def test_centrality(centrality_f, reverse=False):
node_good = (pnode_good, version_good)
node_bad = (pnode_bad, version_bad)
dg_good, nodes_good = graph.make_graph(good)
name = histogram.get_name(dg_good, node_good)
if reverse:
rank_good = centrality_f(dg_good.reverse(copy=True))
else:
rank_good = centrality_f(dg_good)
counts_good = histogram.aggregate(dg_good, rank_good)
kdes_good = histogram.counts_to_kdes(counts_good)
# REMOVE
print counts_good[name]
print "mean", np.mean(counts_good[name])
print
print
print ">>>>>GOOD ON GOOD"
r_good = rank_good[node_good]
pre_good = histogram.kde_predict(kdes_good[name], rank_good[node_good])
print node_good, name, r_good, pre_good
dg_bad, nodes_bad = graph.make_graph(bad)
name = histogram.get_name(dg_bad, node_bad)
if histogram.SHOULD_HARDCODE_GCC:
name = "/usr/bin/gcc"
print ">>>>>BAD ON GOOD"
if reverse:
rank_bad = centrality_f(dg_bad.reverse(copy=True))
else:
rank_bad = centrality_f(dg_bad)
r_bad = rank_bad[node_bad]
pre_bad = histogram.kde_predict(kdes_good[name], rank_bad[node_bad])
print node_bad, name, r_bad, pre_bad
print ">>>>>COMPARISON"
print "diff (good - bad)", name, "rank diff", r_good - r_bad, "prediction diff", pre_good - pre_bad
def test_centrality(centrality_f, reverse=False):
node_good = (pnode_good, version_good)
node_bad = (pnode_bad, version_bad)
dg_good, nodes_good = graph.make_graph(good)
name = histogram.get_name(dg_good, node_good)
if reverse:
rank_good = centrality_f(dg_good.reverse(copy=True))
else:
rank_good = centrality_f(dg_good)
counts_good = histogram.aggregate(dg_good, rank_good)
kdes_good = histogram.counts_to_kdes(counts_good)
# REMOVE
print counts_good[name]
print "mean", np.mean(counts_good[name])
print
print
print ">>>>>GOOD ON GOOD"
r_good = rank_good[node_good]
pre_good = histogram.kde_predict(kdes_good[name], rank_good[node_good])
print node_good, name, r_good, pre_good
dg_bad, nodes_bad = graph.make_graph(bad)
name = histogram.get_name(dg_bad, node_bad)
if histogram.SHOULD_HARDCODE_GCC:
name = "/usr/bin/gcc"
print ">>>>>BAD ON GOOD"
if reverse:
rank_bad = centrality_f(dg_bad.reverse(copy=True))
else:
rank_bad = centrality_f(dg_bad)
r_bad = rank_bad[node_bad]
pre_bad = histogram.kde_predict(kdes_good[name], rank_bad[node_bad])
print node_bad, name, r_bad, pre_bad
print ">>>>>COMPARISON"
print "diff (good - bad)", name, "rank diff", r_good - r_bad, "prediction diff", pre_good - pre_bad
def test1():
node_good = (pnode_good, version_good)
node_bad = (pnode_bad, version_bad)
dg_good, nodes_good = graph.make_graph(good)
good_kdes = histogram.make_kdes(dg_good)
good_vals = histogram.kde_predict_all(good_kdes, dg_good, node_good)
name = histogram.get_name(dg_good, node_good)
# extract all nodes with same name:
proc_nodes = []
proc_vals = {}
for node_num, t in nodes_good.iteritems():
for v in t:
test_node = (node_num, v)
gnode_name = histogram.get_name(dg_good, test_node)
if gnode_name == name:
proc_nodes.append(test_node)
vals = histogram.kde_predict_all(good_kdes, dg_good, test_node)
proc_vals[test_node] = vals
for (k,v) in sorted(good_kdes[name].items()):
print k, v[1] if v else v
print
print
print ">>>>>GOOD ON GOOD"
for (k,v) in sorted(good_vals.items()):
print k,v
mins = {}
min_nodes = {}
print ">>>>>GOOD ON GOOD (on %d nodes)" % len(proc_nodes)
for node, vals in proc_vals.items():
for (k,v) in sorted(vals.items()):
if not k in mins:
mins[k] = v
min_nodes[k] = node
elif v < mins[k]:
mins[k] = v
min_nodes[k] = node
for (k,v) in sorted(mins.items()):
print k,v,min_nodes[k], histogram.get_vals(dg_good, min_nodes[k])
print
print ">>>>>BAD ON GOOD"
dg_bad, nodes_bad = graph.make_graph(bad)
bad_on_good_vals = histogram.kde_predict_all(good_kdes, dg_bad, node_bad)
for (k,v) in sorted(bad_on_good_vals.items()):
print k,v
print
print ">>>>>COMPARISON"
diffs = {}
for k in good_vals:
if good_vals[k] is not None and bad_on_good_vals[k] is not None:
diffs[k] = good_vals[k] - bad_on_good_vals[k]
else:
diffs[k] = None
for (k,v) in sorted(diffs.items()):
print k,v
def test1():
node_good = (pnode_good, version_good)
node_bad = (pnode_bad, version_bad)
dg_good, nodes_good = graph.make_graph(good)
good_kdes = histogram.make_kdes(dg_good)
good_vals = histogram.kde_predict_all(good_kdes, dg_good, node_good)
name = histogram.get_name(dg_good, node_good)
# extract all nodes with same name:
proc_nodes = []
proc_vals = {}
for node_num, t in nodes_good.iteritems():
for v in t:
test_node = (node_num, v)
gnode_name = histogram.get_name(dg_good, test_node)
if gnode_name == name:
proc_nodes.append(test_node)
vals = histogram.kde_predict_all(good_kdes, dg_good, test_node)
proc_vals[test_node] = vals
for (k, v) in sorted(good_kdes[name].items()):
print k, v[1] if v else v
print
print
print ">>>>>GOOD ON GOOD"
for (k, v) in sorted(good_vals.items()):
print k, v
mins = {}
min_nodes = {}
print ">>>>>GOOD ON GOOD (on %d nodes)" % len(proc_nodes)
for node, vals in proc_vals.items():
for (k, v) in sorted(vals.items()):
if not k in mins:
mins[k] = v
min_nodes[k] = node
elif v < mins[k]:
mins[k] = v
min_nodes[k] = node
for (k, v) in sorted(mins.items()):
print k, v, min_nodes[k], histogram.get_vals(dg_good, min_nodes[k])
print
print ">>>>>BAD ON GOOD"
dg_bad, nodes_bad = graph.make_graph(bad)
bad_on_good_vals = histogram.kde_predict_all(good_kdes, dg_bad, node_bad)
for (k, v) in sorted(bad_on_good_vals.items()):
print k, v
print
print ">>>>>COMPARISON"
diffs = {}
for k in good_vals:
if good_vals[k] is not None and bad_on_good_vals[k] is not None:
diffs[k] = good_vals[k] - bad_on_good_vals[k]
else:
diffs[k] = None
for (k, v) in sorted(diffs.items()):
print k, v