from sequence_utils import trim_inactive
from build_models import filter_criteria, log_series
import sys, cPickle
if __name__ == "__main__":
inpath = sys.argv[1]
goodpath = sys.argv[2]
rejpath = sys.argv[3]
good_records = []
rej_records = []
with open(inpath) as datafile:
data = cPickle.load(datafile)
window_size = data['window_size']
records = data['records']
for record in records:
trimmed = trim_inactive(record['relays_out'])
new_rec = {
'ident': record['ident'],
'create': record['create'],
'destroy': record['destroy'],
'relays_in': None,
'relays_out': trimmed,
}
if filter_criteria(log_series(trimmed)):
good_records.append(new_rec)
elif len(trimmed) > 0:
rej_records.append(new_rec)
n_gone = len(records) - len(good_records) - len(rej_records)
print "%i good records" % len(good_records)
print "%i reject records" % len(rej_records)
print "%i len 0 after trimming" % n_gone
def preprocess(series): return map(lambda s: log_series(trim_inactive(s)), series)
def do_summarize(records, direc_key):
"""
Display summary histograms for the series in records.
@param records: the circuit records
@param direc_key: 'relays_in' for incoming relays, 'relays_out' for
outgoing
"""
circ_len_aggr = []
mean_cells_per_window_aggr = []
min_cells_per_window_aggr = []
max_cells_per_window_aggr = []
median_cells_per_window_aggr = []
stddev_cells_per_window_aggr = []
inst_counts_aggr = []
unique_vals_aggr = []
percent_active_aggr = []
time_active_aggr = []
for record in records:
relays = record[direc_key]
circ_len_aggr.append((record['destroy'] - record['create'])/1000.0)
mean_cells_per_window_aggr.append(1.0*sum(relays)/len(relays))
median_cells_per_window_aggr.append(median(relays))
min_cells_per_window_aggr.append(min(relays))
max_cells_per_window_aggr.append(max(relays))
stddev_cells_per_window_aggr.append(std(relays))
inst_counts_aggr += relays
# unique_vals_aggr.append(len(set(filter(lambda o: o > 2, relays))))
time_active = len(trim_inactive(relays))
percent_active_aggr.append(100.0*time_active/len(relays))
# time_active_aggr.append(time_active)
fig = plt.figure()
summarize(max_cells_per_window_aggr, "Max")
meansplot = fig.add_subplot(421)
plt.title("Mean Cells/Window")
plt.xlabel("Mean Cells/Window")
plt.ylabel("Frequency")
plt.yscale('log')
meansplot.hist(mean_cells_per_window_aggr, bins=N_HIST_BINS)
cellsplot = fig.add_subplot(422)
plt.title("Median Cells/Window")
plt.xlabel("Median Cells/Window")
plt.ylabel("Frequency")
plt.yscale('log')
cellsplot.hist(median_cells_per_window_aggr, bins=N_HIST_BINS)
minsplot = fig.add_subplot(423)
plt.title("Min Cells/Window")
plt.xlabel("Min Cells/Window")
plt.ylabel("Frequency")
plt.yscale('log')
minsplot.hist(min_cells_per_window_aggr, bins=N_HIST_BINS)
maxsplot = fig.add_subplot(424)
plt.title("Max Cells/Window")
plt.xlabel("Max Cells/Window")
plt.ylabel("Frequency")
plt.yscale('log')
maxsplot.hist(max_cells_per_window_aggr, bins=N_HIST_BINS)
stddevsplot = fig.add_subplot(425)
plt.title("Std Dev. of Cells/Window")
plt.xlabel("Std Dev. of Cells/Window")
plt.ylabel("Frequency")
plt.yscale('log')
stddevsplot.hist(stddev_cells_per_window_aggr, bins=N_HIST_BINS)
cellsplot = fig.add_subplot(426)
plt.title("Single Window Cell Count")
plt.xlabel("Single Window Cell Count")
plt.ylabel("Frequency")
plt.yscale('log')
cellsplot.hist(inst_counts_aggr, bins=N_HIST_BINS)
lenplot = fig.add_subplot(427)
plt.title("Circuit Length (seconds)")
plt.xlabel("Circuit Length (seconds)")
plt.ylabel("Frequency")
plt.yscale('log')
lenplot.hist(circ_len_aggr, bins=N_HIST_BINS)
# uniqueplot = fig.add_subplot(338)
# plt.title("Number of Unique Values > 1")
# plt.xlabel("Number of Unique Values > 1")
# plt.ylabel("Frequency")
# uniqueplot.hist(unique_vals_aggr, bins=N_HIST_BINS)
# timeactiveplot = fig.add_subplot(428)
# plt.title("Percent of Time in Active State")
# plt.xlabel("Percent of Time")
# plt.ylabel("Frequency")
# timeactiveplot.hist(percent_active_aggr, bins=N_HIST_BINS)
fig.tight_layout()
