key, loz_count[key], '({})'.format(loz_edges[key]))
print('TOTAL lozenges:', sum(loz_count.values()))
print('TOTAL edges in lozenges:', sum(loz_edges.values()))
def generate_graphs(settings):
"""
Draw SVG graphs for each of the documents in the corpus
"""
discourse_only = [k for k in settings.corpus if k.stage == 'discourse']
report = settings.report
# generate dot files
for k in discourse_only:
try:
gra = egr.DotGraph(egr.Graph.from_doc(settings.corpus, k))
dot_file = report.subreport_path(k, '.dot')
create_dirname(dot_file)
if gra.get_nodes():
with codecs.open(dot_file, 'w', encoding='utf-8') as fout:
print(gra.to_string(), file=fout)
except educe.graph.DuplicateIdException:
warning = ("Couldn't graph %s because it has duplicate "
"annotation ids") % dot_file
print(warning, file=sys.stderr)
# attempt to graphviz them
try:
print("Generating graphs... (you can safely ^-C here)",
file=sys.stderr)
for k in discourse_only:
dot_file = report.subreport_path(k, '.dot')
svg_file = report.subreport_path(k, '.svg')
if fp.exists(dot_file) and settings.draw:
subprocess.call('dot -T svg -o %s %s' % (svg_file, dot_file),
shell=True)
except OSError as oops:
print("Couldn't run graphviz. (%s)" % oops, file=sys.stderr)
print("You should install it for easier sanity check debugging.",
file=sys.stderr)
def main(args):
"""
Subcommand main.
You shouldn't need to call this yourself if you're using
`config_argparser`
"""
args.stage = 'discourse|units'
corpus = read_corpus(args, verbose=True)
output_dir = get_output_dir(args)
keys = [k for k in corpus if k.stage == 'discourse']
for k in sorted(keys):
try:
gra = stacgraph.Graph.from_doc(corpus,
k,
pred=_keep(corpus[k],
args.rel_types))
dot_gra = stacgraph.DotGraph(gra)
if dot_gra.get_nodes():
write_dot_graph(k, output_dir, dot_gra, run_graphviz=args.draw)
else:
print("Skipping %s (empty graph)" % k, file=sys.stderr)
except graph.DuplicateIdException:
warning = "WARNING: %s has duplicate annotation ids" % k
print(warning, file=sys.stderr)
def dump_graph(dump_filename, graph):
"""
Write a dot graph and possibly run graphviz on it
"""
dot_graph = stac_gr.DotGraph(graph)
dot_file = dump_filename + '.dot'
svg_file = dump_filename + '.svg'
mk_parent_dirs(dot_file)
with codecs.open(dot_file, 'w', encoding='utf-8') as dotf:
print(dot_graph.to_string(), file=dotf)
print("Creating %s" % svg_file, file=sys.stderr)
subprocess.call('dot -T svg -o %s %s' % (svg_file, dot_file), shell=True)
def _main_rel_graph(args):
"""
Draw graphs showing relation instances between EDUs
"""
args.stage = 'discourse|units'
corpus = _read_corpus(args)
output_dir = get_output_dir(args)
if args.live:
keys = corpus
else:
keys = [k for k in corpus if k.stage == 'discourse']
for k in sorted(keys):
if args.highlight:
highlights = [anno_id_from_tuple(x) for x in args.highlight]
for anno in corpus[k].annotations():
if anno.local_id() in highlights:
anno.features['highlight'] = 'orange'
try:
gra = stacgraph.Graph.from_doc(corpus, k)
if args.strip_cdus:
gra = gra.without_cdus(mode=args.strip_mode)
dot_gra = stacgraph.DotGraph(gra)
if dot_gra.get_nodes():
write_dot_graph(k, output_dir, dot_gra, run_graphviz=args.draw)
if args.split:
ccs = gra.connected_components()
for part, nodes in enumerate(ccs, 1):
gra2 = gra.copy(nodes)
write_dot_graph(k,
output_dir,
stacgraph.DotGraph(gra2),
part=part,
run_graphviz=args.draw)
else:
print("Skipping %s (empty graph)" % k, file=sys.stderr)
except graph.DuplicateIdException:
warning = "WARNING: %s has duplicate annotation ids" % k
print(warning, file=sys.stderr)
def _main_rfc_graph(args):
"""
Draw graphs showing relation instances between EDUs
"""
args.stage = 'discourse|units'
corpus = _read_corpus(args)
output_dir = get_output_dir(args)
if args.rfc == 'basic':
mk_rfc = BasicRfc
elif args.rfc == 'mlast':
mk_rfc = ThreadedRfc
else:
raise NotImplementedError
if args.live:
keys = corpus
else:
keys = [k for k in corpus if k.stage == 'discourse']
for key in sorted(keys):
gra = stacgraph.Graph.from_doc(corpus, key)
for subgra_nodes in gra.connected_components():
subgra = gra.copy(subgra_nodes)
sub_rfc = mk_rfc(subgra)
for node in sub_rfc.frontier():
gra.annotation(node).features['highlight'] = 'green'
rfc = mk_rfc(gra)
for link in rfc.violations():
gra.annotation(link).features['highlight'] = 'red'
dot_gra = stacgraph.DotGraph(gra)
if dot_gra.get_nodes():
write_dot_graph(key, output_dir, dot_gra, run_graphviz=args.draw)
else:
print("Skipping %s (empty graph)" % key, file=sys.stderr)